The labile interactions of cyclic electron flow effector proteins

The supramolecular organization of membrane proteins (MPs) is sensitive to environmental changes in photosynthetic organisms. Isolation of MP supercomplexes from the green algae Chlamydomonas reinhardtii, which are believed to contribute to cyclic electron flow (CEF) between the cytochrome b6f complex (Cyt-b6f) and photosystem I (PSI), proved difficult. We were unable to isolate a supercomplex containing both Cyt-b6f and PSI because in our hands, most of Cyt-b6f did not comigrate in sucrose density gradients, even upon using chemical cross-linkers or amphipol substitution of detergents. Assisted by independent affinity purification and MS approaches, we utilized disintegrating MP assemblies and demonstrated that the algae-specific CEF effector proteins PETO and ANR1 are bona fide Cyt-b6f interactors, with ANR1 requiring the presence of an additional, presently unknown, protein. We narrowed down the Cyt-b6f interface, where PETO is loosely attached to cytochrome f and to a stromal region of subunit IV, which also contains phosphorylation sites for the STT7 kinase.

Photosynthesis converts solar energy into electrochemical energy, which eventually enables CO 2 fixation by the Calvin-Benson-Bassham cycle enzymes (1). To ensure assimilation of fixed carbon, ATP and NADPH serve as energy currencies that are produced during photosynthetic electron transfer. Two photosynthesis modes exist in higher plants and green algae (2), termed linear electron flow (LEF) 2 and cyclic electron flow (CEF). Their contribution to total electron transfer depends on environmental cues. Whereas LEF theoretically provides ATP and NADPH at a ratio of 2.6:2, enzymatic CO 2 fixation requires a ratio of 3:2. To overcome this metabolic restraint, CEF balances out the increased demand for ATP. Indeed, CEF allows only ATP synthesis because electrons produced by photooxidation of the primary donor of photosystem I (PSI) re-enter the electron transport chain either at the plastoquinone (PQ) pool or at the stromal side of the Cyt-b 6 f. Thus, depending on the site of re-injection of reducing equivalents into the membrane, several CEF pathways may exist. One involves PQ reductases from the chlororespiration pathway (3), a type-1 NADH dehydrogenase-like complex (NDH) in higher plants (4), and a type-2 NAD(P)H dehydrogenase in the algal model organism Chlamydomonas reinhardtii (5), hereafter referred to as Chlamydomonas. However, with less than 3 e s Ϫ1 PSI Ϫ1 the kinetics of electron flow through this pathway is too sluggish (6) (reviewed by Nawrocki et al. (7)) to match the rates of electron transfer though CEF, which are well above 10 e s Ϫ1 PSI Ϫ1 , transiently reaching up to 130 e s Ϫ1 PSI Ϫ1 in plant chloroplasts (8). Another pathway, the PGR-dependent CEF, involves the proteins proton gradient regulation 5 (PGR5) and pgr5-like photosynthetic phenotype 1 (PGRL1), which has been proposed to have a ferredoxin-PQ oxidoreductase activity (9 -11). However, a PGRL1-defective mutant of Chlamydomonas is still capable of high rates of CEF (12). Thus, a third route, still awaiting further documentation, involves direct electron donation through the ferredoxin-NADP ϩ oxidoreductase (FNR) bound to Cyt-b 6 f, making use of heme c i next to the PQ binding site Q i (8,(13)(14)(15). Whatever the route, CEF competes for electrons with other, oxygen-consuming reactions involving flavodiiron proteins (16,17) or the Mehler reaction. When oxygen is lacking in the algae, CEF is stimulated upon establishing a strongly reduced PQ pool in response to a lowered intracellular ATP/ NADPH ratio (18). The reduced PQ pool also activates the STT7 kinase, which facilitates migration of light-harvesting complex II (LHCII) and Cyt-b 6 f to the unstacked regions of thylakoid membranes (19,20). Recently, profound CEF-related rearrangements were described in Chlamydomonas, resulting in the formation of a CEF supercomplex (21). The purported supercomplex was reported to consist of PSI and Cyt-b 6 f, as well as of proteins from the light-harvesting complex I (LHCI) family and mobile LHCII. PGRL1 and FNR were also detected within this supercomplex. When illuminated, an in vitropurified CEF supercomplex accepted electrons from NADPH via ferredoxin (Fd). Furthermore, the PSI primary electron donor P700 ϩ was photoreduced by added plastocyanin. This CEF supercomplex preparation was devoid of PGR5. Based on various reports, other auxiliary proteins are believed to be part of the CEF supercomplex in Chlamydomonas as well. Among these CEF effectors are the algal proteins PETO (18,21) and anaerobic response 1 (ANR1) (22). Although PETO (23) and ANR1 (22) facilitate CEF, their exact function and stoichiometry with respect to other CEF components remain elusive. ANR1 expression is stimulated when Chlamydomonas is kept in anoxic conditions (24). In a recent study, a decreased ANR1 accumulation was observed in PETO-overexpressing lines (23). Therein, it was suggested that PETO is the main interactor of ANR1. PETO contains a lumenal N terminus that is separated from a stromal C terminus by a single transmembrane helix (25). The phosphoprotein was originally referred to as "subunit V" of the Cyt-b 6 f complex (26). PETO, largely depleted in Cytb 6 f mutant strains, was later considered as a loosely attached subunit that co-purifies with the Cyt-b 6 f in the early stages of a multistep protocol (25). The two algal CEF effectors were indirectly intertwined with the Cyt-b 6 f, like FNR in higher plants (13,14). However, they are not structurally characterized (27); nor were they shown to be genuine Cyt-b 6 f interactors. Still, ANR1, PETO, FNR, and PGRL1 frequently formed high molecular weight complexes, which has been suggested to result from the ability of these relatively small CEF effector proteins to adopt an oligomeric configuration (23).
In previous studies, the CEF supercomplex was extracted from thylakoid samples that were solubilized with alkyl maltosides, specifically ␣and ␤-glycosides of n-dodecyl maltoside (22,28) as well as n-tridecyl ␤-maltoside (18,21,23). The solubilized proteins were separated according to their apparent molecular weight via rate-zonal centrifugation using sucrose density gradients (SDGs). However, the loose attachment of CEF effectors with the CEF core (18,25,27,29) and the partial disassembly of the CEF supercomplex at elevated ionic strength (21) have hampered its further molecular and functional characterization over the last few years. Indeed, the preparation procedures after the solubilization step span a 16 -24-h time period during which original interactions within the extracted CEF supercomplex might be compromised. The preservation of the genuine components of a transmembrane supercomplex is thus a major challenge when characterizing membrane proteins (MPs) in a detergent environment (reviewed by Seddon et al. (30)). In that respect, replacing detergents with amphipol surfactants holds promises to stabilize solubilized MPs in aqueous solution (reviewed by Zoonens and Popot (31)).
In this work, we carefully revisited the experimental conditions that were previously reported to produce solubilized CEF supercomplexes in heavy SDG fractions. Despite our efforts, which used a series of different approaches to stabilize critical protein interactions and are outlined in the first two sections under "Results," we were not able to isolate a CEF supercomplex at a reproducibly high yield. Therein, we address limitations of the SDG approach by showing that the chlorophyll (Chl) content of high-sucrose density fractions is highly sensi-tive to sample handling during the solubilization step. Green bands in heavy SDG fractions, containing PSI, LHCI, LHCII, and other CEF effectors, were obtained independently of the presence of Cyt-b 6 f, which necessitates caution in the identification of a CEF green band. We also explored how the detergent concentration within SDG, which varied in published protocols, alters migration of CEF effectors, and we discuss potential caveats for the interpretation of SDG experiments. Nevertheless, our experimental conditions, although destabilizing the most labile protein interactions, enabled us in the later sections to obtain new insights into CEF-related protein interactions. Assisted by cross-link experiments and MS, we utilized the lability of the system to characterize the interface of the algal CEF effector PETO with Cyt-b 6 f at a molecular level, thus finally proving that PETO is a subunit of the algal Cyt-b 6 f. We also trapped solubilized MPs with amphipols, concluding that heavy assemblies of CEF effectors in SDG were likely aggregates resulting from detergent-induced destabilization during ratezonal centrifugation. The amphipol approach in combination with cross-link experiments argues for the algal CEF effector ANR1 to be cross-linked to Cyt-b 6 f via a yet unknown protein(s). The potential role of the loosely bound algal Cyt-b 6 f subunits is discussed in the context of CEF.

The high-density fractions in n-tridecyl ␤-maltoside (TDM)containing SDG lack Cyt-b 6 f
Here, we will tie in with the protocol by Takahashi et al. (18), which we used previously to assign heavy, Chl-containing SDG fractions as CEF supercomplexes. Our protocol was developed, like other variants (22,28,32), to build on the original findings by Iwai et al. (21), which we will return to in the next section as an attempt to improve the poor Cyt-b 6 f retention in heavy SDG fractions in our hands.
Rate-zonal centrifugation is routinely used to isolate solubilized MPs in different SDG fractions. This method has the advantage over native PAGE of allowing subsequent in vitro analysis of a CEF supercomplex. For its characterization in Chlamydomonas, various procedures were described that use alkyl maltosides as detergents (18,21,22). Four crucial steps differed between these studies: first, the heterotrophically grown algae were exposed differently to state 2-inducing conditions by inhibiting mitochondrial respiration, thus depleting ATP and increasing reducing equivalents in the chloroplast due to the Pasteur effect (33). This was done either in the presence of the membrane uncoupler FCCP (21) or by removing oxygen from the medium (18,22). Then cells were disrupted either with a nebulizer (21,22) or a French press (18), followed by thylakoid membrane isolation using a modified protocol published by Chua and Bennoun (34). Another main difference for thylakoid isolation was the buffer system, which was either MES at pH 6.5 (21) or HEPES at pH 7.5 (18,22). Finally, the use of the detergent for solubilization and SDG centrifugation differed, whereas the solubilization step was at ϳ1% (w/v) detergent per mg of Chl. Both ␣and ␤-glycosides of dodecyl maltoside (22,28) have been used, with n-dodecyl ␣-maltoside (␣DM) producing milder conditions. An even milder detergent, Interaction study of CEF effector proteins TDM, has been introduced (18,21) and was used in this study. The sucrose concentration in the discontinuous gradient ranged from 1.3 to 0.1 M, and the corresponding detergent concentration (w/v) in the SDG was 0.05% (21,22) and 0.005% (18), respectively.
The protocol from Takahashi et al. (18) uses French press cell disruption and frozen/thawed thylakoids before the solubilization step. The latter treatment proved not necessary (Fig. S1) and was thus omitted in our experiments. As compared with the use of a nebulizer, breaking the cells more effectively via French press (see "Experimental procedures") could disrupt weak protein interactions. We show in Fig. 1 that the two cell disruption modes produced the same set of major green bands. The Chl-binding LHCs were probed in Fig. S2. A set of overlapping bands culminated in a diffuse high-density green band in fractions 4 -7, which was greener when using the nebulizer. Independently of the disruption method, the heavy PSI-containing fractions were greener if the reaction tubes were inverted every 5 min during the solubilization step ( Fig. 1) instead of flicked (Fig. S3). Nonetheless, the distribution of the other core component of the CEF supercomplex, the Cyt-b 6 f, was confined between the PSI and the PSII band in the upper part of the gradient (fractions 16 -22, Fig. 1). We saw, however, different Cyt-b 6 f distributions in the SDG that resulted even in faint signals at the very bottom fraction. Clearly, the Cyt-b 6 f/ PSI signals were not well-aligned in the heavy SDG fractions, and thus, we were unable to produce a CEF supercomplex in these experimental conditions. In agreement with other reports (18,23), the auxiliary CEF effectors were detected in the very bottom fractions of the SDG as well. In particular, the green high-sucrose density fractions 3-7 contained PSI, LHC from both photosystems, and the auxiliary CEF proteins PETO, ANR1, and PGRL1 ( Fig. 2A). Again, neither cytochrome f (Cyt-f) nor subunit IV (su-IV) were detected in the lower SDG region. Despite the absence of the Cyt-b 6 f, a ⌬Cyt-f mutant produced a very similar protein distribution (Fig. 2B). However, PETO was hardly detectable in solubilized ⌬Cyt-f thylakoid membranes, consistent with the report that PETO accumulation depends on that of Cyt-b 6 f (23, 25, 26). Interestingly, the same holds true for ANR1. It should be noted that, due to the was observed upon SDG ultracentrifugation. As described under "Experimental procedures," the harvested cc-124 WT cells were subjected to state 2-inducing anoxic conditions, and cell disruption was carried out with a French press (FP) (A) or by using the gentler nebulizer method (Neb) (B). Thylakoids were isolated in a HEPES buffer as described previously (18). Independently of the disruption method, su-IV of the Cyt-b 6 f was found mainly in fractions 16 -22 in the SDG. The smearing PSI, probed via PsaA, peaked in fractions 15-19 with minor peaks in fractions 12 and 13 and at high-sucrose density fractions 5-8. R.U., relative units of normalized Chl absorbance at 680 nm.

Interaction study of CEF effector proteins
absence of Cyt-b 6 f, the ⌬Cyt-f mutant was locked in state 1 despite the pretreatment (26,35). Nevertheless, state transitions are not a prerequisite for CEF and for the production of Chl-containing heavy SDG fractions (18).
In the next section, we tested whether the original protocol by Iwai et al. (21) was more suitable to retain Cyt-b 6 f in highsucrose density regions together with PSI.

The original isolation protocol, in combination with dithiobis(succinimidyl propionate) (DSP) cross-link treatment, still produced poor Cyt-b 6 f retention in the high-sucrose density fractions
To use experimental conditions closer to the initial report on a CEF supercomplex isolation in Chlamydomonas (21), we inhibited mitochondrial respiration with FCCP before cell disruption via French press, and a MES buffer was used at pH 6.5 for thylakoid membrane isolation. In accordance with the studies by Iwai et al. (21), the protocol yielded four distinct green bands. However, in our hands, the Cyt-b 6 f content in the lower green band remained very limited (fractions 7-10 in Fig. 3A).
At variance with the experimental conditions used in Figs. 1 and 2, Fig. 3 results from a 10-fold higher TDM concentration in SDG, hereafter referred to as high-TDM SDG, which contained 0.05% (w/v) TDM. In contrast, the SDG from Figs. 1 and 2 contained 0.005% (w/v) TDM, referred to as low-TDM SDG. When following the procedure used by Takahashi et al. (18) but using high-TDM SDG (Fig. 3B), the ultracentrifugation of solubilized state 2 thylakoids produced three green bands and a pale band (fractions 7-9) below the PSI band. The heaviest SDG fractions were devoid of PSI and colorless, unlike in the original Takahashi et al. (18) conditions ( Fig. 2A). The FNR distribution was also markedly changed. In low-TDM SDG ( Fig. 2A), FNR migrated in low-density fractions, whereas in high-TDM SDG (Fig. 3B), FNR mainly peaked with PSI in fractions 9 -14 (both panels are assembled in Fig. S4 for comparison). However, FNR was also present in upper fractions, such as 16 -21, where Cytb 6 f was detected with a slight offset to the PSII core marker CP47. Given their respective molecular mass, SDG comigration of the monomeric PSII core complex (36) with the ϳ220-kDa Cyt-b 6 f homodimer is expected (27,37).
Here again, the experimental conditions used above were not suitable to obtain significant comigration of PSI and Cyt-b 6 f as core elements of a putative CEF supercomplex. We then combined the conditions of Iwai et al. (21) with a low-TDM SDG at 0.005% but did not get a major enrichment in PSI/Cyt-b 6 f comigration (Fig. S5).
Because we found no conditions to recover a significant proportion of Cyt-b 6 f in high-sucrose density fractions, we attempted to overcome a putative dissociation of the purported CEF supercomplex by using the reversible cross-linker DSP. This homobifunctional cross-linker is thiol-cleavable and targets Lys within a 1.2-nm distance. Before their solubilization, isolated thylakoids were incubated with DSP. This pretreatment did not modify the distribution of green bands in the Iwai et al.   Fig. 1A. A, the top three green bands of solubilized cc-124 WT thylakoids migrated similarly as in Fig. 1A. Due to disperse pigmentation, the fourth green band in fraction 10 was hardly distinguishable, and at the bottom of the gradient (fractions 4 -7) a fifth green band appeared, which contained CEF supercomplex components (21) except for Cyt-b 6 f and FNR. The latter was found in low-sucrose density fractions 20 -25. PETO, ANR1, and PGRL1 migrated up to the bottom of the SDG. PGRL1 showed a second peak cofractionating with PSII and Cyt-b 6 f. B, the green high-sucrose density band was produced in a mutant devoid of Cyt-b 6 f, harboring a mutation in petA (66). The strain showed protein distribution similar to that of the WT, whereas PGRL1 showed broader signals in the upper half of the SDG. ANR1 and PETO were strongly underrepresented in solubilized thylakoids. R.U., relative units of normalized Chl absorbance at 680 nm.

Interaction study of CEF effector proteins
plex protein distribution across the SDG, we observed a larger portion of Cyt-b 6 f now retained in high-density fractions 8 -10, where PSI, PETO, ANR1, and FNR were also found (Fig. 3C).
When exploring the DSP effect in solubilized HEPES-buffered samples (Fig. 3D), the green intensity of fractions 7-9 showed a marked increase, due to the presence of minor but significant amounts of PSI in these higher-density fractions. However, these fractions only contained minute amounts of Cyt-b 6 f, far from fulfilling the expectations for a CEF supercomplex.
The FNR detection in heavy fractions was not only dependent on the detergent concentration in the SDG (Fig. S4).
As compared with untreated samples, DSP pretreatment increased FNR bimodal distribution due to the loss of signal in the Cyt-b 6 f-containing fractions (Fig. 3, compare B and D). The DSP effect on bimodal FNR distribution was also seen in low-TDM SDG conditions (Figs. 2A and 4A; see assembly in Fig. S6). A large FNR proportion was found in higher-density SDG fractions, where a subpopulation of PSI also was found. This suggests that the destabilization of PSI-FNR interactions in low-TDM SDG was in part counteracted by the cross-linking treatment before solubilization. A most striking effect of the cross-link treatment was on ANR1 and PETO, which comigrated in SDG fractions with either PSI or Cyt-b 6 f, thus pre- . At variance, the SDG was supplemented with 0.05% TDM (w/v). Where indicated, 2.5 mM DSP was used as Lys cross-linker as under "Experimental procedures." A, the densest green band in fractions 7-10 contained PETO, ANR1, PSI (PsaF), FNR, and traces of Cyt-b 6 f (Cyt-f). PETO, ANR1, and FNR migrated below the major PSI band in fractions 11-15. The Cyt-b 6 f majority was in fractions 15-19. B, the an.HEP protocol without DSP pretreatment yielded three green bands and a faint band in fractions 7-9. PETO, ANR1, and Cyt-b 6 f distribution resembled that in A. The abundance of PSI (PsaA) was confined to the major green band in fractions 10 -15, and FNR migration peaked with PSI while smearing to lower sucrose densities (up to 24). C, the DSP pretreatment in unc.MES samples did not change the green bands but, compared with A, increased Cyt-b 6 f abundance in the densest band (fractions 7-10). Strong FNR signals were detected between fraction 5 and the PSI peak in fraction 13. Low-density fractions contained minor amounts of FNR (smearing up to 25), PSI (up to 17), and ANR1, which comigrated with the smearing PSI and Cyt-b 6 f majority (up to 18). D, compared with B, DSP treatment produced a greener band in fractions 7-9, which contained minor populations of PSI and FNR. The FNR peak around fraction 11 was slightly displaced from the PSI peak (fractions [11][12][13]. A DSP-dependent FNR peak at fractions 23-25 was produced. The Cyt-b 6 f majority remained in fractions 17-20, but traces were detected in higher sucrose densities (down to 9). Both PETO and ANR1 were retained in low-density fractions comigrating with smearing PSI and peaking with Cyt-b 6 f majority. R.U., relative units of normalized Chl absorbance at 680 nm.

Interaction study of CEF effector proteins
venting their aggregation in high-sucrose density fractions after their solubilization (see Figs. 2-4). In the following three sections, we will focus on these two proteins, which were hardly detectable in solubilized membranes of a Cyt-b 6 f mutant.

The effect of DSP on the SDG migration behavior of ANR1 and PETO is independent of the other auxiliary CEF protein
ANR1 has been described as interacting with PETO (23), and indeed ANR1 and PETO had a strikingly similar migration behavior in the TDM-containing SDG, with and without DSP pretreatment. We also observed that the accumulation of ANR1 and PETO in whole-cell extracts was compromised in stationary cultures of two independent Cyt-b 6 f-lacking mutant lines (Fig. S7). These experiments confirmed earlier reports on a Cyt-b 6 f-dependent accumulation of PETO (23,25,26) and demonstrate a similar behavior for ANR1.
The DSP-treated WT sample in Fig. 4 showed SDG peak fractions of both ANR1 and PETO distribution where the major population of Cyt-b 6 f fractionated. Their low amount in the absence of Cyt-b 6 f (Figs. 2B and 4B) precluded further investigations on their migration upon SDG centrifugation in the total absence of this protein complex. We thus looked at whether the purported interactions of ANR1 and PETO with Cyt-b 6 f were mutually dependent. A ⌬ANR1 strain was available in the CLiP library, and PETO-knockdown strains were described previously (23). Fig. 5 shows the SDG centrifugation result of solubilized thylakoid membranes from these two mutants and demonstrates the effect of DSP treatment. It is apparent that the DSP-induced comigration of a subpopulation of either ANR1 or PETO with the Cyt-b 6 f did not require the participation of the other protein, meaning that their putative interactions with Cyt-b 6 f were independent from each other. In the following sections, we investigated these interactions by looking at possible cross-link events between ANR1 and PETO with the Cyt-b 6 f.

Cross-link-assisted description of the PETO interface with Cyt-b 6 f
To understand whether mutually independent changes in ANR1/PETO migrations upon DSP pretreatment involved actual cross-links with Cyt-b 6 f, we used the H 6 F 5 strain, which harbored a modified Cyt-f fused with a His 6 tag at its C terminus (38) and followed the same isolation, cross-link, and solubilization protocol as for the WT and mutants in Fig. 5. Instead of analyzing samples via SDG ultracentrifugation, we used a twostep anion-exchange/metal-affinity purification protocol and examined the Cyt-b 6 f preparation on a nonreducing/reducing 2D PAGE, as shown in Fig. 6. This purification procedure has been used previously for crystallization of the algal Cyt-b 6 f core, which contained neither PETO nor ANR1 (27). A mix of antibodies to Cyt-f, su-IV, and PETO was used for immunoblotting the 2D gels in which cross-links were cleaved in the second dimension; thus, off-diagonal signals resulted from cross-links (or interprotein disulfides) in the first dimension. It is of note that proteins with covalent heme binding, cytochrome b 6 and Cyt-f, were readily detected by the enhanced chemiluminescence methods used for the immunoblots (39). As shown in Fig.  6A, PETO could not be detected with Cyt-b 6 f when omitting The samples were isolated and analyzed as in Fig. 2 but pretreated with DSP as in Fig. 3D. Therefore, see Fig. 2 for the DSP effect and Fig. 3D for the high-TDM SDG effect in cc-124. A, the green band and protein migration profile of DSP-treated cc-124 thylakoids were changed in low-TDM SDG. The heavy green band (fractions 4 -7 in Fig. 2A) disappeared, and a relative pigmentation increase of the PSI major band (fractions 12-16) was observed. The DSP effect on ANR1, PETO, and Cyt-b 6 f migration resembled high-TDM SDG (Fig. 3D). In addition to the low-density FNR peak in fractions 24 -26 (see also Fig. 2A), the DSP effect on FNR produced high-sucrose density peak fractions 9 -13, cofractionating with PSI but slightly offset to the PSI major band. DSP pretreatment broadened PGRL1 distribution. B, the DSP effect on the green band patterns in ⌬Cyt-f was identical to the WT, as were the changes in FNR distribution. R.U., relative units of normalized Chl absorbance at 680 nm.

Interaction study of CEF effector proteins
the DSP pretreatment of the thylakoids. On the contrary, PETO was easily detected in the DSP-treated samples, showing a cross-link with Cyt-f and su-IV ( Fig. 6B; see legend for more details). These observations expand earlier studies that proposed PETO as a loosely attached algal Cyt-b 6 f subunit (25,26).
Next, we submitted the affinity-purified Cyt-b 6 f DSP sample to MALDI-TOF/TOF. Before the measurements, the disulfide bridges were cleaved to identify which surface Lys residues were reproducibly labeled by DSP, yielding a variable Lys modification (ϩ145.01975) from the carboxyamidomethylated, cleaved DSP in three biological replicates (Table S1). Because ionization efficiency of cross-linked peptides is often low, thus interfering with cross-link identification, the DSP-labeling approach gave us a first hint which Lys residues were easily reacting with the chemical. Ultimately, we attempted to identify the cross-linked residues with the ECL2 algorithm (40). Therefore, as outlined under "Experimental procedures," we used the noncleavable homobifunctional disuccinimidyl suberate (DSS) and thylakoid samples, which were solubilized with ␣DM. Similar to experiments with DSP, the Lys cross-linker DSS was added before solubilization. The ECL algorithm (40) deconvoluted the measured LC-MS/MS spectra of Cyt-b 6 fcontaining SDG fractions and identified cross-linked peptides from PETO and su-IV. We were not able to detect DSS-crosslinked peptides from PETO and Cyt-f. The LC-MS/MS spectra for peptides KPDLSDPVLkAK (PETO) and SENGVTFP-kVAAGGTK (su-IV) with cross-linked Lys in lowercase type are shown in Fig. S8. By using RaptorX structure prediction Interaction study of CEF effector proteins server (41), a PETO model was placed in proximity to the published structure of Chlamydomonas Cyt-b 6 f (27). PETO was reported to have a lumenal N terminus that is separated from the stromal C terminus by a single transmembrane helix (25). Interestingly, the C terminus of PETO was cross-linked to the stromal N terminus of su-IV through their respective lysines, Lys-121 and Lys-15 (Fig. 7). This cross-link develops in close vicinity of the Ser/Thr kinase STT7-dependent phosphorylation sites Thr-110 and Ser-113 of PETO and Thr-4 of su-IV (28). It is notable that a nearby peripheral stromal loop, formed by su-IV residues 113-128, interacts with STT7 (42).

ANR1 is not cross-linked to the Cyt-b 6 f by DSP, but they cofractionate in the amphipol-trapped state
By following the approach described for PETO in Fig. 6, we failed to detect any cross-link supporting ANR1 as an interaction partner of the affinity-tagged Cyt-b 6 f in the H 6 F 5 strain. This is consistent with the immunoblot of a DSP-treated sample probed with an antibody to ANR1 (Fig. 8A). The absence of an ANR1 cross-link with Cyt-b 6 f appears to conflict with their DSP-induced comigration in SDG (e.g. see Fig. 5F). We then assumed that not ANR1 itself but an unknown interactor(s) would be cross-linked to Cyt-b 6 f. The unknown Cyt-b 6 f crosslink partner(s) only interacts with ANR1 at low ionic strength and gradually detaches from TDM-solubilized Cyt-b 6 f during the lengthy SDG ultracentrifugation. Using an anion-exchange step (27) that resembled CEF supercomplex-destabilizing conditions (21) during the Cyt-b 6 f purification procedure could therefore wash away ANR1 even upon DSP pretreatment. To test these assumptions, we searched for better conditions to stabilize protein/protein interactions between detergent-solubilized proteins. We used a PETO-Strep tag strain as well as trapping of TDM-solubilized proteins by amphipols.
The PETO-Strep strain allowed purification of Cyt-b 6 f complexes at slightly less harsh conditions regarding the ionic strength (170 mM instead of 250 mM NaCl). This strain had been designed in a previous study to show that, among CEFinvolved proteins, ANR1 was the main interactor of PETO (23). Whereas we could reproduce a limited ANR1/PETO coelution in the control experiment, DSP pretreatment increased the abundance of Cyt-b 6 f, but not of ANR1, in the batch-eluted PETO fractions (Fig. 8B). At variance with the previous report (23), FNR remained below detection in the eluted fractions in our experimental conditions. As a control, none of the above proteins were detected in the elution fractions using strains that overexpressed WT PETO (Fig. S9).
We tested whether minimizing exposure time to TDM could stabilize ANR1-specific Cyt-b 6 f interactions. It has been observed previously that Cyt-b 6 f homodimers disassemble over time in certain maltoside detergents (43,44). Therefore, we used the amphipol PMal-C8 in a second approach (45,46). These short amphipathic surface-active agents can substitute for detergents, thus creating stabilizing interactions of various solubilized MPs (reviewed by Zoonens and Popot (31)). However, the amphipol capacity to solubilize MPs is poor, and   (27). The corresponding MS2 spectrum is shown in Fig. S8. A model structure of PETO (green) was generated using the RaptorX service (41) with manual refinement of the transmembrane helix (TMH; residues 71-104). The Cyt-f polypeptide is shown in yellow, and the remaining subunits are shown in gray. The transmembrane helices in the su-IV, Cyt-f, and PETO models are depicted as solid cylinders, and simplified loops were overlaid by a semitransparent surface representation. Various lumenal PETO and Cyt-f residues were DSPlabeled (cyan; see Table S1), but cross-links could not be identified, and Lys residues that were not labeled are shown in blue. Pink STT7-related elements in proximity to the lumenal cross-link highlight phosphorylation sites in PETO and su-IV (28) as well as a putative STT7 interaction interface located in the stromal su-IV loop, fg su-IV (42).

Interaction study of CEF effector proteins
therefore a detergent-mediated solubilization is necessary. By choosing the buffer composition (46,47) and adding an excess of PMal-C8 during detergent removal (see "Experimental procedures"), we favor the coverage of hydrophobic protein domains by the zwitterionic amphipol while disfavoring electrostatic and/or hydrophobic protein aggregation processes in the detergent-free solution. Even if the latter artifact existed, it was eliminated by the centrifugation step before supernatant loading. The samples shown in Fig. 8C were solubilized by TDM, followed by substitution with PMal-C8 and TDM removal. After spinning down the mixture, the supernatant was loaded on an SDG that was not supplemented with TDM. The samples now contained the solubilized photosynthetic machinery with PSI and PSII showing a bimodal distribution upon SDG centrifugation. PSI showed two defined populations (fractions 8 -13 and 18 -20), as did PSII, peaking in fractions 8 -11 and 17-19. Fractions 8 -11 also showed minute amounts of Cyt-b 6 f (arrowhead in Fig. 8C), which accumulated mostly in fractions 19 -21. Interestingly, ANR1 now peaked together with Cyt-b 6 f in fraction 20, which suggested stabilization of the ANR1/X/ Cyt-b 6 f interface in the detergent-free state. Here, X refers to the unknown Cyt-b 6 f-cross-linkable protein(s), as assumed above in the TDM detergent system. Fraction 20 contained only trace amounts of PETO. The latter was mostly detected in the upper fractions 25-26, at moderate sucrose densities that correlated with its apparent molecular weight. Significant amounts of ANR1 were also detected in the high-sucrose density fractions 8 -11, as were PSI (PsaA), PSII (CP47), PGRL1, and the ATP synthase (AtpB). The slight relative offset migration of CEF effectors FNR and PGRL1 in the upper region was also observed in low-TDM SDG ( Fig. 2A and Fig. S1).

Discussion
Here we provided a molecular characterization of the Cyt-b 6 f interaction with the loosely attached CEF effector PETO. PETO was cross-linked to either su-IV or Cyt-f, rather than producing a cross-linked tripeptide. This raises the question of the actual PETO stoichiometry per Cyt-b 6 f. The following attempts to compensate the missing PETO-Cyt-f cross-link identification and explain Fig. 6B are summarized in Fig. 9A. Disregarding higher stoichiometries, PETO isoform-specific cross-links could occur in the case of two PETO per functional Cyt-b 6 f homodimer (PETO/Cyt-b 6 f ratio of 1). When assuming a ratio of 0.5, two scenarios are possible. (i) Supported by reproducible DSP reactivity of various lumenal Lys residues in Cyt-f and PETO, it is likely that there was a yet-to-be-pinpointed lumenal PETO-Cyt-f cross-link. Upon PETO cross-linking on one side of the membrane (to stromal su-IV or lumenal Cyt-f), structural alterations in PETO are transmitted to the opposite side of the membranes, which hamper subsequent cross-links on that side.

Interaction study of CEF effector proteins
In this view, Fig. 6B would be a stochastic snapshot of the PETO/ Cyt-b 6 f population. Signal transduction across the membrane indeed has been suggested for other proteins, such as the STT7 kinase (48). STT7 is important for state transition, which is a CEF-independent process to optimize photosynthesis under suboptimal conditions (18) (reviewed by Rochaix et al. (49)). STT7 was shown to interact with the Cyt-b 6 f (42) and relies on both the redox state of the PQ pool (50) and a functional, lumenal Q o site of the Cyt-b 6 f (35) opposite to the stromal site where the catalytic activity of the kinase takes place. (ii) An alternative to cross-link-mediated alterations in PETO across the membrane would be that of an STT7-induced phosphorylation of stromal Ser/Thr in both PETO and su-IV, which have been identified in proximity to our reported cross-link site (28). These posttranslational modifications could reorganize salt bridges and other electrostatic networks (51), which would then favor stromal PETO-Cyt-f interactions as well. Here, Fig.  6B would be a metabolic snapshot of the PETO/Cyt-b 6 f population. Even without further elaborating on the yet-to-be-identified Cyt-b 6 f/ANR1 interface and autophosphorylation of the STT7 kinase domain that results from the interaction with a proximal su-IV loop of Cyt-b 6 f (42), the suggested stromal scenario involves highly dynamic and partially competing protein interactions. These processes could participate in CEF and/or state transition regulation.
Because ANR1 and PETO are algae-specific CEF effectors (22,23), they might play a role in CEF features that are not found in higher plants. Chlamydomonas has a versatile anaerobic metabolism (52,53), and the lack of oxygen induces several anaerobic responses like increased ANR1 expression (24) and increased CEF due to an altered PQ pool redox equilibrium (18,22,23). Algal and plant PQ pool redox equilibria differ due to changes in chlororespiration, which is electrogenic in plants only (reviewed by Nawrocki et al. (7)). It has been documented that plant type-1 NDH can use Fd as electron donor (54), which may not be the case for the algal type-2 NAD(P)H dehydrogenase (5). Recently, ANR1 was shown to share a motif with NdhS, which is the Fd-binding subunit in plant NDH (23). Interestingly, both ANR1 and PETO interact with Fd FDX1 (55). ANR1 also develops specific FDX2 interaction, whereas PETO interacts with FDX5 (56). Thus, the two algal CEF effectors could be Fd-shuttling proteins between PSI and Cyt-b 6 f, the latter acting as an Fd-PQ reductase (Fig. 9B). In this view, Cyt-b 6 f receives Fd from PETO/ANR1 via FNR, which was shown to bind plant Cyt-b 6 f (13,14). It should be noted that FNR in Chlamydomonas is encoded by a single gene only, unlike in plants that express multiple isoforms with specific membrane association preferences (14,57). Algal FNR is bound to the membrane in a PGR5-dependent manner (58). Its membrane tethering differs in general because higher plant FNRbinding proteins are missing in Chlamydomonas (59 -61). Presently, we do not know whether PGRL1/PGR5 contribute to FNR tethering in algae directly or play a role in a quality control of the FNR pool that participates in CEF and LEF.
Although the cross-link-assisted disintegration analysis enabled us to identify further CEF-related protein interactions, we conclude that the commonly applied SDG approach, due to its liability to artifacts, is not appropriate for further biochem-ical characterization of the CEF supercomplex. Chl-containing heavy SDG fractions can easily be misidentified with the putative CEF supercomplex. Its isolation would require, for instance, an additional affinity chromatography purification step to demonstrate genuine interactions between comigrating proteins. Unfortunately, the buffers that ensure specific binding to the immobilized phase also promote disruption of the CEF supercomplex (21).
SDG served as a tool to separate solubilized MP complexes of different apparent molecular weight. However, there is no straightforward correlation between their migration behavior and their molecular weight because shape, density, and hydrophobic interactions also contribute to their behavior upon SDG centrifugation. Here we observed that, when (in)directly crosslinked to Cyt-b 6 f assemblies, PETO and ANR1 were retained in The cross-link sites were identified (red asterisk) or suggested (white asterisk). STT7-dependent phosphorylation is marked in magenta. B, a cartoon model of CEF (red arrows) is shown. LEF is depicted with blue arrows. The FNR is tethered to the membrane dependent on PGRL1 (L1) and PGR5 (58). Bound to Cyt-b 6 f (yellow; marking f/b l /b h /c i hemes and Rieske Fe 2 S 2 protein), the FNR injects electrons into the PQ pool via heme c i (13)(14)(15). Our study proposes Fd-shuttling functions of ANR1 and PETO (55,56) during this process by interacting with the Cyt-b 6 f. As outlined under "Discussion," the CEF supercomplex model proposed by Iwai et al. (21) requires improved protocols to be further validated and/or characterized.

Interaction study of CEF effector proteins
lower-sucrose density fractions. Compared with the control, relative SDG migration of the ϳ220-kDa Cyt-b 6 f homodimer was not altered. This was unexpected because of the apparent gain in molecular mass of the Cyt-b 6 f homodimer assembly by at least 35 kDa, depending on the actual (yet unknown) CEF effector ratio per Cyt-b 6 f. Other unexpected SDG results were related to solubilization processes during the rate-zonal centrifugation step. Accordingly, the behavior of FNR was odd when changing the TDM concentration in the gradients. Considering the various parameters that participate in the solubilization process (30), the membrane-peripheral FNR co-solubilized with PSI and/or Cyt-b 6 f in high-TDM SDG (0.05% TDM), probably because FNR binding to its CEF sites or to tethering proteins was inaccessible at low-TDM SDG (0.005% TDM). Thus, optimal detergent conditions for tethering FNR to a CEF supercomplex interfered with stable core component assemblies (43,44) during lengthy rate-zonal centrifugation steps. Keeping integral MP complexes in their original oligomeric state, outside their natural environment, remains a major biochemical challenge (reviewed by Seddon et al. (30)), particularly when facing loose interactions as in the CEF supercomplex. The presence of CEF effectors in heavy fractions at the bottom of TDM-SDG, previously considered to represent either TDM-inaccessible membrane microdomains or high-molecular weight complexes (23), could in fact be interpreted as MP aggregates. When stabilizing TDM-solubilized MP by trapping with amphipol PMal-C8, we observed a striking co-fractionation of ANR1 with the Cyt-b 6 f homodimer. The amphipolstabilized CEF effectors PETO, PGRL1, and FNR, which we routinely detected in high-sucrose density fractions or smearing across the TDM-SDG, also migrated in well-defined regions in a detergent-free state. In conclusion, the heavy, detergentcontaining SDG fractions contained broadly migrating CEF effector aggregates and the purported CEF supercomplex core. Regarding fractionation of the latter, PSI and Cyt-b 6 f signal intensities were poorly correlating. Therefore, further purification steps of the pooled fractions of interest are crucial. These will, however, increase the chances of additional destabilization. Using subsequent affinity chromatography allowed us to purify putative CEF supercomplexes from SDG at very low yield (73), suggesting the development of novel purification protocols of the labile CEF complex. In that respect, amphipols might be an additional promising approach for future CEF supercomplex research. Furthermore, alternative methods like interval zone free-flow electrophoresis will also facilitate further biochemical progress. Using this technique in Arabidopsis thaliana, Cyt-b 6 f/PSI particles were observed recently (62).
Taken together, we conclude that the published CEF supercomplex isolation procedures need further improvement to assess their actual existence and to proceed with functional and structural characterization because the MP assemblies are easily destabilized using previous protocols. Our results demonstrate mutually independent interactions between the two Fdinteracting CEF effectors ANR1 and PETO with the Cyt-b 6 f in C. reinhardtii. Based on our current observations, we favor the concept proposed by Joliot et al. (15), where Cyt-b 6 f-associated FNR (13,14) facilitates the reinjection of electrons at the Q i site via heme c i . During this process, the CEF effectors PGR5/ PGRL1 and ANR1/PETO likely play FNR-coordinating (58) and Fd-coordinating roles (55,56) that are functionally redundant because none of the single knockout/knockdown strains is truly CEF-impaired (12,22,23,63).

Strains and growth conditions
In the experiments, C. reinhardtii WT cc-124 (nit2 Ϫ , mt Ϫ ) and cc-4533 (cw15 mt Ϫ ) were used. The latter was provided as reference for ⌬ANR1, which was an insertional mutant (locus Cre03.g164000) designated as LMJ.RY0402.098359 in the CLiP library (64). Strains that overexpressed PETO (WT or coding for Strep-tagII) and the PETO-knockdown strain 3B8 were described previously (23) and generated in the arg7 background (65). The H 6 F 5 strain coded for a His 6 -tagged Cyt-f (38). Cyt-b 6 f-lacking strains ⌬f on the level of petA/Cyt-f (66) and Fud6 on the level of petD/su-IV (67) have been described previously. All cells were grown in Tris acetate phosphate (TAP) medium at 25°C under 50 mol of photons m Ϫ2 s Ϫ1 and kept in the exponential phase by repeated dilution in fresh TAP.

State 2 induction and thylakoid isolation
Cells in the early to mid-log phase were harvested (5 min at 5000 rpm in Beckman JA-12 rotor, 22°C) and resuspended in spent TAP medium to a density of ϳ5 ϫ 10 7 cells/ml. State 2 induction relied on oxygen removal either by incubating the harvested cells (15-25 ml) in the dark with 5 M FCCP (21,68) or bubbling with argon gas (22,28) for ϳ45 min. The latter is referred to here as "anoxic conditions." State 2 transition was routinely checked by recording low-temperature fluorescence emission spectra at 77 K (not shown), followed by cell disruption either with a French press (18) at ϳ20 kilopascals or nebulization twice with a BioNeb cell disruptor at ϳ10 kilopascals (68). The thylakoid isolation was basically as described previously (34,69), and CEF supercomplex studies were further modified from the original protocols (18,21,22). After French press or nebulizer passage, samples were handled throughout at 4°C, and disrupted cells were diluted 1:1 in a prechilled 2ϫ buffer, yielding a final mix of 25 mM HEPES (pH 7.5), 0.33 M sucrose, 10 mM EDTA, 10 mM NaF, 1.5 mM KCl, 5 mM 6-aminocaproic acid, 1 mM benzamidine, 1 mM phenylmethylsulfonyl fluoride (H25 buffer). Where indicated, HEPES was replaced by 25 mM MES, pH 6.5 (M25 buffer). Cells were pelleted (20 min at 40,000 rpm in a Beckman Type 45 Ti rotor) and homogenized in H5/M5 buffer using a potter. The H5/M5 buffer was identical to H25/M25 but only weakly buffered at 5 mM in the presence of 1.8 M sucrose. Thylakoid floatation via a three-step sucrose gradient was as described previously (69), and thylakoids were Interaction study of CEF effector proteins resuspended in sucrose-free H5/M5 buffer devoid of 6-aminocaproic acid and benzamidine. For further use, the Chl concentration (70) was set to 2 and 1 mg/ml for experiments with and without the cross-link step, respectively.

Cross-link reactions
DSP and DSS are homobifunctional and membrane-permeable cross-linkers. Both compounds contain amine-reactive N-hydroxysuccinimide esters at each end of an 8-carbon spacer arm (distance: DSP, ϳ12.0 Å; DSS, ϳ11.4 Å), thus cross-linking side chains of Lys. Chemical protein cross-linking was performed at 2.5 mM DSP or DSS. Thylakoids were diluted to a final concentration of 1 mg Chl/ml with freshly prepared crosslinker, added from a prewarmed 2ϫ stock solution (2% (v/v) DMSO in sucrose-free H5/M5 buffer for mock controls). Cross-linking was carried out in the dark for 2 h on rotating disks. After 15-min quenching of nonreacted cross-linker with 50 mM Tris-HCl, pH 6.8, samples were solubilized.

Thylakoid membrane solubilization and SDG ultracentrifugation
Thylakoids, equivalent to 1 mg of Chl/ml (or 0.95 mg of Chl/ml in cross-link experiments) were brought to 0.8 mg of Chl/ml by solubilizing in 0.85% (w/v) TDM for 20 min. TDM was added to the thylakoids from a 4.25% stock (5.38% in crosslink experiments), and during solubilization, the reaction tubes were inverted once every 4 -5 min, unless otherwise stated. Before loading 375 l of supernatant on the SDG, samples were spun down for 5 min at 14,000 rpm. The SDG (a discontinuous step gradient with 1.7/1.5/2.0/2.0/2.0/1.0 ml of 1.3/1.0/0.85/ 0.7/0.65/0.4 M sucrose) contained 5 mM Tricine-KOH (pH 8.0) and either 0.005 or 0.05% (w/v) TDM. Upon reaction, crosslinks were present during SDG ultracentrifugation and cleaved in selected SDG fractions by incubation with 50 mM DTT for 1 h at 37°C, followed by SDS-PAGE (71) and Western blotting, using a ChemiDoc imager (Bio-Rad).

Affinity purification protocols
The His-tagged Cyt-b 6 f (27) and Streptactin-tagged PETO (23) were purified according to previously published protocols using HiTrap Q HP and HisTrap HP columns for the former and Strep-trap HP columns for the latter (GE Healthcare). The His-tagged Cyt-b 6 f (15 g of protein) was analyzed in nonre-ducing conditions in the first dimension, optionally stained by SYPRO Ruby, and gel slabs were incubated with 50 mM DTT for 1 h at room temperature, followed by SDS-PAGE in the second dimension and Western blotting.

Mass spectrometry analysis
For in-gel digestion of the DSP labeling experiments, samples were reduced in 10 mM DTT and 50 mM NH 4 HCO 3 for 30 min at 56°C in darkness, followed by alykation in 50 mM iodoacetamide and 50 mM NH 4 HCO 3 for 30 min at 56°C in darkness. Trypsin digestion was performed overnight at 37°C. The extracted peptides were then dried in a SpeedVac and resuspended in 40 l of UPLC loading buffer (3% acetonitrile, 0.1% formic acid). For LC-MS/MS analysis, samples were analyzed by Nano LC-MS/MS using an EASY-nLC 1000 liquid chromatography system coupled to a Q Exactive plus mass spectrometer (Thermo Fisher Scientific). See Table S2 for details on LC and MS settings. MASCOT (Matrix Science, London, UK) interrogation was performed via the protein identification software Proteome discoverer (Thermo Scientific) with C. reinhardtii Uniprot DB. The search parameters allowed for up to three missed cleavages. Modifications with a peptide mass tolerance of 10 ppm were set as fixed for carbamidomethyl of Cys. Met oxidation, 3-(carbamidomethylthio)propanoyl of Lys, and protein N terminus were selected as variable modifications. Mass tolerances were 10 ppm on the precursor level and 0.02 Da on the MS2 level. False discovery rates for protein identification were fixed at 0.01. For cross-link site identification, isolated thylakoids were cross-linked with DSS and solubilized with ␣DM. After an SDG ultracentrifugation, the Cyt-b 6 f fraction was tryptically digested using the FASP method (72). Peptides were then analyzed on an LC-MS/MS system consisting of an Ultimate 3000 nano-LC coupled to a Q Exactive Plus mass spectrometer (Thermo Fisher Scientific). See Table S2 for details on LC and MS parameters. The cross-linked peptide combination was found in multiple spectra that were evaluated with an ECL algorithm (40) using a C. reinhardtii Uniprot DB limited to chloroplast-localized proteins. Carbamidomethylated cysteine and oxidized methionine were set as fixed and variable modifications, respectively. Mass tolerances were as described above. Results were filtered to satisfy a false discovery rate of 0.01. The spectrum was plotted with ECL viewer version 2.0.