Chimeric Fusions of Subunit IV and PetL in theb 6 f Complex ofChlamydomonas reinhardtii

The cytochromeb 6 f complex of Chlamydomonas reinhardtii contains four large subunits and at least three small ones, PetG, PetL, and PetM, whose role and location are unknown. Chimeric proteins have been constructed, in which the C terminus of subunit IV is fused to either one or the other of the two putative N termini of PetL. Biochemical and functional analysis of the chimeras together with mass spectrometry analysis of the wild-type (WT) complex led to the following conclusions: (i) neither a free subunit IV C terminus nor a free PetL N terminus is required for assembly of theb6f complex; (ii) the first AUG codon in the sequence of the gene petL is used for initiation; (iii) the N terminus of WT PetL lies in the lumen; (iv) in the WT complex, the N terminus of PetL and the C terminus of subunit IV are within reach of each other; (v) the purifiedb 6 f complex from C. reinhardtii contains an eighth, hitherto unrecognized subunit, PetN; and (vi) the ability to perform state transitions is lost in the chimeric mutants, although (vii) the Q-cycle is unaffected. A structural hypothesis is presented to account for this peculiar phenotype.

Cytochrome b 6 f catalyzes electron transfer from plastoquinol to a hydrosoluble acceptor (plastocyanin or cytochrome c 6 ), while building up a transmembrane proton gradient. The b 6 f complex is found in higher plants, in algae, and in cyanobacteria (1). Purified cytochrome b 6 f from the freshwater unicellular alga Chlamydomonas reinhardtii is a superdimer (2). Each "monomer" contains four large subunits: cytochromes f and b 6 , subunit IV (suIV) 1 and the Rieske iron-sulfur protein (3). Cytochrome b 6 and suIV are homologous to the N-and C-terminal moieties, respectively, of cytochrome b; the chloroplast Rieske protein is homologous to its mitochondrial homonym. Cytochrome f and cytochrome c 1 , despite their functional similarities, are not evolutionarily related (see Ref. 1). Purified preparations of cytochrome b 6 f also contain several very small subunits (ϳ4 kDa), which have no homologues in cytochrome bc 1 . Each of them is thought to span the membrane as a single ␣-helix. Subunits PetG and PetL have been encountered in all b 6 f complexes studied thus far (4 -7). Subunit PetM (formerly called PetX) has been found in C. reinhardtii (3,5,6,8), in higher plants (6), and in cyanobacteria (9). Subunit PetN hitherto has been observed only in Nicotiana tabacum (10), but a gene encoding a homologous protein is present in the nuclear DNA of C. reinhardtii (11). The role of the 4-kDa subunits is uncertain. Deletion of the petG gene in C. reinhardtii (12) or of the petN gene in tobacco (10) prevents accumulation of the b 6 f complex, while deletion of the petM gene in Synechocystis PCC 6803 does not interfere with the assembly of a functional complex (9). As discussed below, the case of PetL is intermediate.
The structure of several forms of cytochrome bc 1 has been solved by x-ray crystallography (reviewed in Ref. 1). As regards the b 6 mh;1qf complex, x-ray data are limited to the extramembrane, catalytic domains of cytochrome f and the Rieske protein (for a review, see Ref. 1). Low resolution electron microscopy projection maps of the whole complex reveal an arrangement of transmembrane helices around the C2 axis of symmetry of the dimer that looks similar to that in cytochrome bc 1 , making it possible to propose tentative positions in the map for most of the b 6 f transmembrane helices (see Ref. 13). Since cytochrome bc 1 does not contain peptides homologous to the small b 6 f subunits, its structure on the other hand is of little help in trying to understand their arrangement and role in the b 6 f complex; yet, some of the structural divergences between the two types of cytochromes, including their dissimilar subunit and prosthetic group (see Ref. 14) complements, must bear on functional differences.
One of those, which is particularly relevant to the present work, is the role of cytochrome b 6 f in the so-called state transitions, a regulatory process whereby photosynthetic organisms balance the supply of excitons between the reaction centers of the two photosystems (PS) (15,16). Transition from state 1 to state 2 results from the transfer of a fraction of the outer PSII light-harvesting complex (LHCII) to PSI, a process triggered by the phosphorylation of LHCII (16,17). During a state 2 3 state 1 transition, LHCII is dephosphorylated and reassociates with PSII (16). Modulation of the phosphorylation state of antenna proteins result from the opposite actions of an LHCII kinase, the activation of which is redox-dependent (18), and a phosphatase, which is generally considered to be permanently active (19). Recent data, however, have suggested a possible regulatory role of an immunophilin-like protein (20). Phosphorylation is activated by the reduction of the plastoquinone (PQ) pool (18,21) and requires the presence of cytochrome b 6 f (22,23). The nature of the kinase is still obscure, although its presence has been reported in partially purified preparations of higher plant cytochrome b 6 f complexes (24). In Arabidopsis thaliana, the consequences of expressing antisense RNAs (25) suggest the involvement in state transitions of a family of thylakoid-associated, presumably transmembrane, kinases (thylakoid-associated kinases) (26). Although the molecular mechanism by which the redox state of the PQ pool controls the kinase is not known, it has been shown in vitro with thylakoid preparations from spinach (27,28) and in vivo in C. reinhardtii cells (23) that it depends on plastoquinol (PQH 2 ) binding to the oxidizing (Q o ) site of the cytochrome b 6 f complex.
The original aim of the present work was to gather information about the location and transmembrane topology of subunit PetL. PetL is strictly required neither for the accumulation nor for the function of cytochrome b 6 f; in its absence, however, the complex becomes unstable in vivo in aging cells and labile in vitro (7). The mRNA sequence deduced from that of the chloroplast gene petL features two possible AUG codons (7). The N terminus of PetL being blocked (3), it is not known which is used for initiation. The distribution of basic residues in the predicted sequence of PetL suggests that, whatever the N terminus is, it is likely to lie in the thylakoid lumen (7). If this prediction is correct, the C terminus of suIV and the N terminus of PetL lie in the same subcellular compartment. In the present study, we have fused the genes coding for suIV (petD) and for PetL by linking either the first or the second of the putative initiation codons for PetL to that coding for the last residue of suIV ( Fig. 1) and examined the expression and accumulation of the chimeric constructs. In vivo functional analysis using time-resolved spectroscopy and fluorescence measurements revealed unusual properties; the chimeric mutants are unimpaired as far as the Q-cycle is concerned, but their state transitions are blocked. To narrow down the range of possible structural interpretations of these observations, the length of mature PetL has been directly investigated by mass spectrometry analysis of the WT complex. The phenotype of the chimeric strains provides interesting insights into the nature of transconformations that could account for the activation of the kinase. In the course of the mass spectrometry study, evidence was also obtained regarding the presence in Chlamydomonas b 6 f of a hitherto unrecognized subunit, PetN, which was confirmed immunologically.

EXPERIMENTAL PROCEDURES
Materials-Sources of chemicals not indicated were as described in Ref. 3.
Strains, Media, and Growth Conditions-A WT C. reinhardtii strain (mtϩ) derived from strain 137c and a ⌬petL deletion strain (7) were used as controls. The deletion strain ⌬petD (mtϩ) (29) was used as recipient strain in chloroplast transformation experiments. WT and mutant strains were grown on Tris acetate-phosphate medium (pH 7.2) at 25°C under dim light (5-6 microeinsteins⅐m Ϫ2 ⅐s Ϫ1 ) (30). Cells were harvested during exponential growth phase and resuspended in a minimal medium (31). They were placed in state 1 and state 2 conditions in darkness, either by vigorous stirring to ensure a strong aeration (state 1; Ref. 32) or by the addition of 5 M FCCP (state 2; Ref. 33).
Plasmids, Oligonucleotides, and Mutagenesis-Plasmids encoding chimeric constructs were created by PCR-mediated site-directed mutagenesis. To generate the fusion between the petD and petL genes, plasmid pdD⌬HI.I (29), carrying the entire coding sequence of petD, was used as template in PCRs using oligonucleotides petDDir (CGCG-CTTAAGTTAAGATCTAAAATTTTAAATTTCCCTCTA) and petDRev (CGCGCTTAAGAATAAACCTAAAGTTAAAGAAATATCAA) as primers and the Arrow TM Taq DNA polymerase according to the manufacturer's instructions. The PCR product was digested with AflII at a restriction site (underlined in the sequence) introduced along with the BglII site (indicated in boldface type), and religated onto itself to yield plasmid pdD⌬Fus. Plasmid pR23 (34), which carries the psaC operon (7), was used as template in PCRs using oligonucleotides petLDirL (GCGCTTAAGTATGATTTTTGATTTTAATTATATCCATAT) or pet-LDirS (GCGCTTAAGTATGTTAACAATCACAAGTTACGTAGGT), homologous to the regions of the putative first or second AUG initiation codons, respectively, of the petL gene (see "Results"), and reverse oligonucleotide petLRev (CGCAGATCTCGAGTTAGATAAGTTTTACAAC-TTTTAAAAGACCT) as primer. PCR products were digested with AflII and BglII and cloned into plasmid pdD⌬Fus digested with the same enzyme, yielding plasmids pDLL and pDLS. The sequence of these plasmids was checked.
Plasmids pDLL and pDLS were introduced by biolistic transformation (35) into the chloroplast genome of the deletion strain ⌬petD (29). Phototrophic transformants were selected for growth on minimum medium according to Ref. 29. The resulting mutant strains, DLL and DLS, were in turn used as recipient strains for biolistic transformation by plasmid pycf7::aadA (a kind gift of Y. Takahashi, Okayama University), which carries an aadA cassette conferring resistance to spectinomycin inserted at the SnaBI site within the petL coding sequence (7). Transformed clones were selected on Tris acetate-phosphate medium containing spectinomycin (100 g⅐ml Ϫ1 ) and subcloned several times on selective medium until they reached homoplasmy. At least three independent transformed strains were characterized for each construct.
Preparative and Analytical Techniques-Cells grown to a density of 4⅐10 6 ml Ϫ1 were broken in a "bead beater" (Biospec-Products) according to the manufacturer's instructions. The membrane fraction was collected by centrifugation and resuspended in 10 mM Tricine, pH 8, at a chlorophyll concentration of 3 g⅐liter Ϫ1 . For SDS-PAGE, membrane proteins were resuspended in 100 mM dithiothreitol and 100 mM Na 2 CO 3 and solubilized by 2% SDS at 100°C for 1 min. Polypeptides were separated on a 12-18% polyacrylamide gel containing 8 M urea (36). Immunoblotting was performed as described in Ref. 3. The antiserum against PetL (7) was a kind gift of J.-D. Rochaix (Université de Genève). For the present work, antisera were prepared (Neosystem, Strasbourg, France) against peptides covering three regions of the predicted sequence of C. reinhardtii PetN precursor, namely PAAQA-AQEVAMLAEG*, IVQIGWAATCVMFS*, and *FSLSLVVWGRSGL (cf. Fig. 8; the asterisk indicates the site of coupling to the carrier protein). The only antiserum that yielded a positive reaction on immunoblots was that raised against the C-terminal peptide, coupled to ovalbumin via its N terminus. Other antipeptide antisera have been described in Ref. 3. Cytochrome b 6 f purification and was performed as described in Ref. 3.
Optical and Fluorescence Measurement-Fluorescence measurements were performed at room temperature on a home built fluorimeter; samples were excited using a light source at 590 nm, and the fluorescence response was detected in the far red region of the spectrum.
Absorbance measurements were performed at room temperature with a home built spectrophotometer described in Refs. 37 and 38. Cells were resuspended in the presence of 10% Ficoll to avoid sedimentation. The slow phase of the electrochromic signal ("phase b" according to Ref. 39), which is associated with electron transfer through the cytochrome b 6 hemes, was measured at 515 nm, where a linear response is obtained with respect to the transmembrane potential (40). Deconvolution of phase b from the membrane potential decay and calculation of cytochrome f redox changes were performed as described in Ref. 41.
Protein Phosphorylation Assays-Cells were resuspended in a phosphate-depleted medium containing 1 Ci ml Ϫ1 33 P i . They were treated as described in Ref. 32. Polypeptides were separated by denaturing SDS-PAGE as described above. Autoradiography was performed as described in Ref. 23.
Mass Spectrometry-WT b 6 f complex was purified, and its subunits were separated by SDS-PAGE and transferred onto nitrocellulose sheets (Millipore Corp., Bedford, MA) according to Ref. 3. Samples were localized on parallel lanes by combining immunoblotting with specific antibodies and staining with Ponceau red. The spots of interest were excised and fixed to a stainless steel target using double-sided tape. After several tests with different matrices, ␣-cyanohydroxycinnamic acid (Sigma) was selected as the matrix of choice. Blot pieces were soaked into isopropyl alcohol and covered with a drop of the supernatant of a saturated solution of ␣-cyanohydroxycinnamic acid in acetone. After drying, samples were washed with acetonitrile. Extraction of chlorophyll by acetone prior to mass spectrometry resulted in the loss of part of the low M r peptides, the proportion of the smaller species (peaks around 3.5 kDa) diminishing considerably as compared with heavier ones (peaks around 4 kDa; not shown). All analyses therefore were performed without chlorophyll extraction. MALDI-TOF measurements were carried out on a STR Voyager mass spectrometer (Applied Biosystems, Framingham, CA) equipped with a nitrogen laser (237 nm, 20 Hz). Spectra were acquired in the linear positive mode (accelerating voltage 20 kV, grid voltage 95%), with a delayed extraction time of 300 ns. They were calibrated using a mixture of adrenocorticotropic hormone (residues 7-38; m/z ϭ 3660.19 Da) and bovine insulin (m/z ϭ 2867.80 and 5734.59 Da), which was applied directly to blot pieces.

Construction of C. reinhardtii Mutants Expressing Chimeric
Proteins-Two chimeric proteins were constructed; both of them comprised of a full-length suIV fused, at its C terminus, to the N terminus of PetL ( Fig. 1). They differed with respect to which of two AUG codons was considered as the initiation codon for PetL synthesis. As a result, the last transmembrane helix of suIV was connected to the single putative transmembrane helix of PetL by either a short or a long intervening loop (ϳ20 and ϳ30 residues, respectively). The corresponding plasmids were named pDLS and pDLL. The chloroplast genome of the nonphototrophic ⌬petD strain, which lacks the gene encoding suIV (29), was transformed by either plasmid. Both chimeric constructs yielded phototrophic clones. SDS-PAGE followed by immunoblotting with an anti-suIV antiserum showed that thylakoid membranes prepared from the transformed strains still lacked WT suIV. They accumulated instead a larger protein, whose size correlated with the expected size of the chimeras (Fig. 2). The same protein indeed also reacted with an antiserum directed against PetL (not shown). The restoration of phototrophy therefore is not due to the presence of WT-like suIV but to the fact that either of the two chimeric proteins can substitute for it.
In the absence of suIV, most other b 6 f subunits are synthesized at the WT rate but rapidly degraded and therefore do not accumulate (29,42). As expected given the phototrophy of the DLS and DLL strains, expression of either chimeric protein restored accumulation of the other b 6 f subunits (Fig. 2), con-firming that they assembled into a functional complex. A difference between the two strains, however, was consistently observed regarding WT PetL. This subunit, which was expressed in both cases along with the chimeric protein, accumulated to WT levels in the strain expressing the chimera with the short loop, DLS, but not in the presence of that with a long loop, DLL (Fig. 2). Since the b 6 f complex is present and functional in DLL cells, it seemed likely that the chimeric protein could structurally and functionally substitute for both suIV and PetL.
Chimeric PetL Is Able to Stabilize the b 6 f Complex in the Absence of the Endogenous Subunit-In order to test this hypothesis, the chloroplast genomes of strains DLL and DLS were transformed with plasmid pycf7::aadA. This plasmid carries a petL coding sequence disrupted by the insertion of an aadA cassette, which confers resistance to spectinomycin (7). Transformed strains, named DLS⌬ and DLL⌬ depending on the recipient strain, were selected on spectinomycin-containing plates. Both types of strains were phototrophic and grew at a rate similar to that of the WT (not shown). Immunoblots of cells grown exponentially showed that they failed to accumulate either WT-like suIV or PetL (Fig. 2). On the other hand, the DLS⌬ and DLL⌬ strains accumulated the DLS or DLL chimeras, respectively, to levels similar to those observed for suIV in the WT strain (Fig. 2). The same held true for the other b 6 f subunits, indicating that in both cases the whole complex was properly assembled, with, apparently, a ϳ1:1 stoichiometry between the chimera and WT subunits (Fig. 2).
The stability of cytochrome b 6 f is affected in ⌬petL strains obtained by transformation of the WT with the ycf7::aadA plasmid. During exponential growth, the complex accumulates, although to somewhat reduced levels; when cells enter the stationary phase, however, it disappears from thylakoid membranes (7). This behavior suggests that the absence of PetL renders the b 6 f complex more sensitive to proteolytic degradation. The accumulation followed by disappearance of the b 6 f complex in ⌬petL mutants is reflected in their fluorescence induction kinetics upon illumination with actinic light; during exponential growth phase, fluorescence transients do not reach the maximum fluorescence yield obtained by adding the PSII were subjected to SDS-PAGE, and blots were probed with subunit-specific antisera against subunits IV and PetL (see "Experimental Procedures"). Sera were used at a 1:10,000 dilution and detected using the ECL system (Amersham Biosciences, Inc.). The PSII oxygen-evolving enhancer protein 3 (OEE3) protein was used as an internal control.
inhibitor DCMU, a consequence of the PQ pool being reoxidized by cytochrome b 6 f (43); in the stationary growth phase, on the contrary, fluorescence transients rise to a level similar to that measured in the presence of DCMU, due to the near absence of cytochrome b 6 f (Ref. 7 and Fig. 3, A and B).
All chimeric mutant strains, when growing exponentially, exhibited fluorescence induction kinetics similar to those of the WT strain (not shown). This phenotype is consistent with the biochemical data, which show accumulation of cytochrome b 6 f. In aging DLL and DLS mutant strains, fluorescence induction kinetics again reflected WT-like electron transfer (Fig. 3C). Aging DLL⌬ and DLS⌬ strains, on the other hand, behaved differently one from another: in DLL⌬, fluorescence transients indicated WT-like electron transfer, while in DLS⌬ they betrayed a very slow rate of reoxidation of the plastoquinone pool (Fig. 3D). This point was further studied by comparing the accumulation of the main subunits of the b 6 f complex in mutant strains expressing or lacking the endogenous PetL subunit. During the exponential phase (2⅐10 6 cells/ml), all chimeric strains resembled the WT, the presence or absence of WT PetL having little or no effect on the level of accumulation of the other b 6 f subunits (Fig. 4A). In aging cells (9⅐10 6 cells/ml), on the other hand, accumulation of the complex tended to be somewhat lower in strain DLL⌬ and, even more so, in strain DLS⌬ than in either WT or the DLL and DLS mutant strains (Fig. 4B).
Biochemical Stability of Cytochrome b 6 f Complexes Incorporating the DLL Chimeric Protein-PetL-free cytochrome b 6 f complexes containing WT suIV, as accumulated during the exponential phase by ⌬petL mutant strains, are markedly unstable following solubilization; upon sucrose gradient fractionation, they monomerize and release the Rieske protein (7). Cytochrome b 6 f complex from the chimeric mutant strain DLL⌬, on the other hand, was still dimeric and functional after solubilization and purification on sucrose gradient. Subunit distribution along the gradient was similar to that for the WT complex (not shown); the Rieske protein, in particular, comigrated with the other subunits, which is a reliable criterion of the integrity of the complex and its dimeric state (2). Nonetheless, DLL⌬ complexes are more fragile than WT ones, and purification to homogeneity proved problematical.
Mass Spectrometry Analysis of the 4-kDa Subunits-A structural interpretation of the above results depends in part on whether the putative 11-residue N-terminal extension of PetL, which makes up the difference between long and short loops in the chimeras, is part of the mature WT PetL subunit or not. In order to directly probe this point, preparations of purified WT cytochrome b 6 f complex were submitted to SDS-PAGE, and the peptides present in the low M r region were analyzed by mass spectrometry. The results are summarized in Table I and Fig. 5.
Of the three small subunits previously identified in C. reinhardtii b 6 f complex, two yielded identifiable peaks. PetM appeared under two forms, one free and one acetylated. The sequence of the mature subunit starts at the position determined by Edman degradation (3,5,6) and runs to the end of the open reading frame of the petM gene (8). PetG, although its presence has been established both immunologically (3,5) and by Edman degradation (6), was undetectable. PetL yielded two identifiable peptides. The first one was observed at three different masses, namely as its H ϩ , Na ϩ , and K ϩ adducts. It starts with the methionine residue corresponding to the first AUG codon but stops after 30 residues rather than the 43 expected. The second, observed as an  H ϩ adduct only, starts with residue 5 (numbering from the first methionine) and ends at residue 39. It thus appears to be clipped by four residues at both termini. It should be noted that, given that this analysis failed to identify a peptide, PetG, that is undoubtedly present in the purified complex, the fact that no PetL peptide starting with the second methionine residue was recovered cannot be taken as a definite proof that this putative initiation site is not used at all. Interestingly, MALDI-TOF spectra also revealed the presence in purified preparations of WT C. reinhardtii b 6 f of a fourth small subunit, PetN ( Fig. 5 and Table I). Until now, PetN had been identified in tobacco only, with strong evidence that in this organism it is an essential subunit of the b 6 f complex (10). The nuclear genome of C. reinhardtii does contain a gene related to N. tabacum petN (11). Antisera were raised against one C-terminal peptide and two putative Nterminal peptides predicted by the sequence of C. reinhardtii petN (Fig. 6A, boxes). Immunoblots of purified WT b 6 f gave a positive signal with the anti-C terminus serum only (Fig. 6B). Analysis of WT and ⌬petD thylakoid membranes using this serum showed that PetN is absent in cells that do not accumulate the b 6 f complex (Fig. 6B).
State Transitions Are Abolished in the Chimeric Mutants-The occurrence of State Transitions in the chimeric mutants was examined by measuring the fluorescence yield of intact algae in the presence of the PSII inhibitor DCMU (43). PSI, at room temperature, acts as a strong fluorescence quencher (15). Fluorescence emission therefore is proportional to the size of the PSII antenna and inversely proportional to the yield of PSII photochemistry (44). In the presence of DCMU, fluorescence changes during the transition from state 1 to state 2 thus directly reflect the decrease in PSII antenna size. Transitions were elicited in total darkness (see "Experimental Procedures"), in order to be independent of the electron transfer properties of the strains. The DLL⌬ and DLS⌬ mutants were compared (i) with the WT, used as a positive control; (ii) with a strain lacking cytochrome b 6 f, the ⌬petD strain, which undergoes no state transitions (22), as a negative control; and (iii) with the ⌬petL strain (7). The maximal fluorescence yield of the WT strain dropped by about 40% in state 2 as compared with state 1 (Fig. 7A). This reflects the transfer of a major fraction of LHCII from PSII to PSI. The same effect was observed in the case of the ⌬DpetL mutant (Fig. 7B), showing that the absence of WT PetL by itself does not block state transitions. On the contrary, neither the b 6 f-free ⌬DpetD mutant (Fig. 7E) nor the DLS⌬ or the DLL⌬ ones (Fig. 7, C and D) displayed any decrease of fluorescence yield under conditions promoting state 2. Actually, the fluorescence yield increased slightly under these FIG. 5. Mass spectrometry analysis of low M r subunits from purified wild-type b 6 f complex. MALDI-TOF analysis following SDS-PAGE and electroblotting (see "Experimental Procedures"). The mass of each identified peak corresponds uniquely to that calculated for the indicated stretch from the predicted sequences of the four 4-kDa subunits. See Table I. conditions, a phenomenon previously observed in strains locked in state 1 when the PQ pool is fully reduced (33). Very similar results were obtained with the DLL and DLS strains, both of which express WT PetL along with the fusion protein (not shown).
DLL⌬ and DLS⌬ Mutants Fail to Activate LHC Kinase under State 2 Conditions-Fluorescence measurements indicate that in the DLL⌬ and DLS⌬ mutants LHCII is not transferred from PSII to PSI under state 2 conditions. To assess whether locking in state 1 is due to the absence of LHCII kinase activation, we examined in vivo protein phosphorylation. Thylakoid membranes were purified from cells that had been preincubated for 90 min with 33 P i and placed for 20 min under state 1 or state 2 conditions in a 33 P i -free medium (32). Fig. 8 shows the labeling pattern of the thylakoid membrane polypeptides of the WT and of the DLL⌬, DLS⌬, ⌬DpetD, and ⌬DpetL mutants in the 25-40-kDa region. In the WT, the phosphorylation of LHCII polypeptides, LHC-P13 and LHC-P17, increased in state 2 as compared with state 1, whereas the PSII phosphoprotein D2 showed an opposite behavior, as previously reported (32). Consistently with fluorescence measurements, a similar phosphorylation profile was observed in the ⌬petL strain, while a significantly lower level of phosphorylation of LHC-P13 and LHC-P17 was observed in the DLL⌬, DLS⌬, and ⌬petD mutants under conditions that promote state 2. This phosphorylation profile is typical of state 1 (32). In the WT and ⌬petL strains, several minor phosphoproteins were detected in the 15-20-kDa region under state 2 conditions. Those included PetO, a protein that interacts with cytochrome b 6 f (45). None of these polypeptides showed significant phosphorylation in the DLL⌬, DLS⌬, and ⌬petD mutants (Fig. 8).
The Cytochrome b 6 Fig. 3 indicate that the overall connection between PSII and PSI via the b 6 f complex is functional in the chimeric mutant strains. A blockade of state transition was therefore unexpected, all b 6 f mutants affected in state transitions isolated so far exhibiting impaired redox activity (reviewed in Ref. 46). We therefore measured the rate of several reaction steps of the Q-cycle in order to check whether the inhibition of state transitions might be associated with some functional deficiency not involving the rate-limiting reaction (which is the only one affecting the fluorescence measurements presented in Fig. 3).

f Complex of the DLL⌬ and DLS⌬ Mutants Exhibits WT-like Plastoquinol Oxidase Activity-The fluorescence measurements presented in
The catalytic cycle of cytochrome b 6 f comprises oxidation of PQH 2 at a luminal (Q o ) site of the protein complex and reduction of PQ at a stromal (Q i ) site. According to the "Q-cycle" hypothesis (47,48), PQH 2 oxidation results in injecting electrons into two distinct electron transfer chains, one comprising the Rieske protein and cytochrome f and the other involving the two b 6 hemes. This process can be studied spectroscopically by measuring the redox changes of cytochrome f (49). In addition, since the oxidation of the b 6 hemes results in a transfer of charges across the membrane, electron flow through cytochrome b 6 generates a measurable increase of transmembrane potential in the millisecond time range (causing the slow phase, called "phase b," of the electrochromic signal; see Ref. 39). Fig. 9 shows the results of such measurements in the WT and in the DLL⌬ and DLS⌬ strains. The slow phase of the electrochromic signal is shown in Fig. 9, A-C. Amplitudes are normalized to that of the fast phase ("phase a"), which, when PSII activity is inhibited by the addition of DCMU and hydroxylamine, is driven solely by PSI and is therefore proportional to the number of positive charges injected into the plastocyanine pool (43). Under our experimental conditions, reduction of the PQ pool is assured at the expense of cell metabolism (50), and the availability of PQH 2  respond to negative and positive absorption changes, respectively. Fig. 9 clearly indicates that the electron transfer properties of the complex were not affected by the DLL⌬ and DLS⌬ mutations; the rates of the single reactions (t1 ⁄2 Ϸ 5-6 ms) were comparable with those observed in the WT (41). In green algae incubated in the dark, an electrochemical proton gradient builds up (51), which selectively slows down the reactions occurring at the Q o site (38,51). Such a gradient was also observed in the two mutants, as indicated by the effects of the protonophore FCCP, the addition of which accelerated phase b (Fig. 9, A-C, circles) and the reduction of cytochrome f (Fig. 9, D-F, circles) in much the same manner as observed in the WT (t1 ⁄2 Ϸ 2 ms). We conclude, therefore, that the main electron and proton transfer steps of the b 6 f catalytic cycle are not affected in the DLL⌬ and DLS⌬ mutants.

Ability of suIV-PetL Chimeras to Substitute Structurally and Functionally for the Two Subunits
In all mutant strains, the chimera obtained by fusing suIV and PetL was expressed at a level comparable with that of suIV in WT strains. Immunoblots showed no trace of any proteolytic cleavage that could have regenerated WT-like subunits. In strains lacking suIV but retaining endogenous PetL, expression of any of the two chimeras restored phototrophy, accumulation of the b 6 f subunits, and fluorescence transients and electron transfer rates characteristic of a native-like complex. Either of the two constructs therefore is able to structurally and functionally substitute for suIV, despite the presence of an unnatural C-terminal extension. The intensity of the bands containing the chimeric proteins is consistent with a ϳ1:1 stoichiometry of the chimera with respect to the other large subunits, suggesting that any excess fusion protein is degraded, in the same manner as nonassembled WT suIV is degraded in WT cells. There is little doubt that the suIV-like moiety of the chimeric proteins must fold and assemble correctly and that it functionally replaces the missing suIV within the complex.
What then is the fate of the PetL-like moiety of the chimeras? Our data indicate that this is a function of (i) the length of the intervening loop and (ii) the presence or absence of the WT PetL subunit. In the DLS strains, which contain the short-loop construct and express WT PetL, the latter accumulates to WT-like levels. It is therefore not accessible to proteolysis, as is the case with nonassembled PetL, and must be incorporated stoichiometrically into the modified complexes. Immunoblots indicate that the PetL-like extension of the chimera is not proteolytically trimmed. The DLS complexes therefore must comprise two copies of the PetL sequence, one free and one fused to suIV. The latter is likely to form an extra transmembrane helix. The fact that it interferes neither with the assembly nor with the functioning of the complex is compatible with the outlying position of the third helix of suIV that is suggested by electron microscopy data (Ref. 13; see below, Fig. 10). It is an interesting observation that, although it cannot occupy its proper position in the complex, this extra PetL-like sequence segment is not degraded, while free, nonassembled PetL is (7). Among several possible interpretations, a simple one would be that degradation of free WT PetL starts at the N terminus (i.e. as shown below, from the lumen).
In constructs with a long loop, on the contrary, there is every evidence that the PetL-like moiety can, and does, displace and substitute for the endogenous peptide: (i) in strains that coexpress WT PetL along with the long-loop chimera (DLL), PetL accumulates to very low levels as compared with that in WT cells or in cells harboring the short-loop construct; the most straightforward interpretation of this phenomenon is that the C-terminal moiety of the long-loop chimera occupies the binding site of PetL, which, not being able to assemble, becomes proteolytically degraded; (ii) in strains that express the longloop construct but no WT PetL (DLL⌬), a functional complex is nevertheless assembled; it is much more stable than PetL-free complexes both in vivo (persistence in aging cells) and in vitro (resistance to detergent). Altogether, these observations strongly suggest that the PetL-like moiety of the long-loop construct is able to bind to the site normally occupied by PetL and, to a large extent even if not absolutely with the same efficacy, to exert its stabilizing effect on the complex. Biochemical and spectroscopic data offer evidence that the short-loop construct also confers functionality to PetL-free complexes; the stability of the DLS⌬ complexes in aging cells, however, appears marginal. The affinity of the PetL-like moiety of the short-loop chimera for the binding site of PetL indeed must be lower than that of the long-loop one, since, at variance with the latter, it is unable to efficiently compete with endogenous PetL for its binding site and, thereby, to provoke its degradation.

Implications for the Transmembrane Topology of PetL
The ability of the PetL-like moiety of the long-loop chimera to occupy the binding site of PetL and to functionally substitute for it is a strong indication that this region of the chimera must adopt the same transmembrane topology as PetL does in WT b 6 f. Because the suIV-like moiety of the construct substitutes for suIV, it also must adopt the same topology as the parent subunit, which places the fusion point in the lumen (1,52). This result is consistent with the PetL-like moiety of the chimera and, therefore, WT PetL itself, lying with its N-terminal end in the lumen (7). Can the opposite orientation, however, be totally ruled out? There are two conceivable types of events that could permit the long-loop chimera to generate functional b 6 f complexes even if WT PetL lies with its N terminus in the stroma. Both assume that the PetL-like moiety imposes this topology to the corresponding region of the chimera. Such a phenomenon has been observed in polytopic proteins whose transmembrane topology had been genetically tampered with, such as the MalF subunit of the maltose transporter (53) or lactose permease (Ref. 54; for a discussion, see Ref. 55). The first mechanism assumes the stoichiometry of the chimera to the other b 6 f subunits to remain 1:1, while the second one requires it to be 2:1.
Case 1-A fusion protein present as a single copy per b 6 f monomer would have to insert with its N terminus in the stroma, to fit the natural topology of suIV, and with its C terminus in the lumen, to suit the postulated orientation of PetL. It would not, therefore, adopt the expected four-helix topology. This can occur in two ways: either (i) one of the helices in the suIV-like region does not insert, or (ii) the loop forms an additional transmembrane helix. The first hypothesis seems very improbable: (a) the right positioning of the first two transmembrane helices of suIV is a sine qua non condition for the luminal loop that links them, which forms part of the Q o site, to adopt a correct conformation (23); (b) mutations in the seventh helix of cytochrome b, which is the homologue of the third helix of suIV, affect the assembly of the mitochondrial cytochrome bc 1 complex (56); and (c) topological signals in the PetL-like moiety of the fusion protein would tend to direct its C-terminal end, not the N-terminal one, toward the stroma (7); given that the N-terminal moiety of the chimera is probably already inserted by the time the C-terminal one is released from the ribosome (insertion of chloroplast-encoded subunits appears to be mainly cotranslational; see e.g. Refs. 57 and 58 and refer-ences therein), it is difficult to understand either why or how the PetL-like moiety of the chimera would force the upstream suIV-like one to insert or rearrange with an aberrant topology. The second hypothesis also appears quite far fetched, the hydrophilic character of the loop, and, in the case of the DLS construct, its short length, making it very improbable that it should have any tendency to form an additional transmembrane helix.
Case 2-A second mechanism to be considered is based on the fact that many integral proteins can tolerate the presence of supernumerary transmembrane helices without loss of function (reviewed in Refs. 55 and 59 -61), as is actually observed here in the case of the DLS strains. In b 6 f complexes incorporating two DLL chimeras per monomer, one of the two chimeric molecules could feature four helices and have its two termini in the stroma (Fig. 1), substituting for suIV, while the other would either have a distorted topology, as hypothesized above, or adopt a fully inverted orientation, providing a functional PetLlike region. This kind of mechanism cannot be a priori ruled out (it may well account for erroneous topological conclusions drawn from fusion experiments carried out on cytochrome b 559 (62)). It holds, however, very little appeal in the case of suIV-PetL chimeras. First, as discussed above, insertion of the fused PetL-like sequence with its N-terminal end in the stroma seems unlikely to occur; second, the presence of two copies of the chimera per complex, although difficult to rule out, is not supported by any data (see below).
It seems safe, therefore, to conclude that the transmembrane orientation of WT PetL must be that originally postulated (7), namely that its N terminus faces the lumen.

PetL Length and Location in the Complex
The length of the intervening loop clearly has a strong effect on the ability of the PetL-like moiety of the chimeras to compete with the WT PetL subunit. Whether this can be taken as an indication that, in the three-dimensional structure of cytochrome b 6 f, PetL lies far away from the last transmembrane helix of suIV depends on which AUG codon is used as an initiation site for the translation of PetL. It was not known, at the onset of this work, whether C. reinhardtii PetL contains or not the sequence segment predicted by the gene sequence upstream of the initiation site used in most other photosynthetic organisms (7). To ensure that at least one of the constructs would contain the complete sequence of WT PetL, the extension of the loop was therefore given the sequence of this N-terminal region. In view of the different properties exhibited by the long-loop and short-loop strains, the length of mature WT PetL was examined using mass spectrometry. While the two PetLderived peptides detected had clipped extremities, the first initiation codon clearly had been used for their synthesis. The higher efficiency of long-loop constructs at competing with WT PetL and stabilizing PetL-free complexes then does not necessarily reflect spatial constraints; it may also be due to the N-terminal extension of PetL being functionally important. The functionality of the b 6 f complex in the short-loop DLS⌬ strains and its marginal but improved stability in vivo as compared with ⌬petL complexes indicate, on the other hand, that the extension is at least partially dispensable.
While there is no doubt that the chimeras can structurally and functionally substitute for both suIV and PetL, a structural interpretation of this phenomenon again depends on the number of chimeras per monomer. The simplest and most likely hypothesis is that a single chimera molecule occupies simultaneously both the suIV and the PetL sites. An alternative is that two distinct chimeras with the same transmembrane topology be involved, one providing its suIV moiety and the other the PetL one, which would leave the distance between the two sites undetermined. One may entertain doubts at the idea of the b 6 f dimer accommodating eight redundant transmembrane helices without its functionality being compromised. Such a model, however, is difficult to rigorously rule out. Immunoblots give no indication that the stoichiometry of the chimera to the other b 6 f subunits is 2:1 rather than 1:1; however, the ECL reaction used in the present study is far from being a quantitative assay. It could also be argued that proteolytic removal of unassembled suIV is so efficient that it is unlikely that chimera molecules with only the PetL moiety inserted into the complex would be totally spared and would not generate any fragments, which would have been detected in immunoblots. The argument holds some appeal, but it is weakened by the fact that, in DLS strains, the redundant PetL-like extension of the chimera, which is undoubtedly present, is not degraded.

Inhibition of State Transitions
The fusion of suIV and PetL inhibits state transitions without affecting the electron transfer efficiency of the complex. This phenotype is novel, impairment of state transitions being associated with the loss of PQH 2 oxidizing activity in all b 6 f mutants hitherto studied (22,23).
The WT-like electron transfer properties of the chimeric mutants explain their ability to grow phototrophically. This phenotype is consistent with previous suggestions that PetL plays essentially a structural function and is not involved in the catalytic cycle of the complex (7) and with the idea that the C terminus of suIV is not directly involved in PQH 2 binding and oxidation. The latter is inferred from the comparison of subunit sequences in the b 6 f and bc 1 complexes. The two cytochromes share the same catalytic cycle (reviewed in Refs. 1, 63, and 64). Whereas very few changes are tolerated in substrate-binding sites, a larger variability affects other sequence regions (1). This is indeed the case of the site of gene fusion in our constructs; the C terminus of suIV is free in the b 6 f complex (1,63,64), while the seventh helix of cytochrome b (its homologue in the bc 1 complex) is connected to the eighth and last transmembrane helix (65,66). Fusing PetL at this position actually recreates a local topology similar to that in the corresponding region of cytochrome b.
Quinol binding to cytochrome b 6 f is not modified in the mutants. The impairment of state transitions thus suggests that they are affected in the transduction of the activating signal from the Q o site to the kinase. This might take place at two levels: (i) the interaction of the kinase with the b 6 f complex and (ii) its diffusion away from the cytochrome, where LHCII phosphorylation takes place (reviewed in Ref. 46). The absence of PetO phosphorylation in the mutants suggests that the fusion of suIV and PetL inhibits state transitions at step (i). At variance with LHCII, this b 6 f-associated peptide indeed is phosphorylated upon PQH 2 binding to Q o even when diffusion of the kinase is blocked (reviewed in Ref. 46). The lack of phosphorylation of PetO in the DLL⌬ and DLS⌬ mutants under State 2 conditions therefore suggests that the kinase is unable to interact with the b 6 f complex of the mutants in a way leading to its activation.

A Mechanism for LHCII Kinase Activation in Thylakoid Membranes
The phenotype of the chimeric mutants suggests that at least one of the two fused subunits is involved in the docking of the LHCII kinase to (or in its activation by) the WT b 6 f complex. A direct involvement of PetL in kinase activation seems very unlikely; state transitions occur in the ⌬petL mutant, which lacks this subunit, and they are inhibited in the DLS strain, where a WT copy of PetL occupies its binding site. On the contrary, a role of suIV in both kinase binding and activation appears more readily conceivable. One unsolved issue in understanding LHCII kinase activation is the mechanism by which PQH 2 binding to Q o , on the luminal site of the membrane, activates an enzyme that operates in the stroma. One model involves conformational changes of the Rieske subunit (23,27,28), whose flexibility has been demonstrated in both the bc 1 (65,66) and b 6 f (67, 68) complexes. Recently, we have proposed (69) that the activating signal is transduced to the active site of the kinase via conformational changes occurring in the transmembrane region of the cytochrome b 6 f. Recent electron microscopy data indeed suggest that such changes, which are peculiar of the b 6 f complex, accompany the movements of the Rieske protein catalytic domain (13). They occur in two main regions of the protein (Fig. 10): the monomer to monomer interface (i.e. close to the region where the redox cofactors are probably positioned) and a more outlying region of the dimer (13). Such movements might promote the activation of the kinase, either by transducing directly the activating signal to its stromal catalytic domain or by stabilizing an interaction between a kinase transmembrane domain and the b 6 f complex; the existence of a transmembrane helix has been proposed at least in the case of the thylakoid-associated kinases (26), which are probably involved in state transitions (25), and the outermost region of conformational changes would be readily accessible to diffusing transmembrane proteins (cf. Fig. 10).
In the frame of this model, the phenotypes of the chimeric mutants can be tentatively explained. This is illustrated in Fig.  10, which shows projection maps of the cytochrome b 6 f complex calculated in the absence (gray) and presence (white) of stigmatellin (redrawn from Ref. 13). The probable position of the three helices of suIV is indicated as I, II, and III, as deduced from the comparison of the projection maps of the b 6 f and bc 1 complexes calculated at the same resolution (Ref. 68; see Ref. 13 for a detailed discussion). It can be observed that a rearrangement seems to take place in the vicinity (*) of these three helices upon the addition of a ligand of the Q o site, stigmatellin. In particular, a new density appears close to helix III, which carries the C terminus of suIV. The absence of state transitions in the suIV-PetL chimeras might be due to the linker peptide interfering sterically with movements occurring in this region and thereby preventing, directly or indirectly, either the dock-ing of the kinase or its activation. This effect would be independent of the presence and position of PetL subunit and of the length of the linker peptide, as observed in the present work. A prediction of this model is that similar effects could be expected upon fusion to the C terminus of suIV of any other peptide likely to form a transmembrane helix, possibly even of a soluble peptide.
PetN: A Fourth Small Subunit of the C. reinhardtii b 6

f Complex
Mass spectrometry experiments performed in the course of this work revealed the presence, in preparations of WT C. reinhardtii b 6 f complex, of a fourth small subunit, PetN, homologous to that previously identified in N. tabacum (10). A serum raised against a peptide featuring the predicted C-terminal sequence of C. reinhardtii PetN (11) confirmed the presence of PetN in purified b 6 f preparations and in thylakoid membranes from WT cells. It also demonstrated its absence in cells that do not accumulate the complex. PetN therefore is a bona fide subunit of C. reinhardtii cytochrome b 6 f, inasmuch as it is present in the purified complex and it does not accumulate in its absence.
Genes homologous to Tobacco petN have been identified in cyanobacteria (70) and in all chloroplast genomes analyzed to date. The high degree of conservation of the open reading frame (Fig. 8A) suggests that it codes for a functionally important subunit. In tobacco, indeed, knocking petN out yields plants that are photosynthetically incompetent (10). The predicted mature sequence of C. reinhardtii petN is very similar to that of its chloroplast-encoded homologues (Fig. 8A). It is, however, preceded by a transit peptide, which is exactly conserved in the closely related species Volvox carteri. If the transit peptide is cleaved by the thylakoid processing peptidase (cf. Fig. 8A), the N terminus of mature PetN must lie in the thylakoid lumen and its C terminus in the stroma. The position of the cleavage site, however, remains ambiguous. The peptide identified by MALDI-TOF features the predicted C terminus of PetN, which is conserved in all photosynthetic organisms (Fig. 8A). Its N terminus, on the other hand, is not that expected from the consensus sequence for the thylakoid processing peptidase, which typically cleaves after an AXA motif (57,71). It seems likely that, as observed for PetL, the peptide identified by mass spectrometry does not correspond to the full-length mature protein. Upstream of the N terminus observed by MALDI-TOF lie several AXA motifs. Two of them (underlined in Fig. 8A) are close to the N terminus of the MALDI-TOF fragment and consistent with the specificity of the peptidase. The absence of cross-reaction with an antiserum raised against the synthetic peptide PAAQAAQEVALMAEG (Fig. 8A, dotted box) would be consistent with the mature protein starting only after the AQA triplet (Fig. 8A, solid arrow). Further studies however will be required to directly establish the position of the cleavage site.

Conclusion
In summary, the experiments reported in the present work lead to the following conclusions. (i) Neither a free suIV C terminus nor a free PetL N terminus is required for the b 6 f complex from C. reinhardtii to assemble and function. This observation opens up interesting prospects for multiple tagging of the complex, as well as for the construction of other fusion proteins; it is of interest also that an extra copy of PetL, tethered to the C terminus of suIV, is protected from proteolytic degradation even though it is prevented by endogenous PetL to integrate into the complex. (ii) Initiation of PetL synthesis starts at the first of the two AUG codons. (iii) PetL lies with its N terminus in the lumen. (iv) In the three-dimensional struc- ture of cytochrome b 6 f, the N terminus of PetL and the C terminus of suIV must be within reach of each other. (v) Cytochrome b 6 f complexes incorporating suIV-PetL chimeras correctly assemble and transfer electrons efficiently. (vi) Nevertheless, they are unable to carry out state transitions; it seems possible that the linker peptide interferes with movements occurring in the complex and thereby prevents the docking of the kinase or its activation. (vii) Finally, the purified b 6 f complex from C. reinhardtii contains an eighth, hitherto unrecognized, subunit, PetN.