INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX REFERENCES

In oxygenic photosynthetic organisms, photosystem I (PSI) catalyzes the light-driven electron transport between soluble electron carriers, from cytochrome c₆ at the lumenal side to ferredoxin at the stromal (cytoplasmic) side of the thylakoid membrane. In case of iron deficiency, flavodoxin can also act as a soluble electron carrier at the acceptor side of PSI (Rogers, 1987). Ferredoxin functions as the soluble electron donor to ferredoxin-NADP⁺ reductase (FNR), which catalyzes the reduction of NADP⁺ to NADPH in chloroplasts and cyanobacteria. Furthermore, ferredoxin is involved in the assimilation of nitrogen and sulfur and in the regulation of carbon assimilation (Knaff, 1996). Therefore, the complex of PSI with ferredoxin can be regarded as a model system for the interaction of electron transfer proteins.

Very recently, the structure of PSI was solved at 2.5 Å resolution (Jordan et al., 2001). From this structure, previous evidence that all of the three membrane extrinsic subunits PsaC, PsaD, and PsaE are involved in the interaction of PSI with ferredoxin was strongly supported. Modeling of the interaction of ferredoxin from Spirulina platensis (Tsukihara et al., 1981) to the 6 Å electron density map of PSI from Synechococcus elongatus (Krauss et al., 1993) led to the suggestion of a binding site of ferredoxin, which is located close to the terminal 4Fe-4S cluster of PSI (Fromme et al., 1994). The distance between the 4Fe-4S cluster of PSI to the 2Fe-2S cluster of ferredoxin was estimated to be 14 Å (center-to-center distance). This would lead to an edge-to-edge distance of 11-12 Å. This distance is in reasonable agreement with the fastest kinetics of the electron transport from PSI to ferredoxin, which was determined to exhibit a halftime of 500 ns (Sétif and Bottin, 1994, 1995). The same binding site was found by electron microscopy of cross-linked complexes of PSI with either ferredoxin (Lelong et al., 1996) or flavodoxin (Mühlenhoff et al., 1996). At this binding pocket, ferredoxin would get in contact with all three stromal subunits, which is in good agreement with the functional studies (reviewed in Sétif, 2001).

Nevertheless, information concerning the docking site at the atomic level, including the orientation of ferredoxin and the identification of the specific interactions between the partners, is still missing. To get further knowledge in the interaction of ferredoxin with PSI, we started experiments on the cocrystallization of these two proteins. In this report we describe the preliminary characterization of the PSI/ferredoxin cocrystal and provide evidence for a functional 1:1 PSI/ferredoxin complex in the cocrystals. Moreover, the EPR characterization of the F g-tensor in the cocrystals was performed and was used for deducing the orientation of the C₃ axis of the trimer relative to the unit cell axes. This information was used for deriving the characteristics of the F g-tensor.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Isolation of PSI and ferredoxin

PSI from the thermophilic cyanobacterium S. elongatus was isolated in the presence of the stereochemically pure detergent -dodecylmaltoside as described elsewhere (Fromme and Witt, 1998). Ferredoxin used in this study was isolated from the cyanobacterium Synechocystis sp. PCC 6803 as described in Bottin and Lagoutte (1992). The rates of photoreduction and the properties of PSI binding of this ferredoxin were found to be very similar for PSI from either S. elongatus or Synechocystis 6803 (Sétif and Bottin, 1994).

Preparation of cocrystals for EPR measurements

For single-crystal studies, a cocrystal within a drop of PEG buffer (43% PEG (w/v) in 100 mM HEPES pH 7.5, 150 mM CaCl₂, 0.02% -DM) was gently deposited onto the holder surface (bottom plane of Fig. 1). The PEG 400 concentration was increased from 14% (concentration in the crystallization buffer) up to 43%, to ensure a conservation of the cocrystals during the freezing process needed for performing low temperature EPR. This was done in several steps by increasing the PEG amount by 4% increments with a 5 min incubation at each intermediate PEG concentration. The cocrystals were kept for at least 10 min at room temperature with the final 43% PEG concentration before freezing. Different samples were prepared at various pH values (between 6.0 and 9.2) and with different reductants (sodium ascorbate or dithionite) and redox mediators (2,6-dichlorophenol-indophenol (DCPIP), phenazine methosulfate (PMS), methyl viologen), which were added at the final stage (43% PEG). Incubation in dim room light or in darkness was performed for times as long as several hours. For freezing, cocrystals were dipped into liquid nitrogen. There was no apparent degradation of the cocrystals during these treatments. Incubation of cocrystals at 200 K was performed outside the EPR cavity in a nitrogen gas flow system (BVT-3000 from Bruker, Wissembourg, France). A detailed EPR investigation was performed for two different cocrystals, which were treated as follows: for the first one, designated 1, 10 mM sodium ascorbate and 0.5 mM DCPIP were added to the final solution (43% PEG, pH 7.5). The cocrystal was then incubated in darkness for 10 min before freezing under dim room light; for the second one, 2, 10 mM sodium dithionite and 0.5 mM PMS were added to the final solution (43% PEG, pH 6.5). The cocrystal was then incubated in darkness for 8 min in a vessel filled with argon gas before freezing in darkness.

View larger version (20K): [in this window] [in a new window]   FIGURE 1   Scheme of the crystal holder used for changing the orientation of the cocrystals in two orthogonal directions. The 1 angle was changed with a goniometer by steps of 5° and the 1 angle was adjusted manually, with a precision of ~5°, under a binocular microscope (20× magnification) while the crystal holder was maintained above a bath of liquid nitrogen. The holder was made of Plexiglas and its upper part was adapted to a standard quartz EPR tube.

The unit cell axes were difficult to correlate to the morphological shape of the EPR-studied large cocrystals. The crystals displayed many facets and did not exhibit either an edge or a face that was significantly larger than the others, so they were first placed randomly in the special holder. After mounting, the holder was rotated by 180° in 5° steps and EPR spectra were recorded to find an orientation where the crystal could be rotated around one of the crystallographic axes.

Powder-spectra of cocrystals were obtained using a large number (>50) of small and randomly oriented cocrystals. Samples were prepared in the crystallization medium (150 mM CaCl₂, 14% PEG 400, 0.02% -DM) with addition of redox mediators. Incubation was performed at room temperature in dim room light under argon flow. Cocrystals were then rapidly sedimented by a brief centrifugation (1 min) in a small quartz tube and frozen by dipping into liquid nitrogen.

EPR measurements

EPR spectra were recorded between 10 and 45 K with an ESR300D X-band spectrometer (Bruker), using a TE₁₀₂ resonator equipped with a front grid for sample illumination within the cavity. Samples were illuminated within the cavity using a tungsten-halogen lamp (800 W). White light was filtered with an infrared filter (Calflex) and a water cuvette. Light was focused onto the cavity grid by a Plexiglas lightpipe. The temperature was controlled with an ESR9 helium cryostat (Oxford Instruments, Oxon, UK). Illumination outside the EPR cavity was performed at 200-240 K in a nitrogen gas flow system (BVT-3000 from Bruker) and the same lamp as above as the light source. For spin quantitation, double integration of the spectra was performed after subtraction of the radical signal in the g_{free electron} region, when present. In the case of a light-induced spectrum relative to a single orientation of a single cocrystal, the quasi-isotropic radical signal from P700⁺ was subtracted after measuring it from the average spectrum of a whole series of ₁ angles (from 0 to 175°).

The following procedure was used for estimating the PSI/ferredoxin stoichiometry in the cocrystals: after its EPR study, an EPR sample containing many small cocrystals was dissolved. Part of the sample was studied by flash-absorption spectroscopy. Measurements were performed at 820 nm, from which the P700⁺ concentration (=[PSI]) was estimated, assuming an absorption coefficient of 6500 M¹ cm¹ (Mathis and Sétif, 1981). Another part of the sample was used for preparing an EPR tube under conditions (pH 9.0 + dithionite), which allowed ferredoxin to be fully reduced. This was compared to a reference sample containing a known amount of ferredoxin from Synechocystis 6803. This amount was calculated by assuming an absorption coefficient of 9500 M¹ cm¹ for oxidized ferredoxin at 423 nm (Tagawa and Arnon, 1968; Fee and Palmer, 1971). The ferredoxin sample was reduced with an excess of sodium dithionite at pH 8.7. For a quantitative EPR comparison of the ferredoxin contents of the samples, spectra were recorded at 45 K under conditions of nonsaturating microwave power. At this temperature, the ferredoxin signal could be observed in the PSI/ferredoxin sample with little interference from F and F as the EPR spectra of these last species are severely broadened due to fast spin-lattice relaxation (see Lelong et al., 1994) for a similar quantitation).

For measurements on single cocrystals, the crystal holders were made of Plexiglas (Atecplast, Bezons, France). Their orientation could be changed in two orthogonal directions (Fig. 1): the ₁ angle was changed with a goniometer by 5° steps from 0 to 180°. For measuring light-induced spectra, a series of baselines was recorded before illumination by varying the ₁ angle and spectra after illumination were recorded later on at the same angles. The angle ₁ was changed manually, with a precision of ~5°, under a binocular microscope (20× magnification) while the crystal holder was maintained above a bath of liquid nitrogen. Complete sets of data were recorded for two different cocrystals, but data for only one of these will be shown in the present report. For both cocrystals, P700 was initially reduced and the iron-sulfur centers were initially oxidized before freezing in darkness, as checked by the almost absence of EPR signal before illumination (only a very weak radical was observable). An EPR baseline was then recorded in darkness at 10-20 K at different ₁ orientations before 10 min illumination within the cavity. Illumination was performed while rotating the sample holder for improving the homogeneity of illumination. A longer period of illumination did not increase the size of the light-induced spectra, thus indicating that all PSIs in the cocrystals were photoexcited during the 10-min period. EPR spectra were then recorded at the different ₁ orientations and light-induced spectra were obtained after subtraction of the baseline. A dataset was thus recorded by varying ₁ with the goniometer from 0 to 180° by 5° steps. Similar datasets were recorded at different ₁ angles to get a complete orientation pattern. For doing this, it was necessary to incubate the crystals between two consecutive datasets at 200 K in darkness for periods of ~30 min, which resulted in a complete recombination between P700⁺ and (F_A, F_B), as checked by EPR. The temperature of a cocrystal was therefore changed many times between 10 and 20 K, 77 K (for intermediate storage), and 200 K without any observable induced damage as checked by the absence of change in the light-induced EPR spectrum at a given orientation. It appears therefore that the PSI/ferredoxin cocrystals are much more resistant than the PSI crystals (further named PSI_a for PSI-alone crystals to discriminate them from the PSI/ferredoxin cocrystals). This allows various redox reagents and light treatments to be used for getting a variety of redox states that may be useful for future studies by x-ray crystallography and EPR at low temperature.

EPR simulations were performed using Mathcad (V. 8.0, Mathsoft, Inc., Cambridge, MA). The systematic parameter search was performed using Borland Pascal programs for Windows (V. 7.0, Borland International, Inc., Scotts Valley, CA). All programs were home-written. Fitting with Gaussian components was performed with the software Origin (V. 6.0, Microcal Software, Inc., Northampton, MA).

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Crystallization of the PSI/ferredoxin complex

A large screening of different crystallization conditions was used for testing the stability and the solubility of the intact PSI/ferredoxin complex. In the following, we describe the main parameters that influence the solubility of the complex.

Ionic strength

Crystallization agents

View larger version (67K): [in this window] [in a new window]   FIGURE 2   Cocrystals of photosystem I with ferredoxin and diffraction pattern. (A) Picture of the PSI/ferredoxin cocrystals. The crystals were grown by vapor diffusion: protein-solution (3 µl): 100 µM PSI (P700), 125 µM ferredoxin, 22 mM HEPES, pH 7.5, 50 mM CaCl2, 0.013% -dodecylmaltoside, 4.7% PEG 400; reservoir (900 µl): 0.1 M HEPES pH 7.5, 0.15 M CaCl2, 0.013% -dodecylmaltoside, 14% PEG 400. The crystals shown herein have a size of 0.3 mm. (B) Diffraction pattern of photosystem I/cocrystals obtained at beamline ID14-1 at ESRF in Grenoble; wavelength 1 Å. The resolution is indicated by the circles.

pH dependence

Temperature dependence

Degree of supersaturation/homogeneous/heterogeneous nucleation

Stoichiometry of photosystem I and ferredoxin

Cocrystal characterization

X-ray diffraction data were first collected at room temperature at beamline SRS 9.6 at the Synchrotron in Daresbury, UK, and more recently at beamline ID14 at ESRF in Grenoble. The crystals diffract x-rays to a resolution of 7 Å (see Fig. 2 B); However, the crystals show decrease of diffraction quality during x-ray exposure, so that the data could be evaluated only to a resolution of 9 Å. Preliminary data evaluation revealed a space group of P2₁2₁2₁ with unit cell dimension of a = 194 Å, b = 208 Å, and c = 354 Å. However, due to the high mosaic spread of the crystals, the space group could not be determined unambiguously so that the P2₁2₁2 and P2₁22 space groups could not be excluded. The orthorhombic space group of the crystals implies that each unit cell contains four trimers. The solvent content was estimated to be 64%, which is much smaller than the solvent content of the PSI_a crystals of 78% (Fromme and Witt, 1998).

Powder-type EPR spectra of the cocrystals

Fig. 3 exhibits powder-type EPR spectra of cocrystals, which were obtained from two different EPR tubes containing a large number of small cocrystals. Both samples were prepared under highly reducing conditions (an excess of sodium dithionite and methyl viologen) as described in Materials and Methods and were frozen under dim room light. Spectra A and B were recorded at 8 and 45 K, respectively, from a sample prepared at pH 9.2 and which was not further illuminated. Spectrum B exhibits g-values of ~1.885, 1.955, and 2.05, which can be ascribed to the reduced [2Fe-2S] cluster of ferredoxin. F and F are expected to be hardly visible at this temperature due to severe line-broadening (see, e.g., Lelong et al., 1994). The spectrum of a frozen solution of ferredoxin recorded under similar conditions is indistinguishable from spectrum B. In spectrum A, the main lines are observed at g-values of ~1.86, 1.89, 1.945, and 2.05. These signals correspond to the reduced forms of the [4Fe-4S] clusters of PSI: the g-value at 1.86 is due solely to F, whereas F, F, and (F, F) contribute to various extents to the other lines (1.89, 1.945, and 2.05). Under the EPR conditions used for measuring this spectrum (8 K and 80 mW of microwave power), only a small signal is observed at g  1.96, indicating that the ferredoxin signal is highly saturated. When the spectrum is recorded at 15 K, a distinct shoulder is observable at g  2.064 (not shown), which can be ascribed to F. This signal is not visible in spectrum A presumably because of microwave power saturation.

View larger version (22K): [in this window] [in a new window]   FIGURE 3   Powder-type EPR spectra of cocrystals obtained from samples containing a large number (>50) of small cocrystals. Spectra A and B were obtained with a sample prepared in the crystallization medium at pH 9.2, with 15 mM sodium dithionite and 200 µM methyl viologen. The sample was incubated at room temperature for 5 min in the presence of argon and was frozen under dim room light. The dotted vertical lines in spectrum B correspond to g = 1.885, 1.955, and 2.05, respectively. These resonances were ascribed to reduced ferredoxin. Spectrum C was obtained from another sample prepared under conditions similar as above except pH = 8.3. After freezing, the sample was illuminated at 240 K for 2 h with white light from a 800 W tungsten-halogen lamp. The vertical arrows (g  1.765 and 2.09) were ascribed to FX. For spectra A-C the large radical signals around the g = 2 region were erased for easier data display. The inset shows the spectrum recorded on the same sample under and after illumination at low field (between 300 and 315 mT). EPR conditions (temperature, microwave power, and modulation amplitude at 100 kHz, respectively): spectrum A, 8 K, 80 mW, 1 mT; spectrum B, 45 K, 20 mW, 1 mT; spectrum C, 9 K, 80 mW, 1 mT; inset: 4.2 K, 0.2 mW, 2.1 mT.

The lower part of Fig. 3 shows the spectra measured with another sample that was prepared at pH 8.3 and was illuminated at 240 K for 2 h with white light. Spectrum C was recorded at 9 K and high microwave power (80 mW) to enhance signals due to F, which are indicated by vertical arrows (g  1.765 and 2.09). The other spectral features in between are due to the coupled (F, F) signal. Reduced ferredoxin may contribute to features around 1.89 and 2.05 and is distinctly observed as a g  1.96 signal. The ferredoxin spectrum was also measured at 45 K, giving a spectrum similar to spectrum B. It must be also emphasized that the g-values and the linewidths of the spectra A and C shown in Fig. 3 are indistinguishable from those found for the PSI iron-sulfur centers F, F, and F in a frozen PSI solution.

When studied under illumination at 4.2 K, a small signal is observed around 3066 G, as shown in the inset of Fig. 3 (spectrum "under h"). A similar signal is observed at high field in a symmetrical position compared to the g_{free electron} position (not shown). These features can be ascribed to a chlorophyll triplet state signal exhibiting a zero-field splitting parameter |D| of 0.0282 cm¹. Similar features (with |D|  0.028 cm¹) have been previously shown to arise from the P700 triplet state at low temperature (Rutherford and Mullet, 1981). Therefore, we tentatively ascribe the observed signal to the low-field peak (2z) of the ³P700 state. This signal is rather small, which may be explained by the weak stationary concentration of the excited state under continuous illumination of the dark cocrystals. The ³P700 state exhibits four other innermost peaks, which were not observed in our experiments. This may be due to the fact that the outermost peaks of ³P700 observed under stationary conditions are much larger than the innermost features, which makes these last ones unobservable under our experimental conditions. If one assumes that the ³P700 state is formed, it would indicate that the secondary acceptor, the phylloquinone A₁, has been prereduced in some PSI centers (Bonnerjea and Evans, 1982).

Both samples were used for a quantitative estimate of ferredoxin stoichiometry in the cocrystals (see Materials and Methods). The ferredoxin/PSI ratios were found to be 0.85 ± 0.1 and 0.95 ± 0.1 for the two samples, respectively.

Illumination of PSI at low temperature under moderate redox conditions is known to lead to the formation of P700⁺ and (F_A, F_B) with the preferential photoreduction of F_A in most cases (see, e.g., Brettel, 1997). This charge separation was studied in a third sample containing many small cocrystals, which was prepared at pH 7.5 in the presence of ascorbate and DCPIP. The sample was frozen in darkness to maintain P700 reduced and the iron-sulfur clusters oxidized. The light-induced spectrum was recorded at 25 K. Except for a large signal around g = 2.0 due to P700⁺, the largest signals at g = 1.864, 1.943, and 2.049 are ascribed to F. Spectral features due to F are observable at g-values around 1.88, 1.93, and 2.07 (not shown).

EPR study of single cocrystals: preliminary characterization of the light-induced spectra

For this report, two different cocrystals were studied in detail for light-induced spectra at 10-20 K. Illumination of PSI at low temperature under moderate redox conditions leads to the formation of P700⁺ and (F_A, F_B) with the preferential photoreduction of F_A. The presence of ferredoxin at its binding site in a covalent complex was found to decrease the amount of F relative to F (P. Sétif, unpublished observation). Thus we expected a small contribution of F in the light-induced spectra. Moreover, this contribution was decreased in a series of experiments made at 10 K and 20 mW of microwave power, for which a stronger saturation of the F signal compared to that of F is expected, in line with the relaxation properties of the two clusters (Rupp et al., 1979). These experiments will be described and interpreted first. Nonsaturating conditions were also used (20 K, 5 mW), in which the relative contribution of F is increased. These experiments will be described further below.

A complete orientation pattern was obtained at 10 K by recording difference spectra at many ₁ and ₁ angles. This was composed of different datasets, each dataset being obtained by varying ₁ from 0 to 180° and keeping the ₁ angle fixed (see Fig. 1 for a scheme of the crystal holder). From the crystallographic data, the space group has been found to be orthorhombic and the unit cell dimensions have been determined (see above). However, the present crystallographic data did not allow the derivation of the orientation of the C₃ trimeric axis in the unit cell (a, b, c) framework. As an initial guess for interpreting the EPR spectra, it was hypothesized that all C₃ axes were parallel, as is the case in PSI_a crystals (Krauss et al., 1993). If this would be the case, there should be a particular (₁, ₁) orientation with the C₃ axis parallel to the magnetic field, and this should result in a single F line (see, e.g., Brettel et al., 1992). We searched systematically for such a special orientation of cocrystals. This was done for the two different cocrystals by recording light-induced EPR spectra and varying the two angles ₁ and ₁ from 0 to 180° by steps of 10-15° and 5°, respectively. With both cocrystals, it was not possible to observe a single line due to F at any value of (₁, ₁). Several possibilities can be considered to explain these observations: 1) It might be speculated that the right (₁, ₁) orientation was not found because the two angles were not adjusted with a sufficient precision or because of a large mosaic spread. We estimated that this was very unlikely because a misalignment of up to 10° between the C₃ direction and the magnetic field is not expected to give rise to several lines, but only to broaden the F single line, which should lie at g = 1.931-1.935 (Brettel et al., 1992; Kamlowski et al., 1997). 2) One might have to abandon the above hypothesis that all C₃ axes of PSI trimers are parallel to each other in a single cocrystal. In the following it will be shown, from a detailed characterization of the orientation dependence of the signal due to F, that this is the case, i.e., C₃ trimer axes are not parallel one to each other in the PSI/ferredoxin cocrystals.

During the course of the EPR measurements, several characteristics of the light-induced spectra due to F were noteworthy. First, in some orientations, up to 10 different lines could be resolved. For the whole set of data, a minimum number of three different lines were observed. Second, for several (₁, ₁) orientations, one line was observed that was isolated from the others (either in the low- or high-field region of the spectrum). After double integration, this line was found to correspond to ~1/12 (between 1/11 and 1/14) of the whole integrated intensity of iron-sulfur centers. These observations are consistent with the presence of four PSI trimers in the unit cell, giving rise to 12 different lines due to 12 magnetically inequivalent F for a random cocrystal orientation. However, a smaller number of lines (between 3 and 10) was generally observed, presumably because of overlapping. Because of this large number of overlapping lines, it was not possible to follow reliably a given resonance at varying orientations. For further interpretation of the EPR data we chose to use another strategy, which is summarized in the next paragraphs, with details given in the Appendix.

We also observed EPR signals of reduced ferredoxin in single cocrystals. Such data were obtained with cocrystals that were reduced at room temperature at very low redox potentials in the presence of methylviologen. In such cocrystals, large signals due to F and F were always present. For an unambiguous identification of ferredoxin signals, it was therefore necessary to perform the EPR measurements at 40-50 K. This allowed us to observe specifically ferredoxin signals, as the F/F signals are considerably lifetime-broadened at this temperature. However, the signals thus obtained are considerably noisier than those ascribed to F and F, which are discussed in the present paper. This has two different origins: first, the ferredoxin signals at 40-50 K are broader and their amplitude is smaller than the F/F signals recorded at 8-20 K (see, e.g., Fig. 3); second, no baseline could be subtracted for such lines, contrary to the light-induced spectra shown below. This led to broad and poorly resolved ferredoxin signals (not shown).

Characterization of F resonances at three particular orientations

As mentioned above, a minimum number of three lines was found at some particular orientations for the two cocrystals that were studied. For each of the cocrystals, spectra comprising three lines were found at three different (₁, ₁) orientations. The three corresponding spectra were identical for both cocrystals and are shown in Fig. 4 for one of these cocrystals. The (₁, ₁) angles associated to these spectra correspond to mutually perpendicular orientations (see legend to Fig. 4). For interpreting these spectra, the hypothesis that all C₃ axes of the trimers are parallel was relaxed; in the cocrystal, the C₃ axis of a given trimer was considered to make angles _a, _b, and _c with respect to the cocrystal axes a, b, and c, respectively (Fig. 5 A; (A, B, C) is a permutation of (a, b, c)). When taking into account the point group of the cocrystals (222 corresponding to the orthorhombic space group), the squares of the direction cosines between all C₃ axes and (a, b, c) should be identical. Moreover, the squares of the direction cosines between the g-tensor principal axes and (a, b, c) should be identical for one PSI of a trimer and the corresponding PSI of another trimer resulting from a rotation about a unit cell axis. For cocrystal orientations for which one of the unit cell axes is parallel to the magnetic field, the EPR spectrum of a single EPR species should therefore consist of three different lines due to the three different PSIs in one trimer. The three lines observed in the spectra of Fig. 4 might therefore originate from the three magnetically inequivalent F species in the three different PSI reaction centers of a given trimer, when either the a, b, or c axes are parallel to the magnetic field.

View larger version (18K): [in this window] [in a new window]   FIGURE 4   Light-induced spectra recorded at 10 K for a single cocrystal at three mutually perpendicular orientations. Spectra A-C were recorded for (1, 1) = (0°, 40°), (0°, 130°), and (90°, 5°), respectively. The P700+ signal is shown as a dotted line around g = 2.00 in spectrum A and was subtracted from all three spectra. The nine FA resonances that were used for characterizing its g-tensor are indicated as dotted vertical lines and their numerical values are given in Table 1. Integration of spectrum B gave spectrum /B. The right part of this spectrum was used for fitting with two Gaussian lines (not shown). EPR conditions: temperature, 10 K; microwave power, 20 mW; modulation amplitude at 100 kHz, 1 mT.

View larger version (10K): [in this window] [in a new window]   FIGURE 5   Angular parameters used for characterizing the orientation of the C3 trimer axis in the (A, B, C) framework (parts A and B) and in the (gx, gy, gz) framework (part C). (A, B, C) is an arbitrary permutation of (a, b, c), with a, b, and c being the three mutually perpendicular unit cell axes. (B) (u, v, C3) is an orthonormal framework. The u and x axes lie in the (A, B) plane. The C, C3, v, and x axes lie in the same vertical plane.

Nine different g-values (dotted lines) were derived from the three spectra of Fig. 4 (A, B, and C). The g-values of the two high-field lines of spectrum B were obtained after integration of the signal (spectrum /B) and fitting with two Gaussian lines (not shown). Among the nine different lines, two of them deserve special mention: 1) the high-field line of spectrum B is fairly broad. It was unsuccessfully tried to get it narrower by small changes in the (₁, ₁) angles. It will be shown below that this can be easily explained by a small misalignment and/or by mosaic spread. 2) The low-field peak of spectrum C exhibits some structure. This may have two different origins. First, this line is close to g = 2 and subtraction of P700⁺ might not be perfect; second, it will be shown below that this probably results also from a slight misalignment. The nine resonances thus identified and ascribed to F are given in Table 1.

Determination of the orientation of the C₃ axis and of the g-tensor of the cluster F_A

The above g-values were measured with a precision of <0.001 except for the low-field resonance of spectrum C (±0.004). In the following, it is therefore assumed that the three g-values of spectrum A (called set A of g-values; idem for set B and set C) correspond to the resonances of F in the three PSIs of a given trimer when one unit cell axis, denoted A, is parallel to the magnetic field (idem for B and C). As EPR cannot help in identifying A, B, or C to any unit cell axis, (A, B, C) is an unknown permutation of (a, b, c).

By using a minimization procedure described in detail in the Appendix, these nine resonances were sufficient to derive: 1) the two angles (, ) which relate the framework (u, v, C₃) to the framework (A, B, C):  = 36.5°;  = 23.0°. These angles correspond to the following angles between C₃ and the unit cell axes (see Fig. 5): _A = 71.7°; _B = 76.6°; _C = 23.0°; 2) the principal values of the F g-tensor: g_x = 1.864; g_y = 1.9405; g_z = 2.050; and 3) the unitary transformation between the orthonormal frameworks (A, B, C) and (g_x, g_y, g_z). The three Euler angles for transforming (A, B, C) into (g_x, g_y, g_z) are 185.7°, 106.7°, and 23.3°, respectively. From these angles and the (, ) angles, one can also derive the angles between C₃ and the principal directions of the F g-tensor (see Fig. 5): _x = 51.4°; _y = 50.6°; _z = 62.9°.

It can be noted that the above parameters, including the principal g-values of the F g-tensor, were obtained solely from the minimization procedure. The values obtained herein for the F g-tensor are close to what was determined before from the study of PSI_a crystals and to the values obtained from the powder spectrum of cocrystals (Brettel et al., 1992; Kamlowski et al., 1997; see Table 2 for comparison), thus strongly supporting our calculations.

View this table: [in this window] [in a new window]   TABLE 2   g-Tensor characteristics of the reduced clusters F and F of Synechococcus elongatus PSI

Simulation of rotation patterns of F

By taking into account the parameters derived above together with the knowledge of the (A, B, C) axis system in the laboratory framework, it is possible in principle to simulate any set of data. This is shown in Fig. 6 for one dataset for which the predicted positions of the lines were compared to the experimental light-induced difference spectra. This dataset was obtained with the same cocrystal that was assumed to have a peculiar orientation with the B axis almost parallel to the bottom plane of the crystal holder (Fig. 1). In such a case, it is possible to adjust the ₁ angle so that B lies vertically. This was performed for ₁ = 0. In this configuration, the two directions A and C could be made parallel to the magnetic field within the same dataset (for ₁ = 40° and 130°; see arrows marked A and C).

View larger version (51K): [in this window] [in a new window]   FIGURE 6   Light-induced spectra recorded at 10 K. This series of spectra was recorded with the same cocrystal used for Fig. 4. The angles are 1 = 0 and 1 = 0 to 180° by steps of 5°. The spectra A and C (arrows) correspond to those exhibited in Fig. 4. The continuous lines are simulated g-values assuming that the B axis lies perfectly vertical during the whole series, which results in six different resonances at any orientation. The dotted lines were simulated by taking into account that the B axis is not perfectly vertical, which results in some splitting of all six lines (see text). Same EPR conditions as in Fig. 4.

Several features emerge from this figure:

1.	The calculated g-values (continuous lines) satisfactorily match the observed spectra. However, small deviations between experimental and calculated g-values are seen (see spectra at g  1.95 for 1 around 10 and 70°). This discrepancy is not understood at the moment;
2.	Only six different lines were observed. This peculiar situation is ascribed to the fact that the B axis lies almost vertically for the whole dataset. In such a case, two different PSIs related by the twofold rotation around this axis are magnetically equivalent. This is expected to result in a twofold decrease in the number of lines, as found and simulated in Fig. 6. However, the B axis is not perfectly vertical. This was simulated by taking into account this small deviation. Splitting of the six lines that are expected for B being perfectly vertical is expected. This is shown by the dotted lines in Fig. 6. Such a splitting may contribute to the structure observed for the low-field peak of spectrum C (see Fig. 4; 1 = 130° in Fig. 6);
3.	Sets of data recorded at other 1 angles were also satisfactorily fitted with the above parameters (not shown). In these cases, 12 lines are generally expected. Taking into account the crowding and fortuitous overlapping of lines, this was consistent with the experimental spectra for which up to 10 distinguishable lines were observed.

Determination of the g-tensor of the cluster F_B

F signals in a single cocrystal were studied at 20 K and 5 mW of microwave power. Under these conditions, the size of the F signals was relatively enhanced compared to the conditions used above, but only a few F signals of significant amplitude could be identified. These signals, which were recorded with the same cocrystal used for studying F (Figs. 4 and 6) are shown in Figs. 7 and 8 together with the large F signals. Fig. 7 exhibits the spectra recorded at the three orientations for which one of the unit cell axes is assumed to be parallel to the magnetic field. The P700⁺ contributions were not subtracted, contrary to Fig. 4, and this results in large signals around g  2.00. Apart from this P700⁺ contribution, only one supplementary signal for each orientation was observable compared to the spectra of F shown in Fig. 4. These signals, which are indicated by vertical arrows and are marked as g_A, g_B, and g_C, are ascribed to F. The g_A and g_C signals are clearly visible at values of 1.909 and 2.048, respectively. The g_B signal, which is more difficult to observe around g = 1.885, is also expanded 10-fold. Fig. 8 exhibits the low-field features, observed at g  2.045-2.067, which are found with the B axis being vertical (₁ = 0°) and ₁ = 70-190°. Identical orientations were used for studying F (Fig. 6). As is the case for F, a mirror symmetry is observed around ₁ = 130° (the C axis being parallel to the magnetic field). The highest g-values were observed at g = 2.067 for ₁ = 110 and 150°. A relative maximum of g-value was also found at 2.063 for ₁ = 85 and 175°; g-values of F in the cocrystal were found around 1.88, 1.93, and 2.07 from the powder spectrum (not shown). These approximate values (±0.05), together with the orientations of the C₃ axes in the (A, B, C) framework, were used for simulating the signals at g_A, g_B, and g_C (Fig. 7) as well as the angular dependence of the low-field g signals of F_B between 2.045 and 2.067 (Fig. 8). This was done via a minimization procedure, which is detailed in the legend to Fig. 7 and resulted in the following characteristics of the F g-tensor (cf. Table 2): 1) the principal values g_x = 1.877, g_y = 1.931, g_z = 2.068 are much closer to the g-values generally observed for F (see, e.g., Jung et al., 1996; g_x,y,z = 1.880, 1.930, 2.069) than those that were found in a previous study on PSI_a crystals (Kamlowski et al., 1997); 2) the angles between the C₃ axis and the principal directions of the g-tensor: _x = 72°; _y = 56°; _z = 40° (see definitions in Fig. 5) slightly differ from the values found in PSI_a crystals (Kamlowski et al., 1997).

View larger version (23K): [in this window] [in a new window]   FIGURE 7   Light-induced spectra recorded at 20 K and 5 mW of microwave power for the same cocrystal and the same mutually perpendicular orientations as those used in Fig. 4. The resonances at gA = 1.909, gB = 1.885, and gC = 2.048 (arrows) were ascribed to FB. These three g-values were used for characterizing the FB g-tensor together with a few resonances ascribed as well to FB and shown in Fig. 8 (g = 2.067 at (1, 1) = (0, 110°) and (0, 150°); g = 2.063 at (1, 1) = (0, 85°) and (0, 175°); g = 2.061 at (1, 1) = (0, 75°) and (0, 185°). This was performed by a systematic parameter search for minimizing the sum of the deviations between the above-observed six resonances (at 1.909, 1.885, and 2.048 for sets A, B, and C, respectively; at 2.067, 2.063, and 2.061 for (1, 1) = (0, 110°), (0, 85°), and (0, 75°), respectively) with the calculated ones. For this minimization, we use the orientation of the C3 axis that we determined from the study of FA. The following parameters were varied: gx = 1.88 ± 0.05; gy = 1.93 ± 0.05; gz = 2.07 ± 0.05; all possibilities for the angles defining the orientations of the FB g-tensor in the (A, B, C) framework. The minimization output was used for deriving the properties of the FB g-tensor (see text and Table 2) and the three FB resonances that are expected for each of the three orientations with A, B, or C parallel to the magnetic field. These calculated resonances are shown as thin vertical lines in Fig. 7 (set A, g = 1.9085, 1.911, 2.009; set B, g = 1.885, 1.915, 2.018; set C, g = 1.951, 1.985, 2.054).

View larger version (40K): [in this window] [in a new window]   FIGURE 8   Light-induced spectra recorded at 20 K and 5 mW of microwave power. This series of spectra was recorded with the same cocrystal used for Figs. 4 and 6 and at similar orientations as those used in Fig. 6. The angles are 1 = 0 and 1 = 70-190° by steps of 5°. The continuous lines are simulated g-values assuming the properties of the FB g-tensor (Table 2), which were calculated in the present report from the best-fitting procedure described in the legend to Fig. 7. The same six observed resonances as those described in Fig. 7 were best-fitted assuming the orientation of the FB g-tensor that is reported in Kamlowski et al. (1997) (bottom line in Table 2) and by taking the principal FB g-values as free parameters (gx = 1.88 ± 0.05; gy = 1.91 ± 0.05, and gz = 2.05 ± 0.05). The best fit was obtained for gx = 1.878, gy = 1.932, and gz = 2.064 and the dotted lines are simulated g-values assuming these principal g-values together with the orientation reported in Kamlowski et al. (1997).

From the above characteristics, it was possible to derive the F resonances that are expected for the three preferential orientations with C₃ parallel to A, B, or C. These resonances are indicated in Fig. 7 by vertical lines (see legend for numbers). It was also possible to simulate the orientation pattern corresponding to the low-field F resonances shown in Fig. 8 (continuous lines). This pattern can be compared with the best-fit pattern that could be obtained from the angular characteristics of the F g-tensor derived from the data on PSI_a in Kamlowski et al. (1997) by taking the principal g-values as free parameters. This pattern is shown as dotted lines in Fig. 8 (see figure legend for details on the fitting procedure and the principal g-values found for the best fit). From the comparison of both types of fits, we conclude that the angular parameters that are derived in the present work (_x,y,z = 72°, 56°, 40°) are more satisfactory that those found in Kamlowski et al. (1997). The slight deviations between the two sets of angles (up to 6°) thus appear to be significant and could be ascribed to a slight modification of the F_B orientation due to the presence of ferredoxin. However, one cannot exclude that these small deviations (6°) are due to the uncertainties inherent to the EPR analyses of (co)crystals.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX REFERENCES

In chloroplasts and cyanobacteria, ferredoxin associates reversibly to the reaction center of photosystem I (PSI). Within the complex, ferredoxin is photoreduced in the submicrosecond-microsecond time range (Sétif, 2001) before it dissociates for later interaction with other soluble electron acceptor proteins. Only a few examples of structures of reversible complexes between electron transfer proteins are known to date (Pelletier and Kraut, 1992; Chen et al., 1992, 1994; Adir et al., 1996; Morales et al., 2000; Kurisu et al., 2001; Lange and Hunte, 2002). Among these, the x-ray structure of cocrystals formed between ferredoxin and ferredoxin-NADP⁺ reductase has been recently determined by two different groups (Morales et al., 2000; Kurisu et al., 2001). For the complex between photosystem I and ferredoxin or flavodoxin there is no x-ray structure available.

The present paper reports for the first time the cocrystallization of PSI with ferredoxin, both being purified from cyanobacteria, i.e., from Synechococcus elongatus and Synechocystis 6803, respectively. PSI is in the trimeric form, as in PSI_a crystals used for structure determination at 2.5 Å resolution (Jordan et al., 2001). A number of PSI mutations were used for identifying a docking region for ferredoxin (reviewed in Sétif, 2001), in accordance with a simulated docking model based on the 6 Å structure (Fromme et al., 1994) and with modeling studies using electrostatic potentials for the structural alignment (Ullmann et al., 2000). The interface between both partners presents the peculiarity that at least the three different extrinsic stromal subunits and probably extrinsic loops of PsaA of PSI are involved in ferredoxin binding. The determination of the structure of the complex between PSI and ferredoxin is important for the understanding of the protein-protein interaction and the inter-protein electron transfer, because this will allow the PSI/ferredoxin interface to be characterized in greater detail and the possible participation of other PSI subunits in ferredoxin binding to be identified. The first steps toward elucidation of this structure are reported here on cocrystals diffracting x-rays to a resolution of 7 Å. From our data, an orthorhombic space group was deduced with the unit cell dimensions a = 194 Å, b = 208 Å, and c = 354 Å, which leads to four different PSI trimers, i.e., 12 PSI monomeric functional subunits present in the unit cell.

The cocrystals were characterized by EPR at low temperature

First, samples consisting of a large number of small cocrystals were studied after various redox and illumination pretreatments, which gave rise to powder-type EPR spectra. In these experiments, the EPR signatures of all three iron-sulfur clusters of PSI (F_A, F_B, and F_X) were observed, as well as the EPR spectrum of the reduced [2Fe-2S] cluster of ferredoxin. Under highly reducing conditions it was also possible to observe the triplet state of P700 under continuous illumination. These data indicate that the cocrystals are much more resistant than the PSI_a crystals. During the last several years, the use of spectroscopic investigations on single PSI_a crystals led to a much deeper insight into the function, location, and structure of some of the cofactors (Brettel et al., 1992; Kamlowski et al., 1997, 1998; Bittl et al., 1997; Käss et al., 2001). However, spectroscopic investigations on single PSI_a crystals were limited by their instability. These cocrystals were grown at low ionic strength and show a solvent content of 80%. Moreover, only four salt bridges are involved in crystal contacts (Jordan et al., 2001; Fromme, 2002). The crystals dissolve immediately by addition of redox compounds such as dithionite, or at pH above 7, so that only a limited number of redox states of the cofactors could have been investigated.

In contrast, the PSI/ferredoxin cocrystals are much better suited to spectroscopic investigations because they are stable in the presence of many redox compounds in a pH range from 5.5 to 9.5. This property will be used in future experiments for studying various EPR species, such as the interaction spectrum of the [4Fe-4S] clusters F and F of PSI or the [2Fe-2S] cluster of ferredoxin. Values of