Cytochrome bc1-cy Fusion Complexes Reveal the Distance Constraints for Functional Electron Transfer Between Photosynthesis Components*

Photosynthetic (Ps) growth of purple non-sulfur bacteria such as Rhodobacter capsulatus depends on the cyclic electron transfer (ET) between the ubihydroquinone (QH2): cytochrome (cyt) c oxidoreductases (cyt bc1 complex), and the photochemical reaction centers (RC), mediated by either a membrane-bound (cyt cy) or a freely diffusible (cyt c2) electron carrier. Previously, we constructed a functional cyt bc1-cy fusion complex that supported Ps growth solely relying on membrane-confined ET ( Lee, D.-W., Ozturk, Y., Mamedova, A., Osyczka, A., Cooley, J. W., and Daldal, F. (2006) Biochim. Biophys. Acta 1757, 346-352 ). In this work, we further characterized this cyt bc1-cy fusion complex, and used its derivatives with shorter cyt cy linkers as “molecular rulers” to probe the distances separating the Ps components. Comparison of the physicochemical properties of both membrane-embedded and purified cyt bc1-cy fusion complexes established that these enzymes were matured and assembled properly. Light-activated, time-resolved kinetic spectroscopy analyses revealed that their variants with shorter cyt cy linkers exhibited fast, native-like ET rates to the RC via the cyt bc1. However, shortening the length of the cyt cy linker decreased drastically this electronic coupling between the cyt bc1-cy fusion complexes and the RC, thereby limiting Ps growth. The shortest and still functional cyt cy linker was about 45 amino acids long, showing that the minimal distance allowed between the cyt bc1-cy fusion complexes and the RC and their surrounding light harvesting proteins was very short. These findings support the notion that membrane-bound Ps components form large, active structural complexes that are “hardwired” for cyclic ET.

Gram-negative, purple, non-sulfur, facultative phototrophic ␣-proteobacteria of Rhodobacter species provide excellent model systems for studying photosynthetic (Ps) 4 and respira-tory electron transfer (ET) chains (1)(2)(3)(4). Among them, Rhodobacter capsulatus uses both a freely diffusible and a membrane-anchored cytochrome (cyt) (c 2 and c y , respectively) electron carrier between the ubihydroquinone (QH 2 ):cyt c oxidoreductase (cyt bc 1 complex) and either the photochemical reaction center (RC) or the cbb 3 -type cyt c oxidase (C ox ) under Ps or respiratory growth conditions, respectively (5-7). In contrast, the closely related Rhodobacter sphaeroides relies exclusively on the soluble cyt c 2 for its Ps growth (8), even though it has a membrane-bound cyt c y that is functional only under respiratory growth conditions (9). R. capsulatus cyt c y is 199 amino acids long, with its NH 2 -terminal 28 residues corresponding to an unprocessed signal sequence-like membrane anchor and COOH-terminal 100 residues (Thr 99 to Arg 199 ) to a monoheme cyt c domain (10). The remaining 70-residues (Asn 28 to Thr 98 ) long Ala and Pro-rich region of cyt c y forms a linker attaching the cyt c domain to the membrane anchor and allowing its observed mobility (11). Noticeably, the linker of R. capsulatus cyt c y is markedly longer than that of R. sphaeroides and other non-Ps bacterial cyt c y (Fig. 1A). A seminal finding was that R. sphaeroides mutants lacking cyt c 2 can be complemented for Ps growth by R. capsulatus cyt c y (12), pointing out that the cyt c y linker length might be critical for its electron carrier function during Ps cyclic ET.
Earlier studies, regarding the rate of ET mediated by cyt c y from the cyt bc 1 complex to the RC, and the correlative presence of the cyt bc 1 complex with cyt c y in membranes (7,13), suggested that these proteins must be in close proximity to one another and to the RC and its light harvesting (LH) complexes. Occurrence of larger, nontransient macromolecular structures, ensuring efficient substrate channeling, catalytic enhancement, and sequestration of reactive intermediates during electron transport have been proposed to occur in membranes of purple bacteria (14), plants (15), and mitochondria (16). However, as the membrane-embedded components can diffuse independently from one another within lipid bilayers, differentiating between static "hardwired" electron transport complexes from those undergoing random collisions (17) is difficult to document in Ps membranes (18). In the accompanying work (38), we have demonstrated by creating soluble variants of cyt c y (cyt S-c y ) that this electron carrier needs not be membrane anchored to support Ps growth of R. capsulatus. In this work, we have exploited the functional cyt bc 1 -c y fusion complex that we constructed earlier by fusing genetically cyt c y to the cyt bc 1 complex (19) to probe the physical proximities of the Ps components to one another in R. capsulatus. We surmised that if the native length of the cyt c y linker is optimized for efficient electronic coupling across the distance between the cyt bc 1 complex and the RC, then progressively shortening it might divulge the minimal distances separating these physiological partners in Ps membranes. Characterization of the physicochemical properties of both membrane-embedded and highly purified cyt bc 1 -c y fusion complex variants indicate that the shortest functional cyt c y linker is about 45 residues long. This short linker still insures rapid ET from the cyt bc 1 to the RC, but exhibits decreased electronic coupling to the RC, thereby limiting cyclic ET and Ps growth. Thus, Ps cyclic ET components appear to be tightly packed together, forming membrane-embedded large structural complexes.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-Bacterial strains and plasmids used in this study are listed in Table 1. R. capsulatus strains were grown at 35°C in mineral-peptone-yeast extract-enriched media (MPYE) supplemented with antibiotics as needed (10, 2.5, and 10 g/ml kanamycin, tetracycline, and spectinomycin, respectively) under either semiaerobic/dark (respiratory), or photoheterotrophic/light (Ps) conditions using anaerobic jars and H 2 ϩ CO 2 generating gas packs (BD Biosciences), as described previously (20). Ps growth curves were obtained using screw-cap tubes, incubated at 35°C in an aquarium filled with water and illuminated by tungsten (Lumiline, Sylvania) lamps at an intensity of 150 E/m 2 /s. Culture turbidity was monitored with a Klett-Summerson photometer equipped with a red (number 66) filter as described (21).
A ⌬(petABC::gen) insertion-deletion allele was constructed by replacing the 2.4-kb ApaLI-StuI fragment of petABC in plasmid pMTS1 with the 1.2-kb HindIII-BamHI fragment containing the gentamycin resistance gene by blunt end ligation, yielding pYO34. This ⌬(petABC::gen) allele was then transferred using Gene Transfer Agent into the chromosome of R. capsulatus strain FJ2 (⌬cycA, ⌬cycY) (5) to yield the triple mutant YO2, lacking both the cyt bc 1 complex and electron carrier cyts c 2 and c y .
Spectroscopic Techniques-Optical spectra were recorded on a PerkinElmer UV-visible spectrophotometer Lambda 20. Absorption difference spectra for the c-and b-type cytochromes were obtained using chromatophore membranes (0.3 mg of total protein/ml), oxidized by adding a crystal of potassium ferricyanide, and reduced by a few grains of either solid sodium ascorbate or sodium dithionite, as appropriate. Timeresolved, light-activated kinetic spectroscopy was performed on a dual wavelength kinetic spectrophotometer with chromatophore membranes resuspended in 50 mM MOPS buffer containing 100 mM KCl (pH 7.0) in the presence of the following redox mediators (with their respective midpoint redox potential, E m,7 ): 100 M ferricyanide (430 mV), 8 M 2,3,5,6tetramethyl-p-phenylenediamine (260 mV), 6 M 1,2-naphthoquinone (NQ, 145 mV), 1 M phenazine methosulfate (80 mV), 1 M phenazine ethosulfate (50 mV), 6 M 2-hydroxy-1,4naphthoquinone (HNQ, Ϫ145 mV), 6 M benzyl viologen (Ϫ359 mV), and a membrane potential uncoupler (2.5 M valinomycin), as described (26). The amount of chromatophore membranes used in each assay was normalized to the RC content, as determined by measuring the flash-induced optical absorbance difference between 605 and 540 nm at an E h of 380 mV, and using an extinction coefficient of 29.8 mM Ϫ1 cm Ϫ1 . Transient cyt c re-reduction and cyt b reduction kinetics at an ambient potential of 100 mV, initiated by a short saturating flash (ϳ8 s) from a xenon lamp were followed at 550 -540 and 560 -570 nm, respectively. Antimycin, myxothiazol, and stigmatellin were used as indicated at 5, 5, and 1 M, respectively.
Optical potentiometric titrations were performed with the purified cyt bc 1  . The optical changes that accompanied the E h changes were recorded in the ␣-band region (500 to 600 nm), and the E m values were determined by fitting the normalized absorption data to a single component n ϭ 1 Nernst equation. EPR spectroscopy was performed at sample temperatures of 10 or 20 K using a Bruker ESP 300E spectrometer (Bruker Biosciences), fitted with an Oxford instruments ESR-9 helium cryostat (Oxford Instrumentation Inc.). Spectrometer settings were as indicated in the appropriate figure legends.
Purification of the Cyt bc 1 -c y Fusion Complex-The R. capsulatus cyt bc 1 -c y fusion complex was purified using pYO38/YO2. Chromatophore membranes (5.9 g wet weight) were obtained from frozen washed cells (81 g wet weight) after two passages through a French pressure cell and collected by ultracentrifugation (20). Chromatophore pellets were resuspended to a final protein concentration of 18 mg/ml in 50 mM phosphate buffer (pH 8.0) containing 7.2 mM NaCl, 20% glycerol, 1 mM phenylmethylsulfonyl fluoride, 100 mM EDTA, 10 protease inhibitor mixture tablets (Roche Applied Science) and 1 mM ⑀-aminocaproic acid. Dodecyl maltoside (DDM, 20% w/v stock solution) was added dropwise to this suspension to a final concentration of 1 mg of DDM/mg of total proteins. The mixture was stirred gently for 1 h at 4°C, and then ultracentrifuged (120,000 ϫ g for 2 h) to eliminate non-dispersed membranes. The supernatant was loaded onto a Q-Sepharose ff column (2.6 ϫ 32 cm) preequilibrated with 50 mM phosphate buffer (pH 8.0) containing 20% glycerol, 0.01% (w/v) DDM, and 7.2 mM NaCl (Buffer A). The column was washed with 5 to 6 column volumes (CVs) of Buffer A containing 400 mM NaCl, and then the remaining photosynthetic pigments were washed with 3 to 4 CVs of the same buffer containing 0.05% (w/v) DDM until a red band on top of the column became visible. The adsorbed cyt bc 1 -c y fusion proteins were eluted with 4 CVs of a linear 400 -800 mM NaCl gradient in the presence of 0.05% (w/v) DDM. Fractions were monitored for their absorption at 280 and 420 nm, and 500 to 600 nm for their dithionite-reduced minus ferricyanide-oxidized optical difference spectra, and those containing the highest concentrations of c-type and b-type cytochromes were pooled, concentrated using an Amicon Diaflo apparatus equipped with a PM30 membrane. The concentrated sample (ϳ33 ml) was passed twice through an anti-FLAG M2 affinity column (5 ml), pre-equilibrated with 10 CVs of 50 mM phosphate buffer (pH 8.0) containing 150 mM NaCl and 0.01% (w/v) DDM. The column was then washed with 20 CVs of the same buffer, and eluted with 6 CVs of 100 mM glycine-HCl buffer (pH 3.5) containing 0.01% (w/v) DDM. Eluents were collected into vials containing 1 M Tris-HCl buffer (pH 8.0) to a final concentration of 50 mM. Fractions containing the cyt bc 1 -c y fusion complex were pooled, buffer exchanged with 50 mM phosphate buffer (pH 8.0) containing 0.01% (w/v) DDM using Amicon Ultra (50,000 M r cut off) centrifugal filter devices (Millipore Co., Ireland), and stored at Ϫ80°C in the presence of 20% glycerol until further use.

Design of the Cyt bc 1 -c y Fusion Complexes with Shorter Cyt c y
Linker Lengths-The availability of a functional cyt bc 1 -c y fusion complex (19) allowed us to probe whether the cyt c y linker could be used as a "molecular ruler" to estimate the distances between the Ps ET components. Comparison of R. capsulatus cyt c y with its counterparts from other species indicated that the cyt c domain is highly similar to that from Silicibacter pomeroyi (71%), Paracoccus denitrificans (70%), and R. sphaeroides (65%). On the other hand, the linker regions of P. denitrificans, R. sphaeroides, and S. pomeroyi are about 20 to 30 residues shorter than that of R. capsulatus (Fig. 1A). Generation of computer-assisted hypothetical three-dimensional structures of the cyt bc 1 -c y fusion complexes with shorter linkers directed us to the region between amino acids 65 and 90 of cyt c y to mimic its non-functional variants as Ps electron carriers (Fig. 1B). Considering that the R. sphaeroides cyt c y linker is 26 amino acids shorter than its R. capsulatus counterpart, plasmids pYO30 (with a 24-amino acid long deletion between positions Ala 63 and Pro 88 of cyt c y ) (⌬24-c y ) and pYO33 (with a 19-amino acid long deletion between the positions Ala 67 and Pro 88 of cyt c y ) (⌬19-c y ) were constructed to determine the Ps ET of each electron carrier (Table 1).
Phenotypic Characterization of R. capsulatus Strains Harboring Cyt bc 1 -c y Fusion Complex Variants with Shorter Cyt c y Linkers-Plasmids pYO30 (⌬24-c y ) and pYO33 (⌬19-c y ) were introduced into the R. capsulatus mutant YO2 (lacking the cyts bc 1 , c 2 , and c y ), and the Ps growth abilities of the resulting strains were examined ( Fig. 2A). On enriched MPYE medium under Ps conditions, R. capsulatus strain pMTS1/MT-RBC1 overproducing cyt bc 1 complex, MT-G4/S4 lacking the cyt c 2 , and pYO38/YO2 harboring an intact cyt bc 1 -c y fusion complex exhibited doubling times of about 192, 216, and 300 min, respectively ( Fig. 2A). However, under similar conditions, pYO33/YO2 (⌬24-c y ) and pYO30/ YO2 (⌬19-c y ) grew markedly slower than pYO38/YO2 (native-c y ) (756 and 498 versus 300 min, respectively). As the mutant YO2 cannot be complemented for Ps growth by either the cyt bc 1 complex or cyt c y alone, the data indicated that the cyt bc 1 -c y fusion complexes with shorter cyt c y linkers provided both the oxidoreductase and electron carrier functions required for Ps growth. However, the slower growth rates suggested that shortening the cyt c y linker hampered the growth abilities of PS.
Prosthetic Group Insertion, Enzymatic Activity, and Subunit Assembly of the Cyt bc 1 -c y Fusion Complexes-As the slower Ps growth might also be attributed to decreased amounts of cyt bc 1 -c y fusion complexes, chromatophore membranes of appropriate strains were examined by TMBZ/SDS-PAGE analyses (Fig. 2B). Typical membrane-associated c-type cytochrome profiles comprised of the cyts c p (32 kDa), c 1 (30 kDa), c y (29 kDa), and c o (28 kDa) were observed for pMTS1/MT-RBC1 and MT-G4/S4, whereas the YO2-derived samples contained only cyts c p and c o (both subunits of C ox ) as expected. On the other hand, pYO38/YO2, pYO33/YO2, and pYO30/YO2 harboring cyt bc 1 -c y fusion complexes with shorter cyt c y linkers had both MAY 16, 2008 • VOLUME 283 • NUMBER 20 cyts c 1 and c y bands replaced by a single peroxidase-active band at 61, 59, and 58 kDa, respectively, corresponding to their cyt c 1 -c y fusion subunits.

Functional Cyt bc 1 -c y Fusions with Variable Length Cyt c y Linkers
Steady-state enzymatic activities of the cyt bc 1 complexes were also assayed by measuring DBH 2 -dependent reduction of horse heart cyt c (Fig. 2C). pYO38/YO2, pYO33/YO2, and pYO30/YO2 harboring cyt bc 1 -c y fusion complexes with different lengths of cyt c y linkers showed very similar levels of DBH 2 : cyt c reductase activities as compared with that of the pMTS1/ MT-RBC1 overproducing a native cyt bc 1 complex. Therefore, neither fusing cyt c y to cyt c 1 , nor changing the length of the cyt c y linker region significantly affected the cyt bc 1 complex enzymatic activity of various cyt bc 1 -c y fusion complexes with different cyt c y linker lengths.
The presence of b-type cyts and cyts c to b ratios for all strains were also determined using optical difference spectroscopy to assess the relative amounts of the cyt bc 1 complexes. Changes in the amounts of the c-type (␣ peak max at 551 nm, ⑀ 551-542 of 20 mM Ϫ1 cm Ϫ1 ) and b-type (␣ peak max at 560 nm, ⑀ 560 -574 of 28 mM Ϫ1 cm Ϫ1 ) cytochromes were monitored after reduction by ascorbate and dithionite, respectively, of ferricyanide-oxidized chromatophore membranes from appropriate strains (Fig. 3). Strains pMTS1/MT-RBC1 and MT-G4/S4 exhibited cyt c to b ratios of ϳ1:2, whereas very small amounts of c or b peaks were detected in YO2. On the other hand, strains pYO38/YO2, pYO33/YO2, and pYO30/YO2 had similar amounts of b-type cytochromes with cyt c to b ratios of about 1:1 (Fig. 3). The data indicated that all strains contained  similar amounts of cyt bc 1 -c y fusion complexes regardless of the linker length of their cyt c y . Furthermore, to confirm that the Fe-S subunits of fusion complexes had native-like physicochemical properties, EPR spectroscopy was used. Chromatophore membranes prepared from pYO38/YO2 exhibited the [2Fe-2S] cluster g y and g x signals of 1.891 and 1.806, respectively, which were identical to those seen with a native cyt bc 1 complex (e.g. MT-G4/S4) (Fig. 4A). Moreover, under appropriate conditions, the EPR signals with g z values of 3.778 and 3.411 assigned to cyts b L and b H , respectively, were observed with the same membrane preparations (Fig. 4B).
The overall data established that all cyt bc 1 -c y fusion complexes assembled similarly, and exhibited similar enzymatic activities, indicating that the slower growth rates observed with pYO33/YO2 and pYO30/YO2 could not be correlated with the amounts of fusion complexes.
Purification and Characterization of Purified Cyt bc 1 -c y Fusion Complexes-Purification of cyt bc 1 -c y fusion complexes was pursued to establish that they existed as intact physical entities in the membranes. Earlier, we partially purified the cyt bc 1 -c y fusion complex (19) using the procedure described by Valkova-Valchanova et al. (22). However, as detailed characterization required purer samples, we developed a new procedure. About 8 mg of purified cyt bc 1 -c y fusion complex was obtained starting with about 2.5 g of chromatophore membranes derived from 97 g (wet weight) of cells, followed by detergent solubilization, Q-Sepharose ff ion-exchange and FLAG affinity column chromatographies ( Table 2). The final preparations of the cyt bc 1 -c y fusion complex contained less than 5% of Ps pigments associated with the LH complexes, as estimated by optical spectra (data not shown). The purified cyt bc 1 -c y fusion complex was highly active, reducing horse heart cyt c as an electron acceptor with decylhydroquinone as an electron donor ("Experimental Procedures"). Its specific activity (about 27.2 nmol/mg of protein/min under the assay conditions used) was comparable with that of the purified cyt bc 1 complex (41.0 nmol/mg of protein/ min) (22). Optical difference spectra indicated that the purified cyt bc 1 -c y fusion complex had a cyt b to c ratio of ϳ1:1, similar to that seen in chromatophore membranes from pYO38/YO2 (data not shown). Potentiometric redox equilibrium titration of the heme groups of the cyt c 1 -c y fusion subunit in the presence of 100 mM KCl at pH 7.0 indicated a single component with an E m,7 value of ϩ336 mV (Fig. 5B). Considering that this subunit is a diheme c-type cyt (cyts c 1 and c y of E m,7 of 320 and 365 mV, respectively (10,27), the data suggested that the two hemes were in rapid equilibrium with each other, with a mean E m,7 value high enough to convey electrons efficiently from the cyt bc 1 complex to the RC.
SDS-PAGE analysis of purified cyt bc 1 -c y fusion complex showed three major bands with 61, 41, and 24 kDa, assigned to the cyt c 1 -c y , cyt b, and the Fe-S protein subunits, respectively, by TMBZ staining and immunoblot analyses with specific monoclonal antibodies (Fig. 5A). Additional bands of higher M r seen with the cyt bc 1 -c y fusion complexes were attributed to their aggregated forms, based on TMBZ staining and immunoblot data, as they were also seen with native cyt bc 1 complexes (27). Finally the non-stoichiometric band running between the cyt b and Fe-S protein subunits, and detected weakly by cyt c 1 -specific antibodies, in FLAG affinity purified samples reflected degradation products of the cyt c 1 -c y fusion subunit. The oligomeric state of purified cyt bc 1 -c y fusion complex was determined by size exclusion chromatography using an analytical grade Superose 6 HR 10/30 (GE Healthcare Inc.) column calibrated with high molecular weight standard markers in the presence of 0.05% (w/v) DDM and 150 mM NaCl (Fig. 5B). Under these conditions, the purified cyt bc 1 complex runs as one major peak around 240 kDa (estimated to correspond to its dimeric form), whereas the purified cyt bc 1 -c y fusion complex FIGURE 2. A, photosynthetic growth properties on liquid and solid media (enriched MPYE) of various R. capsulatus strains (YO2 lacking the cyt bc 1 , cyt c y , and cyt c 2 ; pMTS1/MT-RBC1 overproducing the cyt bc 1 complex, MT-G4/S4 lacking the cyt c 2 , and pYO38/YO2, pYO33/YO2, and pYO30/YO2 containing the cyt bc 1 -c y fusion complexes with the native 19 amino acids and 24-amino acid shorter cyt c y linkers, respectively) were determined by monitoring the turbidity of the cultures, as described under "Experimental Procedures." B, the c-type cytochrome profiles of the same strains were revealed using chromatophore membranes (100 g of total proteins per lane) and 15% SDS-PAGE/TMBZ analyses. C, DBH 2 :cyt c reductase activities of chromatophore membranes (20 g of total proteins) derived from the same strains described above were determined as in Atta-Asafo-Adjei and Daldal (20), in the absence (no inhibitor) or presence (10 M stigmatellin or 20 M antimycin), and for comparative purposes the steady-state enzymatic activities are represented as % of the overproduced native cyt bc 1 complex. exhibited two major peaks at about 417 and 257 kDa, respectively. Immunoblot analyses with subunit-specific antibodies showed that both peaks had homogeneous constituents, suggesting that purified cyt bc 1 -c y fusion complexes consisted of a mixture of tetrameric and dimeric forms under the conditions used.
Light-activated Cyt b Reduction and Cyt c Re-reduction Kinetics-A major aim being to probe the extent of ET from the cyt bc 1 -c y fusion complexes with shorter cyt c y linkers to the RC in situ, appropriate strains were examined using light-activated, time-resolved optical spectroscopy (Fig. 6). In all cases, chromatophore membranes were normalized to the same amounts of photooxidized RC by measuring the absorbance changes at 605-540 nm at an E h of 380 mV after a train of 10 saturating actinic flashes spaced 50 ms apart. Transient cyt b reduction and cyt c re-reduction kinetics, initiated by rapid (microsecond time scale) light activation of the RC, were monitored on the millisecond time scale, at an ambient redox potential E h of 100 mV. At this E h , the membrane Q pool contains Q and QH 2 , and the RC as well as cyts c, and the [2Fe-2S] cluster of the Fe-S protein subunits of cyt bc 1 or the cyt bc 1 -c y fusion complexes are fully reduced. First, light-induced time-resolved single turnover cyt b reduction kinetics were examined at 560 -570 nm, in the presence and absence of antimycin A as a specific cyt bc 1 complex Q i site inhibitor. In MT-G4/S4, which has only cyt c y as the sole electron carrier between the cyt bc 1 complex and the RC, and in pYO38/YO2 producing a cyt bc 1 -c y fusion complex with a native cyt c y linker, cyt b was reduced by oxida-tion of a QH 2 via the Q o site, and immediately re-oxidized by Q bound at the Q i site with an expected rate faster than the available time resolution. However, in the presence of antimycin A (2 M), cyt b oxidation was abolished to reveal the light-induced transient reduction phase (Fig. 6, top row). In the case of pYO33/YO2 (⌬19-c y ) and pYO30/YO2 (⌬24-c y ) with shorter cyt c y linkers, fast cyt b oxidation/reduction rates were quasisimilar to those seen with MT-G4/S4 and pYO38/YO2, but the amplitudes of these changes (per RC) were significantly smaller (Fig.  6, top row). As similar amounts of cyt bc 1 -c y fusions complexes were present in all strains tested (Figs. 2 and 3) the smaller cyt b reduction amplitudes cannot be interpreted as lower amounts of these complexes in the membranes.
The cyt c oxidation/re-reduction kinetics exhibited by the cyt bc 1 complexes were subsequently examined using the specific Q o site inhibitors myxothiazol and stigmatellin to probe ET to the RC. Myxothiazol blocks cyt c re-reduction by displacing QH 2 without immobilizing the Fe-S protein, whereas stigmatellin not only displaces QH 2 but also immobilizes the Fe-S protein to block ET to the c 1 heme as well. In the presence of stigmatellin, MT-G4/S4 and pYO38/YO2 showed typical cyt c oxidation without any cyt c re-reduction (Fig. 6, bottom row), due to the absence of ET from the Fe-S protein subunit to the cyt c 1 heme. In the presence of myxothiazol, cyt c re-reduction reached about half of the amplitude seen in the presence of stigmatellin (i.e. full oxidation), revealing the pre-flash, chemically derived electron localized in the Fe-S protein despite the absence of QH 2 oxidation at the Q o site (Fig. 6, bottom row) (see e.g. Refs. 28 and 29 for a detailed explanation of these ET kinetics). Cyt c re-reduction kinetics exhibited by pYO33/YO2 (⌬19-c y ) and pYO30/YO2 (⌬24-c y ) were quasi-similar to those seen with MT-G4/S4 and pYO38/YO2 (native-c y ) in the presence and absence of the Q o site inhibitors. However, in the presence of shorter cyt c y linkers, significant differences in the cyt c oxidation/re-reduction amplitudes were observed. These amplitude decreases could not reflect lower amounts of the cyt bc 1 -c y fusion complexes with shorter cyt c y linkers in the strains examined (Figs. 2 and 3). Thus, they indicated decreased electronic couplings to the RC, revealing the limits of their physical abilities to reach and convey electrons to the photooxidized RCs (Fig. 6). Comparative kinetic data indicated that a cyt c y linker of about 45 amino acids long (i.e. ⌬24-c y ) seems to be the shortest one able to sustain cyclic ET and Ps growth of R. capsulatus.

DISCUSSION
In this work, first, we have characterized the physicochemical properties of a functional cyt bc 1 -c y complex that we constructed earlier (19) by fusing the COOH-terminal last amino acid of the cyt c 1 subunit of the cyt bc 1 complex to the NH 2 -terminal first amino acid of its physiological membrane-bound electron acceptor cyt c y (Fig. 1A). Using membrane preparations, detailed analyses demonstrated that the cyt bc 1 -c y fusion complex contained its prosthetic groups ([2Fe-2S] cluster and b-and c-type hemes) in appropriate amounts and assembled properly. Purification of the cyt bc 1 -c y fusion complex was needed to establish that it is an intact physical entity able to conduct membrane-confined Ps cyclic ET. This task was challenging because the fusion complex was susceptible to proteolytic degradation, and its chromatographic properties were distinct from the previously purified cyt c y (10) and cyt bc 1 complexes (22). We developed a new procedure, using tandem Q-Sepharose ff ion-exchange and FLAG affinity column chromatographies, to obtain highly pure samples with good yields to initiate crystallization efforts. Unlike the dimeric cyt bc 1 complex, the purified cyt bc 1 -c y complex was a mixture of dimeric and tetrameric populations, with different oligomerization properties in vitro, possibly due to the presence of cyt c y . The FLAG affinity purified samples contained small amounts of cyt c 1 -c y degradation products, running as an additional band around 30 kDa, and the Q-Sepharose column fractions were enriched in RC-LH subunits, usually absent in purified cyt bc 1 complexes (nLC-MS/MS data, not shown). During purification, although most of the Ps components are usually washed out from the cyt bc 1 complex in the presence of 0.01% DDM by a buffer of 0.2 M ionic strength, they remained with the cyt bc 1 -c y fusion complex. The significance, if any, of these apparent tighter associations remains to be seen.
Interestingly, both cyt b reduction and cyt c re-reduction rates exhibited by the cyt bc 1 -c y fusion complexes with shorter linkers were quasi-similar to those seen with strains harboring unconnected or fused cyt c y and cyt bc 1 complexes with native cyt c y linkers. These fast rates further evidenced that the monitored ETs were mediated by the membranebound cyts c y , and not by their soluble versions (cyt S-c y ), somehow generated by proteolysis. Otherwise, as described in the accompanying work (38), the ET rates would have become much slower. Sharply contrasting the rates, the amplitudes of ET to the RCs were slightly smaller that the unconnected systems even in the presence of a cyt bc 1 -c y complex with a native cyt c y linker, and became progressively smaller with shorter linkers. That only a fraction of all lightoxidized RC complexes can be reduced by cyt c y has been well documented earlier using native systems (5,12). This  L and b H ), chromatophore membranes were reduced with ascorbate or air oxidized, respectively, as described in Refs. 35 and 36. Spectra were recorded at 20 K, 9.443 GHz, 12 G and 10 K, 9.59 GHz, 10 G temperature, microwave frequencies, and modulation amplitudes, respectively. work therefore demonstrated that the amounts of electronically coupled RC-LH-cyt bc 1 -c y fusion complexes diminished further with shorter cyt c y linkers, reflecting the distance dependence of efficient cyclic electron transport between the donor and acceptor complexes in Ps membranes. Furthermore, the progressive decrease suggested that membranes might contain a distribution of such closely associated photosynthetic units. Even though the cyt bc 1 -c y fusion complexes with shorter cyt c y linkers were produced in comparable amounts irrespective of the linker sizes, they supported Ps growth at different degrees. Comparison of the shortest (⌬24-c y ) and barely functional cyt c y linker with that of non-Ps competent R. sphaeroides cyt c y , which is 26 residues shorter than that of its R. capsulatus counterpart (Fig. 1A), suggested that at least a linker of about 44 -46 amino acids long (provided that the native linker is about 70 residues long) is required to couple electronically the cyt bc 1 -c y fusion complex and the RC surrounded with its LH complexes. Assuming that both the RC (30) and the cyt bc 1 complex (31) extend into the periplasm by ϳ30 Å, then half of the remaining linker would be consumed to bring the cyt c domain of cyt c y to the same plane of the purified cyt bc 1 -c y fusion complex and cyt bc 1 complex in the presence of 150 mM NaCl and 0.05% DDM. Gel filtration chromatography was performed using a Superose 6 HR 10/30 column, which was run at a flow rate of 0.3 ml/min, and the elution profile was monitored at 280 nm. The column was calibrated with blue dextran (2,000 kDa), thyroglobulin (669 kDa), apoferritin (440 kDa), catalase (232 kDa), and aldolase (158 kDa) as standards, run in the presence of 150 mM NaCl and 0.05% DDM, and their elution positions are indicated at the top of the chromatograph. Aliquots of each fraction (0.5 ml) were concentrated and subjected to 15% SDS-PAGE and immunoblot analyses (lower panel) using subunit-specific antibodies as described in A. C, dark equilibrium redox titration of the cyt c 1 -c y subunit of the purified cyt bc 1 -c y fusion complex (0.1 mg/ml). The titration was performed in 50 mM MOPS buffer (pH 7.0) containing 0.1 M KCl and 1 mM EDTA in the presence of 0.01% (w/v) DDM. Redox mediators were as described under "Experimental Procedures" (26). The E m,7 value indicated was determined by fitting the normalized data to a n ϭ 1 Nernst equation. with its physiological partners. The remaining 20 -25-residue long linker controlling its electronic coupling ability then suggest that the RC-LH complexes and the cyt bc 1 -c y complex must be very close to each other, possibly forming large structural complexes. Furthermore, it is noteworthy that both the R. sphaeroides cyt c y (9) and R. capsulatus cyt bc 1 -c y fusion complexes with shorter linkers convey electrons to their cognate cyt c oxidases. 5 Thus, Ps supercomplexes apparently require longer cyt c y linkers than the respiratory counterparts, possibly due to the LH complexes surrounding the RCs.
Kinetic behaviors of functional supercomplexes in R. sphaeroides Ps membranes between the RC and the "trapped" cyt c 2 acting in a locally confined manners (14,(32)(33)(34), strongly suggest that the RC-LH and the cyt bc 1 complexes must be very close to each other. The data presented here provide comple-mentary structural and kinetic information for such Ps supercomplexes, with a ratio of the RC-LH:cyt bc 1 :cyt c 2 (or c y ) being 2:1:1 (13,33). Joliot et al. (33) proposed that in the RC-LH complexes dimerized via LH-PufX interactions, two RC dimers interact with a single cyt bc 1 complex dimer. If this is also the case with the cyt bc 1 -c y fusion complexes in the Ps membranes, then c y cyts might be located between the cyt bc 1 complexes and the RC-LH complexes with their membrane anchors in the vicinity of the quasi-closed LH rings, promoting tighter associations between the Ps components. The motion constraints thus imposed on cyt c y might then only be compensated by a linker long enough to allow its movement between the cyt bc 1 complex dimers and the opposing RC-LH complexes of R. capsulatus (Fig. 7).
In summary, the availability of functional cyt bc 1 -c y complexes with shorter linkers affecting the coupling to the RC-LH complexes now provide compelling indications that hardwired 5 Y. Ö ztü rk, D. Zannoni, and F. Daldal, unpublished data. FIGURE 7. A, crystal structure of the RC-LH1 core complex (PDB 1PYH) from Rhodopseudomonas palustris and hypothetical three-dimensional structural model of R. capsulatus cyt bc 1 -c y fusion complex (pYO30/YO2, ⌬24-c y ) (with the transmembrane helices of cyt bc 1 shown as ribbons, and cyt c y (⌬24-c y ) shown as sticks), are drawn using the program PyMOL. The narrow section of the RC (subunits L, yellow; M, blue; H, purple) surrounded by the LH1 complex (chains ␣, pale green and ␤, green) and the cyt bc 1 -c y fusion complex (subunits cyt b, cyan; cyt c 1 , orange; cyt c y , red; and the Fe-S protein, magenta) are viewed parallel to the membrane plane. B, top view (perpendicular to the membrane plane) of the RC-LH1 core complex and the cyt bc 1 -c y fusion complex with the shortest cyt c y linker (⌬24-c y ) is shown. Note that to reach the central part of RC subunits L and M in the RC-LH1 and cyt bc 1 -c y fusion complexes, the cyt c domain of cyt c y (assuming that its linker is stretched out in a fashion parallel to the membrane) needs to reach out for about 100 Å. C, a schematic representation of the major membrane proteins involved in cyclic ET of purple bacterial photosynthesis (upper panel), and a proposed mechanism for cyt c y -mediated cyclic ET via supraorganization of R. capsulatus photosynthetic unit (lower panel). cyt bc 1 complex, hydroquinone cyt c oxidoreductase; cyt c y , membrane-associated cytochrome c y . Arrows indicate directions of electron (e Ϫ ) and excitation (ex) flows.
Ps units occur in membranes. Future purification of these Ps units will initiate studies addressing how the membrane supercomplexes are formed and regulated in vivo in response to changing environmental conditions, and why some organisms contain both a soluble cyt c 2 and a membrane-anchored cyt c y .