Functional consequences of the organization of the photosynthetic apparatus in Rhodobacter sphaeroides: II. A study of PufX- membranes.

In the bacterium R. sphaeroides, the polypeptide PufX is indispensable for photosynthetic growth. Its deletion is known to have important consequences on the organization of the photosynthetic apparatus. In the wild-type strain, complexes between the reaction center (RC) and the antenna (light-harvesting complex 1 (LH1)) are associated in dimers, and LH1 does not fully encircle the RC. In the absence of PufX, the complexes become monomeric, and the LH1 ring closes around the RC. We analyzed the functional consequences of PufX deletion. Some effects can be ascribed to the monomerization of the RC.LH1 complexes: the number of RCs that share a common antenna for excitation transfer or a common quinone pool become smaller. We examined the kinetic effects of the closed LH1 ring on quinone turnover: diffusion across LH1 entails a delay of approximately 1 ms, and the barrier appears to be located directly against the quinone-binding (secondary quinone acceptor (Q(B))) pocket. The diffusion of ubiquinol from the RC to the cytochrome bc1 complex is approximately 2-fold slower in the mutant, suggesting an increased distance between the two complexes. The properties of the Q(B) pocket (binding of inhibitors, stabilization of Q(B-), and rate of Q(B)-H2 formation) appear to be modified in the mutant. Another specificity of PufX- is the accumulation of closed centers in the Q(A-) (where Q(A) is the primary quinone acceptor) state as the secondary acceptor pool becomes reduced, which is probably the origin of photosynthetic incompetence. We suggest that this is related to the Q(B) pocket alterations. The malfunction of the reaction center is probably due to a faulty association with LH1 that is prevented in the PufX-containing structure.

In the preceding article (1), we examined some functional consequences of the supramolecular arrangement of the photosynthetic apparatus in the purple non-sulfur bacterium Rhodobacter sphaeroides. Two main issues were addressed concerning the excitation transfer between the RC 1 ⅐LH1 com-plexes and the localization of the quinone pool. In both cases, the dimeric association of the RC⅐LH1 complexes was shown to play an important role. However, the dimeric arrangement of the RC⅐LH1 complexes is not a general feature in photosynthetic bacteria. For instance, monomeric RC⅐LH1 complexes from Rhodospirillum rubrum were resolved by electron microscopy from two-dimensional crystals (2,3), whereas the same technique allowed the observation of dimers in native (4) or reconstituted (5) membranes of R. sphaeroides. The three-dimensional structure obtained from x-ray crystallography of complexes from Rhodopseudomonas palustris is also monomeric (6). There are also atomic force microscopy data that confirm the dimeric arrangement in R. sphaeroides (7), whereas monomers were observed by this technique in Blastochloris viridis (8) and Rhodospirillum photometricum (9).
Another important structural feature of the RC⅐LH1 complexes that differs in the various cases described so far is the open/closed character of the LH1 ring surrounding the RC. LH1 is an oligomeric ring-shaped structure whose building block is a heterodimer of two subunits named ␣ and ␤ (each a transmembrane helix) and binds two bacteriochlorophylls and one carotenoid. In R. sphaeroides, LH1 does not form a closed ring, but an open C-shaped structure. Several ␣␤-heterodimers with the associated bacteriochlorophylls are missing with respect to a complete closed ring. From biochemical evidence, the number of ␣␤-heterodimers/RC has been estimated in the range of 10 -13 (4, 10 -13), whereas the "complete" ring is expected to contain 16 heterodimers, as in R. rubrum. From electron microscopy images, a figure of 12 was estimated (4,5). An open ring has also been obtained in the crystallographic structure of R. palustris (6), which includes 15 ␣␤-heterodimers. The other known structures display closed LH1 rings. In such a case, it is not obvious how the quinone acceptors can efficiently shuttle between their reducing (Q B ) site on the RC and their oxidizing site on the bc 1 complex.
To understand the structural causes and functional consequences of the particular supramolecular arrangement found in R. sphaeroides, it is very useful to examine the effects of the deletion of PufX. This small polypeptide, coded in the puf operon (which also contains the genes for the RC and LH1 subunits), appears to play an important role in controlling the supramolecular structure. It is found at a 1:1 ratio with the RC in purified RC⅐LH1 complexes (14). The sequence is indicative of a single transmembrane ␣-helix with hydrophilic C-and N-terminal loops; a study based on gradual clipping investigated the functional roles of each loop (15). The available evidence indicates that the deletion of PufX suppresses the formation of RC⅐LH1 dimers and that about four additional ␣␤-units are incorporated to form a closed ring (5,13,14,16). Such mutants have lost the ability to grow photosynthetically under anaerobic reducing conditions. Photosynthetic growth is recovered, however, when oxidants such as trimethylamine N-oxide and Me 2 SO are added to the medium (17,18). When examined on a single turnover basis, the electron transfer kinetics appear normal. However, a dramatic slowing down of the cyclic flow is observed under steadystate illumination under anaerobic conditions (17,18). It was suggested that the closed LH1 ring around the RC obstructs the shuttling of quinol/quinone between the RC and bc 1 complexes.
In this work, we compare the functional behavior of membranes from a PufX Ϫ mutant and from the WT. The consequences of the monomeric structure appear to be a greater confinement of the quinone pool and more restricted excitation transfer. Special attention is given to the question of quinone exchange on the RC and diffusion to the bc 1 complex. We show that the closed LH1 ring results in a rather moderate (ϳ2-fold) slowing of these processes, indicating that it does not constitute a tight barrier to quinones. However, we found that the PufX deletion results in modified properties of the quinone-binding (Q B ) pocket of the RC, which may account for the blocking of the RC under reducing conditions. We propose that the main role of PufX is to prevent an incorrect interaction of the RC with LH1, which is responsible for this malfunction.

EXPERIMENTAL PROCEDURES
We used the same materials and procedures as described in the preceding article (1), except those specified below. The strain denoted as WT was R. sphaeroides Ga. The PUF⌬X/g strain, with PufX deleted (denoted here as PufX Ϫ ), was a kind gift of Prof. D. Oesterhelt. Its construction has been described (17). It was grown under semi-aerobic conditions in Hutner's medium containing kanamycin (25 g/ml). As reported previously, this strain was unable to grow under photosynthetic conditions (anaerobiosis ϩ light). This strain contains both the LH1 and LH2 antenna complexes. RC⅐LH1 complexes were isolated as described in the preceding article (1). In agreement with previous work (14), the dimer band found in the sucrose gradient with the WT was totally absent in preparations from the PufX Ϫ strain. A single RC⅐LH1 band was observed, and these complexes displayed the characteristic monomer behavior as described in the previous article with respect to excitonic connectivity or sharing of the quinone pool. A difference with respect to monomers prepared from WT membranes was the amount of endogenous quinone retained in the complexes, which was larger (typically 14 Q molecules/RC) than in WT complexes (typically 7 Q molecules/RC), in line with the larger quinone pool found in membranes from this mutant compared with those from the WT (grown under the same conditions).
For monitoring absorption changes at subzero temperatures (see Fig.  6), we used our Joliot-type spectrophotometer equipped with the low temperature attachment as described (19). The cuvette accommodates a thin sample layer (Ϸ0.25 mm) between a glass slide and a reflecting metal sheet on a Peltier element. The transmission is measured from the reflected light. The membrane samples used were 10-fold more concentrated than those used in the setup for room temperature experiments, where the optical path was 15 mm. Due to the small optical path, absorption changes can be detected in frozen aqueous samples without addition of glycerol or antifreeze. The temperature is measured with a thermocouple in close thermal contact with the sample.

RESULTS
Excitation Transfer- Fig. 1 shows a plot of the normalized variable bacteriochlorophyll fluorescence yield (⌽) as a function of the amount of closed centers (c). The curvature of such a plot is indicative of the degree of "connectivity" of the RC⅐LH1 complexes, i.e. the probability that the excitation visiting a complex with a closed RC can be made available for other RCs (20,21). The data were obtained during the P ϩ Q A Ϫ recombination following a saturating flash to stigmatellin-inhibited membranes. The relative fluorescence yield and the amount of P ϩ (603 nm absorption change) were recorded at discrete times, and each data point is a couple of these values. The ⌽(c) relationship obtained with PufX Ϫ membranes (Fig. 1, closed circles) displays a smaller curvature than that obtained with WT membranes (open circles and plus signs for membranes from photosynthetic and semi-aerobic cultures, respectively). These curves can be characterized (22) by a parameter (J) obtained by a fit with the following function (Equation 1).
This relation is appropriate for a large and reasonably homogeneous array of RC⅐antenna complexes. As previously shown (see Ref. 1 and the associated Supplemental "Experimental Procedures"), it may also be used in a broader context as a phenomenological fitting function. The parameter J expresses the degree of connectivity of RC⅐LH1 units: the trapping section of an open RC is enhanced by a factor of J ϩ 1 when the neighboring centers become closed. 2 The allowed interval for J ranges from 0 for isolated units to J lake ϭ F m /F 0 Ϫ 1 (where F m and F 0 are the fluorescence yields for closed and open RCs, respectively) for unrestricted excitonic diffusion ("lake model"). J depends both on the efficiency of antenna connectivity and on the trapping properties of the RC, notably the quenching efficiency of the closed RC. The fit of the PufX Ϫ data (Fig. 1, solid line) corresponds to J ϭ 0.32, which may be compared with a value of 0.81 for the WT. This indicates that the RC⅐LH1 monomers in the PufX Ϫ membrane are midway between the situation of completely isolated units and the connectivity observed in the WT, which was determined as essentially due to efficient transfers between the two partners within an RC⅐LH1 dimer. A likely interpretation is that there is a distribution of the distances between monomers, so some of them behave as isolated units, whereas others are close enough to allow excitation transfer.
Photoreduction of the Quinone Pool- Fig. 2 shows the photoreduction kinetics of TMPD (603 nm absorption change) during continuous illumination of PufX Ϫ (closed circles) and WT (open circles) membranes. The reaction medium included the set of inhibitors and uncouplers indicated in the preceding article (1). In particular, the bc 1 inhibitors myxothiazol and FIG. 1. Excitation connectivity in PufX ؊ membranes. Shown is a plot of the relative variable fluorescence ⌽ ϭ (F Ϫ F 0 )/(F m Ϫ F 0 ) as a function of the fraction of closed reaction centers (c) in the P ϩ state. The fluorescence yield and the decay of P ϩ (absorption change at 603 nm) were monitored during the recombination of P ϩ Q A Ϫ in stigmatellin-inhibited membranes. •, data obtained with PufX Ϫ membranes, fitted (solid line) with Equation 1 (J ϭ 0.32); E and ϩ, data obtained from WT membranes from photosynthetic and semi-aerobic cultures, respectively. antimycin A were used for blocking quinol oxidation. As explained previously (1), the photo-oxidation of TMPD under such conditions is a mirror image of the photoreduction of the acceptor pool, and we will use these terms as equivalents. The TMPD concentration was 10 mM, allowing reduction of P ϩ in the 100-s range; as documented below, this reaction is not rate-limiting in these experiments. The vertical scale indicates the number of electrons transferred per RC, using for calibration the absorption change induced by one saturating flash.
The major difference in Fig. 2 between the PufX Ϫ and WT kinetics is the overall size of the acceptor pool. It is ϳ92 electrons/RC (46 quinones) in the mutant, which is ϳ2-fold larger than in the WT. These figures were found to be somewhat variable depending on preparations, but the enlargement of the pool in the mutant was always obvious. Typical figures are 40 -60 quinones/RC in PufX Ϫ membranes and 25-30 and 15-25 quinones/RC in WT membranes from semi-aerobic and photosynthetic cultures, respectively. The experiments in Fig.  2, which were run at moderate illumination intensity, reveal no obvious slowing of the kinetics in the mutant (in the region where both kinetics can be compared) that could be ascribed to the closed LH1 ring. The experiments described below investigate this issue in a more quantitative manner. Fig. 3 shows a magnification of the initial part of the photoreduction kinetics under high illumination intensity. The kinetics in the PufX Ϫ membrane are clearly slower than those in the WT membrane, especially beyond the first 2 electrons (dashed lines). The initial slope (e.g. for the first electron) corresponds (WT curve) to ϳ1 electron/700 s. This is ϳ4-fold slower than P ϩ reduction by TMPD. This initial rate is thus still essentially limited by the photochemical reaction, depending on the light intensity and its collection and trapping efficiencies. The initial rate for the PufX Ϫ curve is slower, probably because of the lower LH2 content. Beyond this initial part of the kinetics, a roughly linear section can be identified, as shown by the dashed lines. The slope of this line can be taken as the quasi steady-state rate of quinone reduction in the presence of the oxidized pool. It is representative of the full turnover of the quinone cycle on the Q B pocket, including the transfer of 2 electrons and 2 protons, the release of the reduced quinol, and the binding of a fresh oxidized quinone. Fig. 4A shows plots of this rate as a function of the illumination intensity for WT and mutant membranes. Clearly, the rate-limiting step for quinone turnover is 2-3-fold slower in the mutant. To analyze these data, we made the assumption that the turnover of the centers involves sequentially a photochemical reaction, with rate constant ␣I (where I is the light intensity and ␣ is an efficiency constant), and a dark recovery reaction, with rate constant k r . The intensity dependence of the turnover rate should then be as in Equation 2.
The lines in Fig. 4 are best fits of the data using Equation 2. The accessible intensity range was not sufficient to reach the saturated rate k r in the WT membranes, but this may be deduced from the fits, which give k r Ϸ 0.57 (PufX Ϫ mutant) and 1.29 (WT) electrons per ms and per RC. Therefore, the full turnover of one quinone (2 electrons) on the RC corresponds to time constants (2/k r ) of 3.5 and 1.6 ms for the mutant and WT membranes, respectively. This is slower than the electron transfer reactions on the donor side (135 s) and acceptor side (Ͻ100 s) (23-26) and must predominantly reflect the time required for releasing Q B -H 2 and binding an oxidized ubiquinone to the Q B pocket. Fig. 4B shows the results of similar experiments run with isolated RC⅐LH1 complexes. When applied to WT membranes, the isolation procedure described in the preceding article (1) yielded a major fraction of dimeric complexes and a small amount of monomeric complexes. With PufX Ϫ membranes, the RC⅐LH1 complexes were strictly monomeric, in agreement with Ref. 14. Both types of complexes retained a 25-30% fraction of the native Q pool. As may be seen, the saturation curves obtained for the dimeric complexes from the WT and for the monomeric complexes from PufX Ϫ are similar to those of the corresponding membranes. With RC⅐LH1 monomers from the WT, we obtained different results depending on the preparation procedure. Fresh WT monomers collected directly from the sucrose gradient had the same maximum rate as the dimers. Further purification on a chromatography column or aging resulted in complexes with a slower turnover, similar to the PufX Ϫ complexes. (For the latter, there was no difference between "fresh" and "purified" preparations.) In a preliminary report (27), we declared that PufX Ϫ and WT monomers behave similarly, but we have now realized that this does not apply to fresh preparations.
When first tackling these experiments, we considered it likely that the closed LH1 ring in the PufX Ϫ mutant would delimit two quinone pool domains. An internal pool within the ring could react directly with the RC at a fast rate, whereas the quinones from the outside would appear as a slower pool (as illustrated below in Fig. 11A). The results of Figs. 3 and 4 do not support the first proposition since a slower rate-limiting step was observed in the PufX Ϫ mutant from the very first quinone turnovers (excluding the initial reduction of the Q B bound in the dark). On the other hand, the photoreduction kinetics in the PufX Ϫ membranes (see Fig. 2) slow down markedly in the final part, which might be a sign of a kinetic barrier due to the closed ring. The experiment in Fig. 5 was devised to test this possibility. The idea is to compare the kinetics during a continuous illumination pulse (curve 1) or after a dark interruption of several hundred milliseconds (400 ms for curve 2 and 700 ms for curve 3) inserted after reducing a large fraction (about two-thirds) of the total pool. If the slower rate observed during this later part of the photoreduction kinetics were limited by oxidized quinone crossing the LH1 barrier, whereas the "internal pool" would be reduced, then the dark period would allow this diffusional process to continue, reconstituting an oxidized proximal pool. This would predict significantly faster kinetics when resuming the illumination compared with the uninterrupted control. The data of Fig. 5 do not support this prediction. The traces observed when resuming the illumination can be superimposed (open symbols) on curve 1 after a suitable horizontal translation. Fig. 5B shows a zoom of curve 1 and the translated curves 2 and 3. A small fast phase is actually observed (arrow) with an amplitude of ϳ2 electrons. It corresponds to the binding of an oxidized Q B in the RC pocket during the dark period.
The preceding results show that the only kinetic barrier resulting from the PufX deletion appears to be located directly at the opening of the Q B pocket and that there is no proximal pool, besides Q B itself, with privileged access to the RC. This suggests that the LH1 ring is in close contact with this region of the RC. Moreover, this barrier exerts a significant but moderate effect, slowing the turnover rate by slightly Ͼ2-fold. This is far from the dramatic blocking suggested by previous reports (17,18), but it makes the photosynthetic competence of other bacteria with a native closed LH1 ring easier to understand.
We wondered whether lowering the temperature would lock the LH1 subunits more tightly and inhibit the quinone turnover in the PufX Ϫ mutant. This was not really the case, as illustrated in Fig. 6, which shows photoreduction kinetics run at Ϫ22°C. As described elsewhere (28), in this experiment, the bulk medium was frozen, but it appears that a liquid water phase was still in contact with the donor side of the RC, allowing submillisecond reduction of P ϩ by TMPD over many turnovers. Briefly stated, this view is substantiated by the following observations. Fig. 6 (inset) shows the kinetics of P ϩ reduction by TMPD following a single turnover flash on both sides (Ϫ5 and Ϫ10°C) of the freezing transition of bulk medium. The freezing temperature was around Ϫ7°C when cooling because of a supercooled state of the medium, (Thawing occurred around 0°C.) The bulk freezing was accompanied by a 5-fold acceleration of the reaction, which presumably reflects an increase in the TMPD concentration in the liquid phase. The reaction rate was still proportional to the concentration of added TMPD (data not shown), implying a collisional process. Together with the finding that many turnovers are still possible (Fig. 6), this renders unlikely the possibility of the RC reacting with bound TMPD and justifies our suggestion of a liquid phase.
The quinone turnover is also preserved at Ϫ22°C, although very much slowed down with respect to room temperature. As pointed out in the preceding article (1), the demonstration of quinone turnover at low temperature supports our proposal that quinones are not diluted among the lipids, but form a particular phase where the Q molecules are still mobile at subzero temperatures and remain accessible to the RC. The slopes of Fig. 6 correspond to turnover rates of ϳ(100 ms) Ϫ1 and (180 ms) Ϫ1 / electron/RC for the WT and PufX Ϫ mutant, respectively, hence a slowing by ϳ100 with respect to room temperature. As seen from the break occurring after the first electron, the Q A Ϫ Q B 3 Q A Q B Ϫ electron transfer has become partly responsible for the slower quinone turnover, as shown elsewhere (28). Irrespective of this complication, the results of Fig. 6 show that the diffusion of quinone across the LH1 ring of the PufX Ϫ mutant still occurs at an appreciable rate at Ϫ22°C.
Diffusion of Quinol to the Cytochrome bc 1 Complex-When the quinone pool is oxidized in the dark, the quinol oxidation reaction on the oxidizing site of the cytochrome bc 1 complex becomes rate-limited by the arrival of a quinol reduced by the RC (29). To monitor this process, we used ferrocene (E m ϭ 420 mV) as an electron donor for the RC. The redox potential of the assay medium was ϳ380 mV, so the high potential chain of the bc 1 complex (Fe-S center and cytochrome c 1 with E m Ϸ 270 -300 mV) was oxidized in the dark. Antimycin A was added to inhibit the reoxidation of the b hemes on the quinone-reducing site of the complex. Fig. 7 shows the kinetics of the absorption changes triggered by the second flash of a series for the wavelength difference (561-569 nm), reflecting the reduction of heme b 561 (29). (Oxidized/reduced spectra of this heme can be found in Refs. 30 and 31.) As shown in the inset, no heme b reduction was observed on the first flash: this is a check that no quinol was present in the dark, so the kinetics observed on the second flash correspond to the release of the quinol formed on the Q B pocket of the RC and its diffusion to the bc 1 complex. The half-time thus obtained for the WT was ϳbout 7 ms, in agreement with literature data (29). For the PufX Ϫ membranes, the process was slowed by 2-fold (t1 ⁄2 Ϸ 14 ms). We thus again observed a "significant but moderate" effect of the PufX deletion. Similar experiments were reported by Barz et al. (18) and Francia et al. (15), who found a larger slowing (Ͼ10-fold) and a pronounced lag of ϳ9 ms before the onset of the reduction kinetics. In our experiments, the time course of heme b reduction is sigmoidal, with a lag of only ϳ2 ms. We do not know the origin of this discrepancy. The major differences in the experimental procedures are the use of a mixture of redox mediators (1-10 M), but no addition of an exogenous donor to P ϩ in the experiments of Barz et al. and Francia et al., whereas we added ferrocene and no other mediator. Tentatively, we suggest that there may be a problem with the use of endogenous cytochrome c 2 as a donor. These authors noted that the PufX Ϫ membranes were not sealed vesicles, which should entail considerable dilution of the soluble cytochrome.
Quinone Domains-As described in the preceding article (1), one can estimate a "size" of the domains where the quinone acceptors can diffuse in the 100-ms time range by measuring the decrease in the extent of the photoreducible pool (denoted as P) when inhibiting a variable fraction of the RCs. This decrease is due to quinones that cannot access an active RC anymore. If the diffusion domain includes n centers, this means that all n centers are inhibited: the probability for this is f n , where f is the probability of having one center blocked by the inhibitor. Fig. 8 (closed circles) shows the results of such an experiment with PufX Ϫ membranes. The WT data already shown in the preceding article are plotted for comparison (plus signs). The solid line is a best fit with the function 1 Ϫ f n , yielding n ϭ 1.7, which suggests that most quinone domains in the PufX Ϫ membranes include only one or two RCs. The average domain size is clearly smaller than in the WT membranes. As previously noticed, the WT data do not match the theoretical function for a single n, suggesting a more complex distribution of domain sizes than in the mutant.
Q A Ϫ Accumulation during the Photoreduction Process-In the preceding article (1), we studied the accumulation of Q A Ϫ during pool photoreduction in WT membranes and isolated RC⅐LH1 complexes. The relevant quantity here is the amount of Q A Ϫ present after the relaxation of the electron transfer equilibrium. Because of the large equilibrium constant implied by the ⌬E m Ϸ 130 mV between Q A and the pool, one expects that the accumulated Q A Ϫ reflects the fraction of RCs whose available quinone pool is entirely reduced. In this manner, the FIG. 6. Photoreduction kinetics at ؊22°C for WT and PufX ؊ membranes. The same conditions as described in the legends to Figs. 2 and 3 were used, except for a 10-fold larger membrane concentration (see "Experimental Procedures"). The inset shows the 603 nm reduction kinetics of P ϩ following a flash before (Ϫ5°C; E) and after (Ϫ10°C; •) the bulk medium freezing. With respect to the base line (dashed line), the initial negative change is due to P ϩ , and the final positive level is due to oxidized TMPD. The TMPD concentration was 10 mM. distribution of the quinone/RC stoichiometry can be estimated. The results obtained with the WT were indicative of a relatively broad distribution of this ratio, which appeared to be roughly consistent with the statistical fluctuations corresponding to the quinone domain size. Fig. 9 (closed circles) shows the results obtained with the mutant. As described previously for the WT, we used variable periods of strong illumination to reduce variable fractions of the pool. The electron transfer relaxation of Q A Ϫ after putting the illumination off was monitored to determine the extent of the slow phase reflecting the equilibrated fraction of Q A Ϫ . This determination is unambiguous because the relaxation was completed in Ͻ100 ms, whereas the slow phase extended over several seconds. The extent of the slow phase is plotted as a function of the reduced pool (closed circles). Comparison with the WT results (open circles) reveals a dramatic difference between the two strains, with a much larger accumulation of Q A Ϫ in the mutant. Fig. 9 indicates a surprisingly small (i.e. close to 1) "apparent equilibrium constant" between Q A and the Q pool in the PufX Ϫ mutant. In the interpretation framework that we have developed thus far, this would imply a very broad distribution function for the Q/RC ratio. This distribution would have to be much broader than that of the WT. Fig. 8 shows that, in terms of the number of RCs sharing a common pool, the quinone domains were significantly smaller in the PufX Ϫ mutant. In itself, this would tend to broaden the distribution, but the effect should be offset by the increased number of quinones/RC present in the PufX Ϫ membranes. Even if the quinone domains included a single RC (although Fig. 8 indicates the average size is close to two), the statistical fluctuations about a mean of 45 quinones would still fail by far to account for the data. The analysis of the pool distribution in isolated complexes (see preceding article (1)) led us to suggest a cooperative association, implying a self-affinity of the isoprenoid quinones. The consequence of such cooperativity is to broaden the distribution. It seems rather unlikely, however, that the deletion of PufX would result in a major enhancement of this cooperative trend. It thus appears difficult to account for the results of Fig.  9 in terms of a drastic modification of the pool distribution in the mutant, and an alternative explanation is called for. As described below, we found out that the deletion of PufX caused important modifications of the equilibria on the acceptor side of the RC. We believe that this is probably the explanation for the results of Fig. 9 and, more importantly, for the inability of the PufX Ϫ strain to grow under anaerobic conditions.
The Q B Pocket Is Modified in the PufX Ϫ Mutant-When performing the experiments of Fig. 8, we noticed that the affinity for stigmatellin was significantly lower in the PufX Ϫ membranes than in the WT. This effect is documented in Fig.  10A, which shows plots of the fraction of inhibited centers as a function of the concentration of added stigmatellin. These membranes contained little cytochrome c 2 , so, in the absence of stigmatellin, the reduction of P ϩ following a flash took place mostly in the seconds range. The fraction of centers with stigmatellin bound to the Q B pocket in the dark was estimated from the extent of the 50-ms decay phase due to the P ϩ Q A Ϫ recombination. The stigmatellin concentration for half-inhibition is ϳ6-fold higher in the mutant (0.89 M) than in the WT (0.16 M). Marked differences in the same direction were also observed with other inhibitors of the Q B pocket (terbutryn, atrazine, and o-phenanthroline) (data not shown).
Another difference between the WT and PufX Ϫ mutant concerns the rate of P ϩ Q B Ϫ recombination, which is ϳ4-fold slower in the mutant. This is illustrated in Fig. 10B, which shows the kinetics obtained in RC⅐LH1 complexes. (The effect was also observed in membranes, but the unambiguous determination of the P ϩ Q B Ϫ kinetics was complicated, on this slow time scale, by the possible interference of cytochrome c 2 .) No difference was found between monomeric (open circles) and dimeric (open inverted triangles) WT complexes. On the other hand, the rate of P ϩ Q A Ϫ recombination in the presence of a saturating concentration of stigmatellin was the same in complexes from the WT and PufX Ϫ (inset). This indicates that the reason for the slower P ϩ Q B Ϫ recombination is a larger equilibrium constant

An increased stabilization of Q B
Ϫ is expected, all things being equal, to decrease the equilibrium constant for the "second electron transfer," Q A Ϫ Q B Ϫ ϩ 2H ϩ 7 Q A Q B -H 2 . This prompted us to search for possible kinetic modifications concerning this step. Ϫ was probed by measuring the P ϩ absorption change induced by a saturating flash triggered at various times after putting off the continuous illumination. This relaxation phase was completed in Ͻ100 ms and was followed by a much slower phase. The data points indicate the extent of this slow phase, reflecting the "equilibrium" of Q A Ϫ with its acceptor pool. The extent of pool photoreduction (horizontal scale) was determined from the TMPD absorption changes. •, data for PufX Ϫ membranes; E, data for WT membranes.
of two exponentials with a half-time of 40 s (80% of the amplitude) and a slow phase with a half-time of 250 s. In the PufX Ϫ membranes (closed circles), the kinetics are markedly slower, with t1 ⁄2 Ϸ 330 s (single exponential). DISCUSSION The deletion of the PufX polypeptide has a number of structural and physiological consequences. This subunit controls the supramolecular arrangement of the RC⅐LH1 complexes, preventing the formation of a complete closed LH1 ring around the RC and inducing a dimeric association of the PufX RC⅐LH1 complexes. The deletion of PufX results in a phenotype incapable of photosynthetic growth under reducing conditions. Other modifications whose relation with the arrangement of RC⅐LH1 complexes is not straightforward are also observed. The structure of the intracytoplasmic membranes is affected (16,17), so membrane preparations do not yield sealed vesicles as in the WT. It may thus be hypothesized that the supramolecular arrangement of the RC⅐LH1 complexes affects the curvature of the membrane and controls in this manner its large-scale structure. We encountered another unexpected consequence of the PufX deletion, viz. an increase in the number of quinones/RC, which is possibly due to the modified membrane structure. The issues investigated in this work can be classified into three categories: (i) those related to the RC⅐LH1 monomers; (ii) those directly related to the closed LH1 ring; and (iii) functional modifications of the RC, which are probably due to a faulty RC⅐LH1 association.
Monomeric Organization-The deletion of PufX results in smaller domains for both energy transfer and quinone diffusion, meaning that a smaller average number of RCs share excitons or quinones. Concerning excitation transfer, the analysis developed in the preceding article (1) led us to conclude that, in the WT, the diffusion occurs in a very effective way within RC⅐LH1 dimers with little dimer-to-dimer transfer, be it direct or mediated by LH2. In the PufX Ϫ membranes, the ⌽(c) curve (Fig. 1) is closer to the diagonal, implying a decreased excitonic connectivity. Using the WT curve as indicative of 100% RC⅐LH1 dimers, the PufX Ϫ data can be fitted as a combination of 55% isolated monomers and 45% dimers, which may reflect a more or less random dispersion of the RC⅐LH1 monomers and LH2 rings over the membrane. Our results are not consistent with a preferential clustering of RC⅐LH1 or with a systematic isolation of the RC⅐LH1 monomers by LH2 spacers.
The experiment aimed at determining the average number of RCs that share the same quinone pool (Fig. 8) yielded n Ϸ 1.7. Compared with the results obtained for the WT, the relationship between the extent of the photoreducible pool and the fraction of inhibited RCs was closer to the theoretical curve for a fixed domain size, suggesting a more homogeneous arrangement in the mutant. In the WT, there was clearly a significant contribution of domains with n Ͼ 2, which was not seen in the mutant. Our interpretation of these results is explained in the preceding article (1). We proposed that the isoprenoid quinones are not dispersed in the lipid phase, but tend to form clusters with a particular affinity for the membrane proteins. In this picture, the dimeric association of the RC⅐LH1 complexes would favor a distribution 3 of domains with pairs of RCs, e.g. with a major n ϭ 2 contribution and smaller contribution of n ϭ 4, etc. In the PufX Ϫ mutant, the dissociation of the RC⅐LH1 dimers into monomers is then expected to decrease the average n, notably by increasing the contribution of domains with n ϭ 1. Our results agree qualitatively with such predictions.
Concerning the distribution of quinones among the domains, one would not expect big differences between PufX Ϫ and WT based on the above data. The domain size in terms of number of RCs/domain is smaller, but the average stoichiometry in terms of quinones/RC is larger. The outcome is that the average number of quinones/"isolated" domain should be similar, and one expects a similar width of the distribution. It was thus a surprise to us to observe a quite different relationship between the fractions of reduced Q A and Q pool (Fig. 9), indicating a Ϫ recombination phase with t1 ⁄2 Ϸ 50 ms following a saturating flash. B, P ϩ Q B Ϫ recombination kinetics in purified RC⅐LH1 complexes. •, PufX Ϫ monomers (t1 ⁄2 Ϸ 4.9 s); E and ƒ, WT monomers and dimers, respectively (t1 ⁄2 Ϸ 1.2 s for both WT preparations). The inset shows the P ϩ Q A Ϫ recombination kinetics in WT (E) and PufX Ϫ (•) complexes after addition of stigmatellin (t1 ⁄2 Ϸ 65 ms). C, normalized kinetics of the 450 nm absorption change following the second flash of a series. Ferrocene (100 M) was present, allowing the reduction of P ϩ with t1 ⁄2 Ϸ 15 ms. This slower contribution to the absorption changes was subtracted. The extent of the changes in the 1-ms range was ϳ10-fold smaller after the first flash than after the second one, confirming the interpretation of the kinetics as due to Q A Ϫ Q B Ϫ 3 Q A Q B -H 2 . The WT data are fitted with a single exponential (dashed line; t1 ⁄2 Ϸ 60 s) or with a sum of two exponentials (solid line; t1 ⁄2 Ϸ 40 s (80%) and t1 ⁄2 Ϸ 250 s (20%)). The PufX Ϫ data are fitted with a single exponential (t1 ⁄2 Ϸ 330 s; a two-exponential fit converged toward the single exponential). much lower apparent equilibrium constant in the PufX Ϫ membranes. This difference is not due to a kinetic effect: it was observed under quasi-equilibrium conditions in the 100-ms range after putting the illumination off. In the interpretation framework developed in the preceding article (1), this would indicate a much broader distribution of the Q/RC stoichiometry in the mutant compared with the WT, which, as argued above, seems very unlikely. We propose below an alternative explanation involving a modified equilibrium constant in the RC.
Quinone Diffusion Rate-From a study of the pool reduction rate as a function of light intensity, we could estimate the rate of the process limiting quinone turnover on the Q B pocket. Thus, under saturating illumination, the average time for a full cycle is ϳ1.6 ms in the WT. This includes the 2-electron transfer steps, the uptake of 2 protons, the binding of an oxidized quinone, and the release of a quinol. The electron and proton transfer reactions are known to occur with half-times of some tens of microseconds, so the rate limitation must originate from either the binding or release step (or both). In the PufX Ϫ mutant, the turnover time is ϳ3.5 ms, thus slowed down by a factor of 2.2. This measurement was carried out in the initial part of the kinetics, with a fully oxidized pool. We wondered whether this slower turnover was indicative of the quinone flow across the LH1 or whether a more severe effect of the LH1 barrier could be evidenced at later times. We reasoned that, in the dark-adapted state, the oxidized quinones could be distributed both inside and outside the LH1 ring and that the inside "proximal" pool could react faster with the RC than the outside pool (Fig. 11A). We thus searched for a kinetic discrimination between a fast and a slow pool, but the outcome was clearly negative. First, there was no obvious break in the photoreduction kinetics under strong illumination indicative of the exhaustion of a fast pool (except Q B ). We then probed specifically the slower part of the kinetics when the pool is ϳ70% reduced. Assuming that the proximal pool is fully reduced and that the kinetics have become limited by the inward flux of oxidized quinone across LH1, a dark lapse of a few hundred milliseconds is expected to replenish significantly the proximal pool. This should appear as a fast phase when resuming the illumination, which was not observed. We thus concluded that there is no proximal pool besides Q B itself and favor the scheme shown in Fig. 11B. The point here is that the LH1 barrier is situated directly against the mouth of the Q B pocket: there may be other quinones inside the LH1 ring, but they have no particularly fast access to the RC pocket.
We also investigated the effects of PufX deletion on the diffusion of quinones on a broader scale by estimating the time required for a quinol generated on the RC to reach the bc 1 complex. For the reduction of heme b 561 , we measured a half-time of ϳ14 ms in membranes of the PufX Ϫ mutant (compared with 7 ms for the WT) and observed an initial lag of ϳ2 ms. We estimated that the quinone turnover time in the PufX Ϫ mutant was ϳ3.5 ms. Compared with the estimate of 1.6 ms for the WT, this implies that the additional time required for crossing the LH1 barrier twice (inwards and outwards) is ϳ2 ms, or 1 ms for a single passage. Whereas this can account for the lag period observed for heme b 561 reduction in the mutant, it clearly does not explain the observed increase from 7 to 14 ms. This indicates that the diffusion time of the quinol, once it has come out of LH1 and until it reaches the bc 1 complex, is significantly lengthened in the mutant. This finding is consistent with the concept of a supercomplex arrangement (32) in the WT membranes, associating one bc 1 complex with an RC⅐LH1 dimer (or, more likely, since monomeric bc 1 is believed to be inactive, a bc 1 dimer with two RC⅐LH1 dimers (33)). This arrangement is disrupted in the PufX Ϫ mutant, resulting in a greater average distance (and/or more obstacles) between the two types of complexes. A similar conclusion was drawn in a recent study (33) on living cells of the two strains, where the reverse process was investigated, i.e. the diffusion of an oxidized quinone from the bc 1 complex to the RC in the context of a reduced pool.
According to our results, the closed LH1 ring is far from constituting a tight barrier to quinone diffusion. The crossing time (per quinone/quinol) is ϳ1 ms at room temperature. At Ϫ22°C, we found turnover times of ϳ200 and 360 ms for the WT and mutant, respectively. Applying the same reasoning as above, this would mean a crossing time of ϳ80 ms. Therefore, the LH1 ring is still not sealed at Ϫ22°C. This suggests either that the association enthalpy of the ␣␤-heterodimers is rather low, allowing substantial "breathing" at low temperature, or that the ring structure presents sufficient openings for the passage of quinones. Clearly, a good permeability of the LH1 ring to quinones is required in bacterial strains endowed with a native closed ring. We have shown that R. sphaeroides is no exception in this respect and that another explanation must be found for the non-photosynthetic phenotype of PufX Ϫ .
Alterations of the RC-The accepted view concerning the PufX Ϫ mutant was that the RC was unaffected by the deletion of PufX and that its handicap was the permeation of quinones through LH1. It was thus a surprise to realize that a number of RC features concerning the Q B pocket were in fact modified in the mutant. We found an increased stabilization of the Q B Ϫ state, a decreased affinity for the Q B pocket inhibitors, and a 6-fold slower rate for the second electron transfer (decay of the Q A Ϫ Q B Ϫ state). In contrast, the P ϩ Q A Ϫ recombination rate and the E m of Q A were the same as in the WT.
Our results on the rate of P ϩ Q B Ϫ recombination do not fully agree with those recently reported by Francia et al. (34). These authors observed a relatively slow recombination (t1 ⁄2 Ϸ 2.3 s) in RC⅐LH1 complexes for WT dimers as well as for PufX Ϫ monomers, in comparison with the faster rate (t1 ⁄2 Ϸ 0.69 s) observed in isolated RCs without LH1 (experiments performed at pH 7.8). Their conclusion is that the interaction of the RC with LH1, irrespective of the presence of PufX, causes a stabilization of the semiquinone Q B Ϫ . In our experiments, the P ϩ Q B Ϫ recombination rate in fresh WT RC⅐LH1 complexes (monomeric or dimeric), with t1 ⁄2 Ϸ 1.2 s (at pH 8), is close to that observed in chromatophores of PufX-containing strains (determined using strain CYC17 (35) devoid of cytochrome c 2 and of its isoform). On the other hand, we found a 4-fold slower rate (t1 ⁄2 Ϸ 4.9 s) in the RC⅐LH1 complexes from the PufX Ϫ mutant. Thus, we believe that the LH1 interaction with the RC is dependent on the presence of PufX. We have no discrepancy, however, regarding the fact that the P ϩ Q B Ϫ recombination rate is faster in isolated RCs. (This FIG. 11. Two models for quinone access to the RC in the PufX ؊ RC⅐LH1 complex. A, a fraction of the quinone pool is present within the internal space between the LH1 ring and the RC, and these quinones have direct access to the Q B pocket. B, LH1 is applied against the mouth of the Q B pocket, and there is no internal pool directly accessible to Q B . This does not exclude the possible presence of quinones in the internal space, as pictured in B. was also observed for R. capsulatus (26).) There are a number of functional differences between isolated RCs and membranes (21,23), and further investigation is required to determine which features are specifically controlled by the interaction with LH1 or by the lipid/detergent environment.
Can the increased stabilization of Q B Ϫ in the PufX Ϫ strain have significant functional consequences? Titrations of the redox couples involved in the Q A Q B system were reported by Rutherford and Evans (36). When lowering the ambient redox potential, one first observes the appearance of a semiquinone wave, ascribed to the Q B /Q B Ϫ couple, with E m Ϸ 40 mV (at pH 8). A semiquinone decrease then takes place around Ϫ40 mV, ascribed to the second reduction step (Q B Ϫ /Q B -H 2 ). A final reduction wave occurs around Ϫ80 mV, reflecting Q A reduction. These E m values predict an equilibrium constant for the second electron transfer (K sec ϭ [Q A Q B -H 2 ]/[Q A Ϫ Q B Ϫ ]) of ϳ5. If such is the case, the decay kinetics of Q A Ϫ Q B Ϫ should be biphasic, with a fast phase of relative amplitude 5/6 reflecting the relaxation of the above equilibrium and a slower 1/6 phase caused by the dissociation of Q B -H 2 . Our results with WT membranes (Fig.  10C, open circles) are consistent with this prediction (40-and 250-s phases with 80:20 weights). In the PufX Ϫ membranes, the higher potential of the Q B /Q B Ϫ couple 4 indicated by the slower P ϩ Q B Ϫ recombination is expected (all things being equal, e.g. the binding energy for the quinol form) to result in a lower potential for the Q B Ϫ /Q B -H 2 couple, thus decreasing K sec , possibly below 1. The slowing of the second electron transfer observed in the mutant (Fig. 10C, closed circles) may tentatively be explained in this manner. Rather than a slowing of the electron transfer rate proper, this would be due to an increase in the slow phase reflecting Q B -H 2 release. We could not resolve a fast phase, however, which, in this framework, implies that K sec has indeed decreased much below 1, so the reoxidation of Q A Ϫ is entirely limited by the release of the quinol. If the above hypothesis is correct, it may explain the "anomalous" accumulation of Q A Ϫ during the photoreduction of the pool in the PufX Ϫ mutant. In domains where a large fraction of the pool is reduced, there would be a significant fraction of Q A this faulty interaction by nucleating an appropriate supramolecular arrangement. This may also be the role of polypeptide W in R. palustris, resulting in a different structure of the RC⅐LH1 complex. If this view is correct, this problem must have been solved in another way by bacterial strains with a closed LH1 ring and no equivalent of PufX or polypeptide W, such as B. viridis, R. rubrum, R. photometricum, etc. This could be achieved by modifying the RC⅐LH1 interface or by reinforcing the Q B pocket.