PGR5 is required for efficient Q cycle in the cytochrome b6f complex during cyclic electron flow

Proton Gradient Regulation 5 (PGR5) is involved in the control of photosynthetic electron transfer but its mechanistic role is not yet clear. Several models have been proposed to explain phenotypes such as a diminished steady state proton motive force (pmf) and increased photodamage of photosystem I (PSI). Playing a regulatory role in cyclic electron flow (CEF) around PSI, PGR5 contributes indirectly to PSI protection by enhancing photosynthetic control, which is a pH-dependent downregulation of electron transfer at the cytochrome b6f complex (b6f). Here, we re-evaluated the role of PGR5 in the green alga Chlamydomonas reinhardtii and conclude that pgr5 possesses a dysfunctional b6f. Our data indicate that the b6f low-potential chain redox activity likely operated in two distinct modes – via the canonical Q cycle during linear electron flow and via an alternative Q cycle during CEF, attributing a ferredoxin-plastoquinone reductase activity to the b6f. The latter mode allowed efficient oxidation of the low-potential chain in the WT b6f. A switch between the two Q cycle modes was dependent of PGR5 and relied on unknown stromal electron carrier(s), which were a general requirement for b6f activity. In CEF-favouring conditions the electron transfer bottleneck in pgr5 was the b6f and insufficient flexibility in the low-potential chain redox tuning might account for the mutant pmf phenotype and the secondary consequences. Models of our findings are discussed. Significance statement The cytochrome b6f complex (b6f) is a plastoquinol (PQH2)-plastocyanin oxidoreductase, deprotonating PQH2 at the lumenal Qo-site while protonating plastoquinone (PQ) at the stromal Qi-site. These reactions require an evolutionary conserved electron bifurcation at Qo to drive the canonical Q cycle, making the b6f a major proton motive force (pmf) generator. Cyclic electron flow (CEF) between photosystem I (PSI) and b6f contributes to the pmf, enhances ATP production and protects PSI. Our data indicate that the pmf phenotype in pgr5 is linked to a diminished electrogenic capacity of the b6f, which is caused by an inefficient Q cycle. We provide evidence that during CEF an alternative Q cycle mode operates that requires PGR5 for sustained b6f function in the light.


Introduction
In linear electron flow (LEF), the two photosystems (PSII and PSI) act in series to ultimately reduce NADP + via the enzyme ferredoxin (Fd)-NADP(H) oxidoreductase (FNR). The cytochrome b6f complex (b6f) functionally interconnects the two photosystems (reviewed in 1), accepting electrons from plastoquinol (PQH2) and donating electrons to plastocyanin (PC). Functional b6f occurs as a homodimer, each monomer consisting of four major subunits (cytochrome b6, subunit-IV, cytochrome f (cyt.f), and the Rieske iron sulphur protein (ISP)), as well as four minor subunits (PetG, L, M and N). In addition, each monomer includes six cofactors: two b-hemes (bl and bh), two c-hemes (cyt.f and ci), one chlorophyll a and one βcarotene. Light indirectly induces b6f turnover: Upon oxidation by the primary PSI electron donor P700, PC extracts one electron from cyt.f, which is re-reduced by the Rieske ISP. The positively charged Rieske FeS domain moves towards the lumenal Qo-site, where an electron flow bifurcation occurs: PQH2 donates one electron to the Rieske ISP (part of the highpotential chain with a midpoint potential Em = 300-350 mV) and a second electron to bl (lowpotential chain; Em = -130 mV). PQ is re-reduced at the stromal Qi-site via bh (Em = -35 mV) and/or ci (Em = 100 mV, flexible as described below). Via the canonical Q cycle, the production of one PQH2 at Qi requires the oxidation of two PQH2 at Qo (supplementary Figure S1). The spatial proximity between bh and ci suggests electron sharing between the two and the presence of a membrane potential () promotes the shared electron to rest on bh red /ci ox (2). Furthermore, presence of a  is a general prerequisite for efficient b-heme oxidation (3). It is of note that heme-ci is unique since it lacks an amino acid axial ligand and thus might ligate with the semiquinone analogue NQNO (4,5) which downshifted the heme-ci midpoint potential from 100 mV to approximately -150 mV (6). Furthermore, heme-ci was proposed to engage in a Qi-site gating function (7): by ligating tightly in the oxidised state either with the phenyl group of F40 in subunit-IV (closed Qi-site), or, after transient heme-ci reduction, with (semi-)PQ. A recent cryo-EM structure of the spinach b6f complex contained the native PQ ligated to heme-ci (8). Following ligation-associated midpoint potential downshift of heme-ci (6), it is not clarified yet whether heme-bh or the quinone is reduced by heme-ci. Upon a single Qo-site turnover ( Figure S1), the heme-ci functions as terminal lowpotential chain redox carrier (6). Accordingly, the b6f low-potential chain harbours bl ox /bh ox /ci red after the first, and bl ox /bh ox /ci ox after the second Qo-site turnover that is associated with PQH2 formation at Qi. The (semi-)PQ in the Qi-site receives the electrons from ci red and/or bh red in the presence of , probably in a concerted and closely spaced process. Since not more than half the b-heme population is reduced per Qo-site turnover in uninhibited complex (9,10,11 and references therein), occurrence of bl red /bh red is unlikely. Part of failing to detect bl red /bh red might be that, in the presence of bh red (during a second canonical Q cycle turnover), oxidant-induced reduction of bl ox provides the strong reducing redox potential in the low-potential chain that is necessary to inject the first electron in the quinoneci ox ensemble (7), so that heme-ci would force the quasi-concerted PQ reduction (5,12). The deprotonation of PQH2 at Qo and the protonation of PQ at Qi couple electron transfer to proton translocation into the thylakoid lumen. The resulting transmembrane electrochemical proton gradient (pmf) fuels ATP synthesis via the chloroplast ATP-synthase. Besides LEF, which produces both NADPH2 and ATP, diverse auxiliary electron flow pathways, including cyclic electron flow (CEF) around PSI, contribute to the pmf and thereby equilibrate the NADPH2 to ATP output ratio of the light reactions (reviewed in 13). In addition, the pmf plays an integral photoprotective role, since the chemical component (pH) induces energy-dependent quenching (qE) and modulates the rate-limiting, pH-dependent oxidation of PQH2 at the Qo site, which is termed photosynthetic control (14, reviewed in 15). Hence, CEF creates a regulatory feedback loop linking the stromal redox poise to the efficiency of light harvesting and the rate of electron transfer. PGR5 (proton gradient regulation 5) has been first identified in Arabidopsis thaliana as component being involved in the regulation of the pmf via CEF (16). The corresponding knockout mutant in C. reinhardtii features multi-faceted phenotypes resembling its vascular plant counterpart (17,18): The algal pgr5 fails to induce qE-dependent NPQ and is extremely susceptible to PSI photodamage in response to high light (19,20) as well as fluctuating illumination (21). These defects have been attributed to an impaired acidification of the thylakoid lumen due to compromised Fd-PQ reductase-dependent CEF and a resulting lack of photosynthetic control in response to enhanced stromal redox pressure (19,22). However, the detailed mechanism of this CEF route is still elusive, as is the molecular role of PGR5. In the past, the association of FNR with the b6f (23-25) has been proposed to induce a switch from LEF to CEF: According to this model, FNR would tether reduced Fd in the vicinity of bh and ci, ultimately facilitating PQ reduction via a modified Q cycle that combines lumenal and stromal electrons (26)(27)(28). Our previous work showed less stable binding of algal FNR to the thylakoid membrane in the absence of PGR5 (20), suggesting a structural or regulatory contribution of PGR5 to this CEF pathway by influencing the localisation of FNR. By contributing to photosynthetic control and potentially providing the Fd-PQ reductase activity required for CEF, the b6f seems to be at the core of the phenotypes the absence of PGR5 produces. Therefore, we spectroscopically reinvestigated the impact of PGR5 on photosynthetic electron transfer in C. reinhardtii with a focus on b6f functionality by probing the behaviour of the high-and low-potential chain as well as the electrogenic capacity of the photosynthetic machinery. We provide evidence that during CEF a Fd-assisted Q cycle is active which requires PGR5 for sustained b6f function in the light.

1.
Alterations in electron transfer in Chlamydomonas pgr5. When assessing the function of PGR5 in photosynthetic electron transfer, PSI photodamage in pgr5 has to be anticipated (19,20). To re-evaluate the role of PGR5 in electron transfer regulation, we combined several in vivo measurement protocols to assess PGR5-dependent electron transfer under low-light autotrophic conditions in C. reinhardtii. For this dataset we compared pgr5 strain with WT, and furthermore investigated a partially PGR5-rescued strain C1 (19). The weak irradiance during growth and measurement provided permissive conditions for the classical pgr5 phenotype by avoiding photodamage (19,20). Thus, we obtained a PSI:PSII ratio of 1.13 ± 0.13, 1.18 ± 0.13 and 1.24 ± 0.18 in WT, pgr5 and complemented C1, respectively (N = 3 ± SD). The PSI:PSII ratio was calculated from the electrochromic shift (ECS) amplitude ratio of laser flash-induced charge separations in presence and absence of PSII activity, by using single turnover saturating flashes and adding PSII inhibitors. The actinic light intensity in the measuring cuvette produced an initial electrogenic signal that was characterized by the slope of the ECS intrinsic voltmeter. Similar initial membrane potential () formation rates were obtained in WT, pgr5 and C1, respectively, separating 216 ± 23, 193 ± 11 and 185 ± 12 e/s/photosynthetic chain (N = 3 ± SD). This actinic background light served to light-adapt cells for at least 30-min in the measuring cuvette. Another parameter that was varied was the stromal redox state which became strongly reduced in anoxic conditions, by depriving oxygen and thus inhibiting mitochondrial respiration (29). Where indicated, oxic control samples were poised with methyl viologen (MV) to abolish PSI acceptor side limitation and inhibit CEF. We functionally analysed both photosystems and the b6f when, except for MV samples, CEF and LEF were operational in light-adapted algae. Before documenting mutant behaviour, we will first describe the WT in more detail in the following three paragraphs.
The experiments were designed so that a steady state, light-adapted system was measured which was briefly perturbed by a saturating light pulse, for multiple turnovers, and then relaxed during several seconds of darkness. We monitored PSI performance in absence of PSII inhibitors first by following the classical Klughammer & Schreiber method (30), in which a saturating ms-pulse is superimposed on actinic background light ( Figure 1A). A net oxidation of P700 caused negative signals. The redox signals of fully re-reduced P700 after several seconds darkness were set to zero. To achieve full P700 oxidation during the strong pulse, algal PSII was poised with hydroxylamine/DCMU (not shown) which, by serving as reference, allowed to deduce partial oxidation ratios in uninhibited cells ( Figure 1A). The calculations are shown in Figure 1B, and the yield of P700 (YI) was about 0.27 in oxic WT controls. In other words, an additional 27% of light-adapted P700 were photo-oxidisable during the saturating ms-pulse. PSI acceptor side limitation (YNA) in these samples was 0.51 and donor side limitation (YND) was 0.22. Addition of MV abolished YNA and CEF which, despite active PSII, resulted in 57% pre-oxidised P700 in the actinic background light due to YND. Regarding YI, an additional 43% of P700 could get oxidised during the saturating mspulse. When oxygen depletion produced strongly reducing conditions, anoxic samples showed an increased YNA in the light and a lower YI fraction of P700. YND was also slightly diminished in anoxic samples.
Next, we monitored the b6f redox changes in the same samples. It is of note that the measured signals, especially during the ms-pulse, were sometimes small in amplitude but they were absent in b6f-lacking mutants (supplementary Figure S2). When setting cytochrome f (cyt.f) redox signals in the background light to zero, the saturating ms-pulse caused a net cyt.f oxidation of about -0.4 a.u. in oxic samples (hatched box in insert of Figure 1C). In the dark, after the pulse, the fast cyt.f net reduction phase had an amplitude of about +0.65 a.u. and finished in ~50 ms. Both amplitudes of cyt.f net oxidation (-0.2 a.u.) and reduction phase (+0.4 a.u.) were smaller in anoxic samples, compared to oxic conditions. The time to reach maximal oxidation during the pulse was shorter in anoxic samples (inset Figure 1C), since it was related to the lower YI fraction under these conditions ( Figure 1B). The cyt.f re-reduction kinetics in darkness was faster in these samples as well (inset Figure 1C). When MV was present, there was almost no cyt.f net oxidation during the pulse. As elaborated further at the end of this section, no distinct fast cyt.f reduction phase was observed during 100-ms darkness. When monitoring redox changes of the hemes bl/bh during the saturating ms-pulse in oxic samples ( Figure 1D), a net reduction of about +0.1 a.u. was observed. Again, redox signals in the background light before the pulse were set to zero. In darkness after the saturating pulse, the hemes bl/bh net oxidation was finished in ~50-ms in oxic samples and the redox state was slightly more oxidised for several seconds darkness compared to the steady state established in the background light. The hemes bl/bh net reduction amplitude during the ms-pulse was a bit more pronounced in presence of MV. After the pulse, however, there was a significantly larger amplitude of hemes bl/bh net oxidation in the MV samples which transiently reached -0.2 a.u. compared to the pre-equilibrated steady state. The oxidation was finished after ~300-ms darkness. In anoxic samples during the ms-pulse, hemes bl/bh net reduction amplitude reached a slightly lower plateau earlier. In darkness, hemes bl/bh net oxidation amplitude was small and the phase finished in ~25-ms. In contrast to oxic samples, a unique hemes bl/bh redox feature was a net reduction phase that started by no later than 100-ms darkness in anoxic WT. PSII performance was monitored by chlorophyll fluorescence measurements of dark-and light-adapted algae. The maximum quantum yield of PSII, Fv/Fm, was expectedly different in oxic and anoxic samples and was comparable in the strains (supplementary Table S1). When dark-adapted WT was illuminated with the background light for a rather short period of 10-s, the photochemical quantum yield of PSII (PSII) as well as the PSII efficiency factor (qP) were indifferent from the light-adapted state after 30-min in both oxic and anoxic conditions ( Figure 1E). In agreement with a previous report (31), qP was slightly lower in presence of MV. On the level of QAre-oxidation after a saturating pulse (32), a multiphasic kinetics was observed in oxic algae during the 6-s of darkness shown in Figure 1F. The kinetics of QAre-oxidation reflect the PQ pool redox state that is in equilibrium with QA in the dark (32). Starting from an almost fully reduced QApool immediately after the pulse (33), the pool became quickly re-oxidized by a fraction of ~0.3 in the first ~50-ms of darkness. From there to ~500-ms darkness, further QAre-oxidation was absent in oxic samples and a second slow re-oxidation phase followed thereafter. It is of note that the retardation phase between 50-ms and 500-ms darkness, representing downstream electron transfer via the b6f (32), was not established in dark-adapted algae after a short 10-s illumination (supplementary Figure S3). The light adaptation-specific dark kinetics (during progressing pmf consumption) link the redox equilibration via the b6f to a reservoir of electrons accumulating downstream of PSI. In MV-treated samples, QAre-oxidation was identical in the first ~20-ms darkness compared to oxic controls ( Figure 1F). The abovementioned QAre-oxidation slow down between 50-ms and 500-ms is lacking in MV-treated samples, suggesting a somewhat faster QAre-oxidation in presence of MV (31) since no electrons could accumulate downstream of PSI. In strongly reduced anoxic cells, QAreoxidation was slower during the first ~500-ms darkness ( Figure 1F).
When the above experiments were performed in the pgr5 mutant, clear differences were observed in the electron transfer chain compared to WT: in oxic pgr5, a careful look at the kinetics during the ms-pulse showed a faster establishment of P700 + plateau (hatched box in Figure 2A), which will be explored in more detail at the end of this section. The fractural P700 redox analysis in anoxic conditions revealed that YNA was strongly diminished in pgr5 and YND appeared to be increased to 0.33 ( Figure 2B). Figure 2B also shows that oxic samples were like WT and MV addition produced slightly less YND in pgr5. The net cyt.f oxidation amplitude during the pulse was smaller in oxic pgr5 (-0.2 a.u., Figure 2C). In this sample, the reduction phase amplitude was slightly larger (+0.8 a.u.) but the decay kinetics resembled WT, which will also be shown further below. Whereas MV samples were indistinguishable from WT, cyt.f redox changes in anoxic pgr5 differed in several aspects. The net oxidation amplitude was almost non-existent in anoxic cells whereas the reduction amplitude was, unlike the oxic vs. anoxic difference in WT, indistinguishable from oxic pgr5. Unlike anoxic WT, fast cyt.f reduction kinetics were absent after the pulse and, in fact, slowed down significantly in pgr5. The hemes bl/bh redox signals during and after the saturating ms-pulse were like WT with two exceptions in anoxic pgr5 ( Figure 2D). The hemes bl/bh oxidation phase after the pulse finished a bit later than in WT and the onset of rereduction was significantly delayed, which will be quantified further below. Finally, we also observed differences on the level of PSII but this time in the oxic samples. Light-adapted oxic pgr5 showed higher PSII which, as proposed previously (19), might be the result of higher LEF rates in the mutant (asterisk in Figure 2E). In line with this finding, oxic light-adapted pgr5 samples also showed higher qP. When plotting the QAredox relaxation after full reduction by a saturating pulse ( Figure 2F), the fast QAre-oxidation phase up to 20-ms darkness and the slow phase after 500-ms were similar in oxic pgr5 and WT (cf. Figure 1F). However, oxic light-adapted pgr5 lacked the QAre-oxidation slow down between 50-ms and 500-ms darkness ( Figure 2F), which may point to less accumulated electrons downstream of PSI. The latter would be in line with previous studies in Arabidopsis pgr5 which assigned a disturbed electron flow downstream of PSI, probably on the level of Fd (34,35). The kinetics in the presence of MV and in anoxic cells were similar, the latter being slightly faster beyond 100-ms darkness in pgr5. Establishing the slowdown phase between 50-ms and 500-ms, as a distinguishable feature of the oxic WT and C1 line, required light adaptation for more than 10-s (supplementary Figure S3).
As mentioned above, the rates of P700 oxidation have been determined as well as dark relaxation kinetics on the levels of P700, cyt.f and hemes bl/bh ( Figure 3). The PGR5complemented C1 has been analysed (supplementary Figure S4) and is also included. Singleexponential functions were used to determine kP-ox, the P700 oxidation rates during the pulse ( Figure 3A). When following a previous assumption (19) by ascribing the origin of higher PSII in oxic pgr5 ( Figure 2E) to higher LEF rates, the faster P700 oxidation rate during the ms-pulse ( Figure 3A) would support this hypothesis of faster electron flow via pgr5 PSI. In the light of undelayed QAre-oxidation as proxy for missing electron accumulation downstream of PSI in oxic pgr5 ( Figure 2F), the faster kP-ox may be tied to unregulated electron flow between PSI and its sinks (34,35) which could contribute to diminished photosynthetic control after the saturating pulse. The C1 line was indistinguishable from WT.
Since the light intensity during the pulse was saturating, a faster oxidation rate in all anoxic samples could result from a combination of events in these conditions ( Figure 3A). Higher kPox might reflect a larger PSI antenna size in anoxic cells because of state transitions (36). Since PSI acceptor side limitation was enhanced in the reducing anoxic WT and C1 line ( Figure 1B, Figure S4B), a faster apparent P700 oxidation during the multiple turnover pulse could result from a smaller pool of photoactive PSI. The latter would stem from inefficient charge stabilisation at the PSI acceptor side and the hampered relaxation to the oxidized P700 -the multiple turnover process during the saturating pulse to empty the immediate donor pool of reduced PC and cyt.f (30). As elaborated above ( Figure 1B, Figure 2B, Figure S4B), PSI donor side limitation was enhanced in the presence of MV which suggests facilitated oxidation of primary and secondary electron donors of P700. It could explain why WT P700 was slightly faster oxidized ( Figure 3A). However, slightly faster P700 oxidation was not observed in MV-containing pgr5 and C1.
P700 reduction in the dark occurred in multicomponent decay kinetics (37). During the first few ms after the pulse and especially in reducing conditions, the probability of charge recombination events between P700 + and its primary and secondary reduced acceptors increases (38). We therefore disregarded the initial 4 ms of darkness and calculated the first (fast) component, k1P-red, of a 2.5-s two-exponential decay as an estimate for the rate of electron flow that was occurring in the light ( Figure 3B). We observed in all anoxic cells that there was a slowdown of k1P-red after the pulse in darkness, to a significant extent in WT and C1. One possibility to explain the P700 reduction slowdown would be enhanced photosynthetic control in these samples. The large build-up of a pH, reported in anoxic algae (39), would slowdown PQH2 oxidation at the b6f Qo-site and therefore reduction of PC. All strains showed a smaller k1P-red in presence of MV (asterisk in Figure 3B), which could result from a strongly pre-oxidized pool of primary and secondary donors. Photosynthetic control would slowdown the total electron transfer rate and the PQ pool would become more reduced due to PSII activity. However, a putatively enhanced photosynthetic control contradicts the faster QAre-oxidation ( Figure 1F, Figure 2F, Figure S4F), which is in equilibrium with the PQ pool in the dark (32) and should be slowed down if the latter was in fact more reduced. Later, we will revisit the smaller k1P-red in presence of MV in the light of b6f performance.
When calculating the reduction rate of cyt.f from exponential decay of the kinetics after the pulse (kf-red, Figure 3C), oxic samples were identical and significantly slowed down in the presence of MV. Using a multiple turnover protocol, it is important to stress that kf-red in Figure 3C also strongly depended on the oxidation level of the PC pool at the end of the pulse which would have influenced the redox equilibration time between PC and cyt.f. Similarly, the PQ pool redox state was also relevant for the calculated rates. The PC pool was expected to be oxidized the presence of MV and to answer the question whether reduction of cyt.f in Figure 3C is truly slowed down, we performed single turnover experiments as well in Results section 3. In addition to the possible impact of the PQ and PC pools redox states in Figure  3C, there is a b6f intrinsic process, explored in detail further below, which regulates the electrons entering the high-potential chain via a mechanism that depends on electrons exiting the low-potential chain (reviewed in 40).
The latter process was measured from exponential decay of the kinetics after the pulse (kb-ox, Figure 3D). MV addition significantly lowered kb-ox in WT and C1. Oxic pgr5 controls produced relatively low rates and the inhibitory effect of MV was statistically not significant. Considering the relatively large oxidation amplitudes after the pulse in the MV samples ( Figure 1D, Figure 2D, Figure S4D) in combination with lower kb-ox ( Figure 3D), it is likely that electrons accumulated in the low-potential chain during illumination which were slowly transferred to PQ at the Qi-site in the dark. According to the above-mentioned coupling concept, the lack of electrons exiting the low-potential chain in presence of MV could result in low kf-red shown in Figure 3C. The anoxic WT produced higher kb-ox ( Figure 3D) which, again, would enhance the efficient release of "trapped" high-potential chain electrons ( Figure  3C). On the contrary, the pgr5 mutant and the partially complemented C1 strain did not produce higher kb-ox ( Figure 3D), thus failing to efficiently reduce cyt.f ( Figure 3C). The C1 line was less severely affected than pgr5. We also calculated the slow dark-reduction rates of hemes bl/bh in anoxia from single-exponential fits (kb-red, Figure 3E) and the WT, which also showed an earlier onset of this phase, was significantly faster than C1 and pgr5. The rereduction phase in anoxic cells most likely represents re-reduction of heme-bh, which is accomplished in a few seconds in the dark (41,42). The origin of the slowdown is unknown, but it could be that delayed hemes bl/bh oxidation in pgr5/C1 lowered apparent re-reduction rates.
2. Electrogenic capacity of the photosynthetic chain in Chlamydomonas pgr5. The data in Figure 1 and Figure 2 indicate that there were alterations in the mutant electron transfer chain. In a similar fashion to the P700 measurements where a superimposed saturating light pulse empties the immediate donor pool of PSI during several turnovers (30), we analysed the charge separation capacity of the photosynthetic apparatus by recording the electrochromic shift (ECS) which is shown in Figure 4 (see supplementary Figure S5 for C1). The ECS, which serves as intrinsic voltmeter (reviewed in 43), was recorded in the background light, during and after the saturating pulse. In the steady state, before the short pulse, the slope of the signal was zero. The additional membrane potential () that was built up at the very onset of the 22-ms pulse depended on the immediate availability of electrons in the photosynthetic chain and the photochemical yield of both photosystems, which we determined above. The rate of the initial  generation, kini, was derived from the positive slope during the first 2-ms of the pulse, which is shown as green symbols in WT (inset of Figure 4A) and pgr5 (inset of Figure 4B). Oxic samples produced ~4 additional, stable charge separations during the pulse, also in the presence of MV. Anoxic cells generated less additional  during the pulse, and WT was slightly more competent than pgr5. During several seconds of darkness when ATP synthase was active, the  collapsed to ~-4 units in oxic and anoxic samples, and to ~-8 units in MV samples. It is of note that we did not check for the total pmf partitioning between pH and . Instead, Figure 4 probed only the  which, if it was the dominating pmf component, would produce large amplitudes when driving ATP synthesis during 2.5-s darkness. It remains to be tested whether this effect was contributing to the large ECS amplitudes after the pulse in the MV samples. At the end of the pulse a new steady state (zero slope) was established in oxic and anoxic WT only. Upon exhaustion of the immediate donor pool, the  generation efficiency at the end of the pulse was derived from a Dark Interval Relaxation Kinetics-based protocol (44). To obtain the progressed  production rate at the end of the pulse, kend, the apparent light-dependent  formation (slope of orange symbols in Figure 4A and Figure 4B) had to be corrected for the ATP synthase-dependent  consumption (negative slope in the dark shown by yellow symbols in Figure 4A and Figure 4B), assuming that the latter was identical just before onset of darkness.
The kini values are shown in Figure 4C and indicate that in presence of MV, with exception of pgr5, the pulse-induced  formation was less efficient compared to controls. Nonetheless, the MV effect existed in pgr5 as well which showed a relatively low kini in oxic controls already. Considering the electrogenic competence associated with qP, PSII and YI in the presence of MV (Figure 1, Figure 2 and Figure S4), the observed slowdown of kini ( Figure  4C) was likely resulting from the poorly functioning b6f ( Figure 3C-D). As will be discussed later, such a situation would imply lower b6f-borne electrogenicity and less immediately oxidisable PC. The enhanced b6f low-potential chain performance in anoxic WT ( Figure 3C-E; in the presence of lower qP, PSII and YI) could explain why kini was least diminished among the anoxic strains in Figure 4C. Note that, compared to oxic controls, YI was unaffected in anoxic pgr5, whereas PSII and qP were like in WT. Still, kini and the b6f rates were lowest in anoxic pgr5. As expected upon exhaustion of the immediate donor pool, kend was diminished compared to kini ( Figure 4D). Once the pool of reduced PC and cyt.f were exhausted, restricted PSII performance could have contributed to the lower kend in anoxia.
It is of note that throughout the study, the partially complemented C1 line, which accumulates ~75% of WT PGR5 levels under the control of its native promoter (19), resembled WT in oxic conditions whereas it tended to partially perform like pgr5 in anoxia. With exception of P700 in anoxic conditions ( Figure S4B), this was most apparent on the levels of b-hemes oxidation (kb-ox in Figure 3D). For cyt.f reduction (kf-red in Figure 3C) and electrogenicity (kini in Figure 4C), both rates were significant faster under anoxia as compared to pgr5 but still slower than WT. To eliminate PGR5 titration effects in C1, we generated an independent PGR5-complemented line which, besides the expected P700 redox behaviour ( Figure 5A), also produced WT-like cyt.f reduction rates after the pulse (kf-red in Figure 5B), as well as kb-ox ( Figure 5C) and kini ( Figure 5D). We noted that hetero-phototrophic cells varied slightly in their rates, compared to photoautotrophic cells (cf. Figures 1 and 2). As we will discuss later, the zeocin system showed that electrogenicity of the photosynthetic chain relied in part on the availability of the immediate P700 donor pool which was governed by Qi-site turnover in a PGR5-dependent manner. For feasible photoautotrophic culturing under low light, the C1 line with a native PGR5 promotor is preferred over the zeocin-induced, complemented strain.
3. PGR5-dependent redox finetuning of the b6f low-potential chain and implications on the Q cycle. In order to rule out possible dark redox equilibration artefacts (owing to different pre-oxidation levels in the steady state), this section introduces single turnover measurements. Here, the light-adapted cells have been assayed in absence of PSII photochemistry and upon a 30-s dark period. This setup ensures reduction of primary and secondary donors of PSI as well as pmf consumption, especially since the pH governs photosynthetic control (14, reviewed in 15). The single b6f turnover passes the electron hole from oxidised c-heme in cyt.f to the Rieske ISP which, after swapping back the FeS domain closer to cytochrome b6 subunit, is reduced at the Qo-site in a bifurcated process that also reduces hemes bl/bh. When electrons are passed on heme-bh, a  was generated which we monitored via ECS signals. Redox changes in the b6f ( Figure 6A) and the corresponding ECS changes ( Figure 6B) were assayed in oxic samples. The ECS kinetics in Figure 6 are relative and are composed of three phases (reviewed in 43,45). First, the unresolved a-phase finished in less than 1 µs before the first measured signal and represented PSI photochemistry in this experiment. The a-phase upon the sub-saturating flash generated ~0.4 -0.5 charge separations per total PSI, serving for ECS kinetics normalisation. In the ~10-ms range, the b-phase corresponded to the b6f-dependent  generation and the cphase was dominated by ATP synthesis activity which consumed the  produced by the flash. In the ECS kinetics in Figure 6, the b-phase was deconvoluted by subtracting from the total ECS kinetics the c-phase, which instantaneously followed first-order exponential decay (46,47). Regarding the decay rate of the c-phase, we observed no significant differences in the two strains (supplementary Figure S7). We also measured the b6f redox kinetics and ECS in presence of MV (Figures 5C and 5D) as well as in anoxic conditions (Figures 5E and 5F). On a time scale after injecting an electron hole into the b6f, cyt.f reduction (kf-red derived from single exponential fits in the b6f redox kinetics panels) preceded the electrogenic b-phase (k derived from single exponential fits in the ECS kinetics panels) and the last phase was the relatively slow oxidation of hemes bl/bh (kb-ox derived from single exponential fits in the b6f redox kinetics panels). A statistical evaluation of kf-red ( Figure 6G), k ( Figure 6H), and kb-ox ( Figure 6I) is shown as well as the amplitude of the b-phase relative to one charge separations per PSI ( Figure  6J).
In oxic samples, independently of MV, cyt.f oxidation was finished before the first record at 300 µs after the flash and resulted in an amplitude of ~-0.1 units compared to the reference signal before the flash (circle symbols in b6f redox kinetics panels). In anoxic samples, oxidation of cyt.f was slowed down slightly, finishing between ~1-ms and ~2-ms after the flash. With exception of MV samples, cyt.f reduction by the FeS domain was initiated at ~1ms, yielding similar kf-red values in oxic and anoxic conditions. When MV was added, kf-red was lowered significantly (5% and 9% residual rates of oxic WT and pgr5, respectively) and a delayed onset of reduction became apparent between ~5-ms and ~10-ms.
Before cyt.f was getting reduced in oxic and anoxic samples (during the first ms), net redox changes of hemes bl/bh were very small (square symbols in b6f redox kinetics panels). Only after the onset of cyt.f reduction, a net reduction of hemes bl/bh became apparent, which coincided with the b-phase (square symbols in ECS kinetics panels). The sequential reduction of cyt.f and hemes bl/bh was expected and was also observed in MV samples but there was a significant slowdown of the low-potential chain turnover. The amplitude of hemes bl/bh reduction in the presence of MV appeared larger since Qo-site turnover, with k as proxy, was less slowed down than kb-ox (23% and 26% residual k of oxic WT and pgr5 in Figure  6H vs. 10% and 28% residual kb-ox in Figure 6I). The inhibitory MV effect on kb-ox in pgr5 was less efficient, which might account for smaller net reduction amplitude during the 10-ms phase ( Figure 6C).
In both strains under oxic and anoxic conditions, the net reduction amplitudes in the hemes bl/bh redox kinetics during the first 10-ms differed ( Figure 6A and Figure 6E). During the initial phase in oxia and anoxia, the electron transfer rate k between hemes bl and bh (which depends on kf-red and the following electron injection into the low-potential chain upon Qoturnover) was not changed in the respective strains. However, k differed between WT and the less efficient pgr5 (Figure 6H), although hemes bl/bh net reduction amplitude was comparable ( Figure 6A and Figure 6E). As discussed later, this suggests that oxidation of heme-bh (yielding PQH2 eventually) was slower during the electrogenic 10-ms phase in pgr5 and thereby allowed a similar net reduction amplitude compared to WT.
The relative amplitude of the b-phase was between 50-85% of the a-phase and tended to be slightly higher in pgr5 ( Figure 6J). Further experiments need to clarify whether less efficient PQH2 formation at the pgr5 Qi-site, which would weaken stromal charge neutralisation compared to WT, was responsible for slightly higher b-phase amplitudes in the mutant. Nevertheless, the  generated by one b6f turnover in our conditions was close to values in earlier reports which attributed similar fractions of one charge separation across the whole membrane when measuring electron transfer 'within' the membrane bilayer from hemes bl to bh in the bc1 complex (48,49). After injection of Qo-site electrons into the low-potential chain, the slower hemes bl/bh oxidation phase (kb-ox in Figure 6I) produced faster rates in anoxic WT only. Whether the dysfunctional kb-ox tuning was responsible for slower k in pgr5 needs to be examined.

Discussion
PGR5 is an important regulator of photosynthetic electron transfer, however, its function has not been linked to the operation of the b6f. Our data indicate a dysfunctional b6f in the absence of PGR5. The b6f enigmatically receives a stromal feedback from PSI. It modifies the Q cycle in strongly reducing conditions or is inhibited when being disconnected from PSI signals in the presence of artificial electron acceptors. We provide evidence that PGR5 is functionally involved in a modified Q cycle which has access to stromal electrons and operates in WT but less efficiently in pgr5. Figure S8 in the supplementary discussion for a detailed interpretation of the results in moderately reduced stroma. Our pgr5 findings under oxic conditions, arguing in support of previous reports (19,34,35), suggest unregulated coordination of electrons between PSI and its sinks that produced higher PSI-borne LEF rates and resulted in a more oxidised electron carrier pool downstream of PSII.

Please refer to supplementary
We did also observe a more oxidised electron carrier pool in anoxic pgr5 -but under these reducing steady state conditions we located it downstream of the b6f. The undersupply of electron donors rendered PSI more oxidisable by the pulse and thus diminished acceptor side limitation in pgr5 (cf. A and B panels of Figures 1, 2, and Figure 5A). Accordingly, the highpotential chain was even more oxidised in the steady state (cf. cyt.f oxidisability in C panel inserts of Figures 1, 2, and Figure S6A) and redox relaxation in the mutant b6f was slowed down in the dark ( Figures 3C and 3D, Figures 5B and 5C). The PGR5-dependent bottleneck at b6f exhausted the immediate PSI donor pool and thus lowered the initial  formation rate significantly when rapidly transitioning anoxic mutant cells to strong light (kini in Figures 4C  and 5D). Notably in the single turnover measurements (Figure 6), which were carried out in absence of PSII activity when pH was collapsed and photosynthetic control was cancelled, the pgr5 mutant was generally less efficient to separate charges across the membrane via the b6f (k in Figure 6H) and failed to accelerate the electron discharge from the low-potential chain in anoxia (kb-ox in Figure 6I). This argues that the absence of PGR5 has a direct impact on b6f function.
The observed steady state pgr5 phenotype in anoxic conditions can also be linked to the b6f function when the WT showed enhanced Qi-site activity upon kb-ox redox tuning ( Figure 7A). This tuning feature prevents deleterious electron back-up in the low-potential chain, which we will explore further below, and thus promotes photosynthetic (self-)control via the b6f in high light conditions. It is of note that light acclimation in the cuvette (i.e., steady state actinic light) did not increase photosynthetic control since we did not observe increased YND in the WT ( Figure 1B). However, k1P-red in anoxic cells was slowed down after the saturating pulse ( Figure 3B), suggesting enhanced photosynthetic control compared to oxic conditions. Importantly, YND (established before the pulse) and k1P-red (measured after saturating pulse) were probed at two different pmf levels. A closer look at the electric pmf component after the pulse shows that the  was consumed back to the steady state level in about 10-ms to 20-ms darkness (cf. insert of Figure 4A). Therefore, it could be that photosynthetic control was established in the course of the pulse and contributed to k1P-red. In anoxic conditions, the pgr5 mutant failed to discharge low-potential chain electrons (for explanation, see scheme in Figure 7B). Manifested in single turnover experiments, the kb-ox redox tuning defect eventually impaired the steady state turnover. The inhibited steady state b6f in anoxic pgr5 would produce less YNA due to increased YND, as observed. The low-potential chain electron back-up in the steady state might slowdown the b-heme re-reduction phase in anoxic pgr5 (kb-red in Figure 3E) once these electrons were discharged upon  consumption in the dark. Given the significant b-heme oxidation (D panels of Figures 1, 2, and S4) and the -dominated pmf when probing the ECS decay amplitude in the dark ( Figures 4A, 4B, and S5), electron back-up in the low-potential chain was probably also contributing to the poor steady state b6f turnover in MV samples (supplementary Figure S8C). Interpretation of the steady state samples can be inferred from the single turnover experiments in the presence of MV where, in both cases, kb-ox was significantly slower ( Figure 3D, and Figure 6I). The inhibition of kb-ox by MV suggests that b6f function relies on stromal redox signals associated with PSI turnover. Figure 7 summarises the multiple-turnover measurements under anoxic conditions in a scheme and proposes a safety mechanism of the b6f turnover (see respective arrow heads in Figure 7), which couples b-heme oxidation to cyt.f reduction via the Rieske ISP (reviewed in 40,[50][51][52]. In the following, we will draw the attention to the b6f turnover and focus on the low-potential chain which, as introduced, underlies a canonical Q cycle. Thus, the b-heme kinetics under oxic conditions in Figure 6A followed the transient net reduction/oxidation and, compared to the pre-flash reference signal, the slightly more oxidised b-hemes after ~100 ms resulted from a small pre-reduced heme-bh population ( Figure S1). Despite the drawback that both b-hemes were spectroscopically monitored as one population at room temperature (41) and we could not determine the heme-ci redox state, our observed changes in the low-potential chain oxidation behaviour under anoxic conditions ( Figure 6E) support a modified Q cycle with access to stromal electrons (see below). However, the initial fast oxidation phase (≤ ~10-ms after the flash, see two left panels in Figures 7A and 7B) can be explained on grounds of a canonical Q cycle since the samples were strongly pre-reduced. The 30-s darkness before the flash were sufficient to pre-reduce heme-bh but not heme-bl (41,42), and to reduce heme-ci in absence of  (2). The heme-bh dark reduction rates in anoxic samples (41,42) indicate that bh red /ci red was very likely to occur, yielding the pre-reduced Qi-site. Yet, pre-reduction was not sufficient to drive PQH2 formation in the dark since the necessary  prerequisite was not met (3) and bh red oxidation was only achieved with the flash. Regarding the probed b-hemes, the anoxic samples contained bl ox /bh red before the flash, and bl ox /bh ox at the end of the measurement (thus the negative b-hemes signals in Figure 6E). The relatively flat b-heme redox signals during 10ms were a result of bl red /bh red disqualification (9,10,11 and references therein) so that the fast bh red oxidation (at a rate ≤ k) was driven by reduction of bl ox upon Qo-site turnover. To explain the faster oxidation phase in WT vs. pgr5 after the electrogenic 10-ms phase (kb-ox in Figure 6I), a modification of the canonical Q cycle is required. With ongoing consumption of , bh red /ci ox would transiently convert to bh ox /ci red (2), which eventually makes the Qi-site 'PQ-accessible' for the next turnover upon weakening the ligation of ci red with F40 in subunit-IV (7). We propose that this canonical scenario produced the corresponding kb-ox in anoxic pgr5 (see two right panels in Figure 8A). In anoxic WT, after ~10-ms darkness, bh red /ci ox has access to stromal electrons so that bh red /ci red is formed and a second PQH2 is produced at the Qi-site (see two right panels in Figure 8B). The accessible substrate in the WT Qi-site allows faster bh red oxidation and is not governed by the -dependent transition to bh ox /ci red , unlike in the mutant (cf. similar ECS decay in Figures 5F, and S7). The second oxidation of bh red (yielding kb-ox beyond ~10-ms) is slower than the one during the first Qi-site turnover due to absence of the reducing pressure on the b6f (Qo-site electrons entering the low-potential chain). A facilitated generation of ci red from stromal electrons could increase the chance for the, compared to oxic conditions, less abundant PQ to enter the Qi-site binding pocket (7). By enhancing low-potential chain oxidation, this WT b6f feature is an optimisation to highly reducing conditions, as evidenced by our steady state observations where the anoxic pgr5 displayed an arrested b6f. In reducing conditions, the mutant b6f underlies the intrinsic, short-circuit-preventing process that influences cyt.f reduction rate by governing the interaction between the Rieske ISP FeS domain and cytochrome b6/cyt.f (12). Unless b-hemes are not oxidised in steady state pgr5, the reduced flexible FeS domain will not swap closer to cyt.f to release the "trapped" high-potential chain electron (reviewed in 40).
In conclusion, PGR5 is involved in the Q cycle operation under anoxic conditions which attributes Fd-PQ-reductase activity to the b6f, as originally suggested by Arnon (53). A stromal redox poise activates this specific b6f function, which increases the H + /e ratio of 2 that is obtained in LEF-favouring conditions during a canonical Q cycle. Under CEFpromoting conditions, the role of PGR5 is to ensure reduction of the heme-ci in the b6f, likely via Fd. The b6f low-potential chain activity in pgr5 is slightly lower during single turnovers and steady state activity in CEF-favouring conditions requires the PGR5 polypeptide. Thus, PGR5 is required for efficient Q cycle in the b6f, which is in turn crucial for lumen acidification to trigger the onset of photosynthetic control and qE. Whether PGR5 action during the redox poise is direct or indirect needs to be further investigated. In yeast-based protein-protein interaction studies, candidates for PGR5 interaction were cytochrome b6 and Fd (34). The PGR5-dependent CEF effector proteins PETO and ANR1, which interact with the algal b6f (54)(55)(56)(57), are part of the Fd interactome as well (58,59) and could have Fd shuttling functions. The yeast two-hybrid study also showed that PGRL1 interacts with PGR5 and FNR (34), and BiFC studies indicated protein-protein interaction between PGRL1 and ANR1 (56). Notably, FNR has been shown to co-purify in b6f preparations (23)(24)(25) and the algal pgr5 retains less FNR on the thylakoid membrane (20). Interestingly, MV-treated thylakoid membranes do not retain FNR either (60). Thus, FNR membrane recruitment itself might (in)directly rely on reduced Fd since it is stimulated in anoxic conditions and requires functional b6f and PSI (61). By disturbing the electron flow downstream of PSI, probably on the level of Fd (34,35), the FNR-anchoring trigger might be weakened in pgr5. This may lower functional Fd availability to reduce the heme-ci, whereas other sinks like hydrogenase might successfully compete for Fd in anoxic pgr5 (62,63). One could imagine that bound FNR could have allosteric effects on Qi-site turnover since the b6f redox kinetics in the presence of MV, in which FNR is expected to be soluble (60), showed attenuated resemblance of b6f samples treated with the Qi-site inhibitors MOA-stilbene (10) and NQNO (64). Further experiments are required to provide mechanistic insights into the PGR5dependent electron transfer, permitting reduction of the b6f Qi-site during alternative Q cycle.

Strains and cell cultures
As described previously (19), Chlamydomonas reinhardtii WT strain t222+, pgr5 and a complemented line, termed C1, were used. The complemented C1 strain accumulated ~75% of WT PGR5 levels (19). Cells were cultivated on agar-supplemented plates in Tris-acetatephosphate medium (TAP, 65) at 20 µmol photons m -2 s -1 . When growing cells for experiments, liquid Tris-phosphate medium was devoid of acetate (TP). Stirred cultures were grown at 10 µmol photons m -2 s -1 (16 h light/8 h dark) and were bubbled with sterile air at 25 °C. Grown cultures were diluted ~6-fold at least once after inoculation and grown to a density of ~2 × 10 5 cells mL -1 before harvesting (5000 rpm, 5 min, 25°C). For experiments with PGR5-complemented lines that express zeocin resistance, WT, pgr5 and pgr5::PGR5 were grown in TAP in same conditions as for TP cells, except without air bubbling. One day before the experiments, cells were diluted in fresh TAP and 5 µg/mL zeocin was added to pgr5::PGR5 cultures. Cells were resuspended at 20 µg chlorophyll mL -1 in TP supplemented with 20% (w/v) Ficoll and shaken vigorously in dim light. Optionally for oxygen-deprived conditions in the dark, cells were supplemented with 50 mM glucose, 10 U glucose oxidase and 30 U catalase in a cuvette, and then overlaid with mineral oil for at least 30 min before measurements. In presence and absence of PSII photochemistry, these illuminated cells will be referred to as anoxic.
Generation of PGR5 complemented lines using dicistronic system.
The PGR5 gene (Cre05.g242400) was amplified from genomic DNA extracts using forward (5'-GCCCCGAATTCATGCTGGCCTCCAAGCCCGTTGTTG) and reverse oligos (5'-CTAGTCTAGATTAAGCCAGGAAGCCAAG) that harboured the underlined EcoRI/XbaI restriction sites. The digested fragment was introduced into a dicistronic expression vector (66,67) for PGR5 under the control of the PSAD promoter. The construct conferred zeocin resistance as well, since PGR5 expression was linked to ble via a skipping peptide FMDV2A, and DNA was introduced to the pgr5 nuclear genome by electroporation (25 µF, 1kV). Transformants were pre-selected on TAP agar plates supplemented with 10 µg/mL zeocin.

Chlorophyll fluorescence analysis
The LED-based spectrophotometer (JTS-10, BioLogic, France) was equipped with a Fluo_59 accessory in fluorescence mode. Probing was performed with 520-nm LED measuring pulses (350 µmol photons m -2 s -1 ) and light-detecting photodiodes were protected from scattered actinic light by using appropriate 3-mm thick filters (reference diode: BG39; measuring diode: LPF650+RG665, Schott, Mainz, Germany). After 30-min dark-adaptation, F0 was detected by probing in the dark and Fv/Fm ((FmF0)/Fm) was obtained by a 250-ms saturating pulse (520-nm LED, 5000 µmol photons m -2 s -1 ). Adaptation to 630-nm actinic light (LEDs emitting 4700 µmol photons m -2 s -1 which yielded ~150 µmol photons m -2 s -1 at the measuring cuvette) was carried out by regularly resuspending the sample for at least 30-min in an open cuvette. Oxygen-deprived cells were not mixed. The photochemical quantum yield of PSII, PSII, was calculated from (Fm'Fs)/Fm' by probing maximal (Fm') and steady state (Fs) fluorescence in the light. The PSII efficiency factor, qP, was derived from (PSII×Fm)/Fv (68). The QAre-oxidation in the dark was calculated as described elsewhere by using the Stern-Volmer relationship (32).
Picked colonies, putatively carrying the dicistronic PGR5 construct, were grown at 10 µmol photons m -2 s -1 and submitted to spot tests performed on TP agar plates (supplementary Figure S9). Selection was based on WT-like chlorophyll fluorescence after 24-h exposure to 200 µmol photons m -2 s -1 , using a Maxi-Imaging PAM chlorophyll fluorometer (Walz, Germany).

Time-resolved absorption spectroscopy
Time-resolved measurements are expressed as I/I and were carried out with the JTS-10. P700 redox changes were measured with combined detection LEDs at 705 and 740 nm and associated interference filters at 705 and 740 nm (FWHM: 6 nm and 10 nm respectively). The light-detecting diodes were protected from scattered actinic light by a RG695 filter (Schott, Mainz, Germany). For reference purpose, PSII inhibitors hydroxylamine (HA at 1 mM final, from 1M aqueous stock) and 3-(3,4-dichlorophenyl)-1,1-dimethylurea (DCMU at 10 µM final, from 10 mM ethanolic stock) were added to obtain maximal P700 oxidation during the saturating pulse. However, the kinetics shown in the paper contained light-adapted cells (at least 30-min with 630-nm LEDs emitting 4700 µmol photons m -2 s -1 ) which had a functional PSII. The continuous actinic red light was hatched by short dark intervals (250 µs) during which 10-µs detecting pulses were placed after 200 µs. The P700 redox kinetics was recorded during saturating pulse (12-ms of 630-nm LED, 3000 µmol photons m -2 s -1 ) and followed for several seconds darkness. The P700 kinetics in the dark were corrected for a linear drift that had developed in light-adapted cells especially in anoxic cells (supplementary Figure S10). ECS signals were measured as the difference of the absorbance changes at 520 and 546 nm (respective interference filters FWHM: 10 nm) using white pulsed LED probing light. The light-detecting diodes were protected from scattered actinic light by 3-mm BG39 filters (Schott, Mainz, Germany). A saturating pulse (22-ms of 630-nm LED, 3000 µmol photons m -2 s -1 ) was used for kinetics of ECS and the cytochrome b6f redox reactions. Using appropriate interference filters (FWHM: 10 nm), the latter were monitored on the level of cytochrome f (554 nm) and cytochrome b (563 nm) with a baseline drawn between 546 and 573 nm (69). At variance (69), 554-nm signals were corrected with 0.23 × (563 nm -546 nm) to subtract the contribution of cytochrome b to cytochrome f kinetics (6), especially in reducing conditions where this correction yielded pre-flash cytochrome f signals after several tens of ms (supplementary Figure S11). In single-turnover kinetics, the b-phase (electrogenic b6f contribution to the ECS signal) was deconvoluted by subtracting the c-phase (ATP synthase activity that results in ECS decay) which followed first-order exponential decay (46,47). All I/I signals in this study were calibrated to ECS changes produced by one charge separation upon a saturating laser flash in the presence of PSII inhibitors HA and DCMU. The PSI:PSII ratios were obtained by referring to uninhibited flash ECS amplitudes which were produced upon two charge separations by both photosystems, as reviewed recently (43). The overall membrane potential formation in saturating light was determined via a Dark Interval Relaxation Kinetics-based protocol (44), using ECS signal slopes at the end of the 22-ms light pulse and in subsequent darkness. Dark signals were disregarded during the first 2-ms, since they might be prone electrogenic PSI charge recombinations (38), and slopes up to 16ms were calculated with the linear fitting function in OriginPro software (OriginLab). The same fitting was used for ECS slopes during the first 2-ms upon changing the light intensity. To normalise to one charge separation, the ECS slope values in uninhibited samples were divided by the flash-induced ECS I/I in presence of HA/DCMU (charge separations/photosynthetic chain/s). Unless otherwise stated for calculated decay rates, kinetics were fit with mono-exponential decay function "ExpDec1" in OriginPro software and the fast P700 component, coined k1P-red in the text, was obtained via two-exponential fitting "ExpDec2" (dark kinetics from 4-ms to 2550-ms). For kf-red and kb-ox the 600-ms phase was used for ExpDec1 fitting. However, a shorter time window for anoxic kb-ox calculation was necessary for a seamless transition to the ExpDec1 fit of kb-red up to 9-s dark. hemes were comparable during the pulse (inset) and most of the oxidation was finished in ~50-ms, ~300-ms, ~25-ms in oxic, oxic +MV and anoxic samples, respectively. The latter samples showed a slow re-reduction phase with an onset at less than 100-ms dark. (E) Chlorophyll fluorescence-derived quantum yield of PSII (PSII) and PSII efficiency factor (qP) are shown (N = 3 ± SD). The pre-steady state after 10-s illumination and steady state cells after 30-min light adaptation were similar and MV addition significantly lowered qP (Student's t-test *P < 0.05) which, like PSII, was also lower in anoxic cells. (F) Following a saturating-pulse in light-adapted cells, redox relaxation of QAin the dark are shown (N = 3 ± SD). The fully reduced QApool was re-oxidised in oxic samples by ~30% in the first ~50-ms and a following ~500-ms retardation phase was missing upon MV addition. QAre-oxidation was slowed down in anoxic cells. Figure 2. Total electron transfer from PSII to PSI is changed in pgr5, most prominently in anoxic conditions. (A) Saturating pulse-induced P700 redox changes in absence and presence of 10 mM methyl viologen (MV) are shown, as well as in anoxic cells. The 12-ms pulse (hatched red box) was applied on light-adapted cells in the steady state (red box), followed by several seconds dark measurements (black box). (B) The different P700 populations were deconvoluted as oxidisable fraction (YI, yield of PSI), non-oxidisable P700 owing to acceptor side limitation (YNA), and pre-oxidised fraction due to donor side limitation (YND). The electron acceptor MV abolished YNA and increased YND despite PSII activity. Unlike WT (cf. Figure 1), anoxic pgr5 did not produce lower YI and maintained a low YNA fraction. YND was slightly increased (C) Cytochrome f redox changes in a similar light/pulse/dark regime show different pulse-induced oxidation magnitudes (inset) which, with oxic > anoxic = oxic+MV amplitudes, compared to the steady state reference. Oxidation amplitudes in oxic controls were smaller than in WT. Mutant kinetics in the presence of MV and in anoxia were similarly slowed down. (D) Redox changes of b-hemes were comparable during the pulse (inset) and showed a similar trend as WT with a longer oxidation phase in the presence of MV and slightly shorter phase in anoxia. The latter samples showed a re-reduction phase which started later and was slower than in anoxic WT. (E) Chlorophyll fluorescence-derived quantum yield of PSII (PSII) and PSII efficiency factor (qP) are shown (N = 3 ± SD). The pre-steady state after 10-s illumination and steady state cells after 30-min light adaptation differed in oxic conditions compared to WT (Student's t-test **P < 0.005, cf. Figure 1E). This effect was inhibited by MV which yielded as low PSII and qP pre-steady state samples without MV. Both parameters were also lower in anoxic cells. (F) Following a saturatingpulse in light-adapted cells, redox relaxation of QAin the dark are shown (N = 3 ± SD). The fully reduced QApool was re-oxidised in oxic samples by ~30% in the first ~50-ms and, unlike as oxic WT, the ~500-ms retardation phase was missing in pgr5 even without MV addition. QAre-oxidation was slowed down in anoxic cells. Figure 3. Depending on the conditions and strains, different rates were observed for the redox reactions in P700 and b6f during the saturating pulse and in the dark. Data from Figure 1, Figure 2, and Figure S4 were used for calculations (N = 3 ± SD; Student's t-test *P < 0.05 and **P < 0.005). As mentioned in the methods, exponential decay functions were used for fitting the data. (A) The pulse-induced P700 oxidation rate, kP-ox, was faster in oxic pgr5. Only MV samples in the WT showed faster kP-ox. Anoxic cells also produced faster kP-ox than oxic controls. (B) The faster of the two P700 re-reduction phases after the pulse yielded k1Pred. This rate was slowed down in MV samples. The anoxic WT and C1 also showed slower k1P-red. (C) Following the pulse, fitting the cytochrome f reduction phase yielded kf-red which was slowed down by MV. Anoxic WT showed faster kf-red, whereas it was slower to a different degree in anoxic pgr5 and C1. (D) Following the pulse, b-heme oxidation rates were calculated as kb-ox which was slowed down by MV in WT and C1. In anoxic WT kb-ox was faster compared to the oxic control and the other anoxic strains. (E) The slow b-heme rereduction rate kb-red in anoxia was fastest in WT. Figure 4. The electrogenic capacity of the photosynthetic electron transfer chain in pgr5 is compromised under anoxic conditions. The ability to generate and dissipate an electric field is shown by measuring the electrochromic shift, ECS, and signals of samples from Figure 1, Figure 2, and Figure S4 were recorded in steady state light, during a saturating pulse and in darkness. The ECS kinetics for the oxic WT (A) and pgr5 (B) indicate that the 22-ms pulse led to equilibration of a new membrane potential  (for ECS kinetics of C1 refer to supplementary Figure S5). The efficiency to generate a higher  level at the onset of the pulse is shown by initial rates in green (inset), yielding kini from the linear slope during the first 2-ms of the pulse. The  generation capacity at the end of the pulse (kend) was corrected with the -consuming activity of the ATP synthase. (C) The kini was similar in oxic strains and slower upon MV addition. A significant MV effect was not observed in pgr5 which had the lowest kini among the controls. In anoxic cells, kini was highest in WT, followed by C1 and pgr5. (D) At the end of the pulse, kend was lower in the strains due to exhaustion of electron carriers. There was a difference between oxic pgr5 and C1. Anoxic kend were the same in the strains. (N = 3 ± SD; Student's t-test *P < 0.05 and **P < 0.005) Figure 5. Assessments of steady state P700 redox state (A), cytochrome b6f dark redox relaxation (B, C), and electrogenic capacity of the photosynthetic chain (D) in anoxic WT (brown), pgr5 (violet), and PGR5-complemented lines (red, pgr5::PGR5) show complementation of the pgr5 phenotype in the rescued line. Cells were grown in Tris-acetatephosphate medium and PGR5 expression was induced by 5 µg/mL zeocin 24 h before experiments. (A) The different P700 populations were deconvoluted as oxidisable fraction (YI, yield of PSI), non-oxidisable P700 owing to acceptor side limitation (YNA), and preoxidised fraction due to donor side limitation (YND). Like in Figure 2B, the anoxic pgr5 maintained higher YI and YND, and failed to develop pronounced YNA like WT and pgr5::PGR5. (B) After the pulse, the reduction rates of cytochrome f (kf-red) were significantly slower in the mutant which recovered to WT-level in pgr5::PGR5. For kinetics refer to Figure S6A. (C) Simultaneously, the oxidation rates of b-hemes after the pulse (kb-ox) were significantly slowed down in the mutant only. Refer to Figure S6B for kinetics. (D) When steady state cells experienced the saturating pulse, the high initial electrogenic charge separation rates (kini) in WT were PGR5-dependent (N = 3 ± SD; Student's t-test *P < 0.05). Figure 6. Redox kinetics and electrogenic signals reveal a PGR5-dependent low-potential chain tuning in anoxia as well as an inhibitory effect of MV on the single b6f turnover. (A) Cytochrome f and b-heme signals are shown for oxic WT and pgr5. Cytochrome f is rapidly (<300 µs) oxidised after the flash and re-reduced within 100-ms darkness (fitted curves). The b-heme net reduction lasted between ~1-ms and 10-ms darkness, followed by a slower oxidation phase (fitted curves) for several tens of ms. (B) The corresponding oxic ECS kinetics were normalised to the signal produced by a flash hitting ~40% of PSI centres (<300 µs, a-phase). The following b-phase (fitted curve), resulting from b6f-dependent charge separation activity in the ~10-ms range, was deconvoluted from raw ECS kinetics by subtracting exponential decay phase produced by ATP synthase (c-phase, fitted curve, refer also to Figure S7). In this sample, the b-phase developed slower in pgr5 and showed a slightly larger relative amplitude. (C) Addition of MV slowed down cytochrome f reduction and b-heme oxidation in both strains, and the mutant was slightly less affected. The inhibition of b-heme oxidation allowed larger reduction amplitudes compared to panel A. (D) Evolution of the b-phase was slowed down by MV but the amplitude was not altered. MV also slowed down the c-phase upon -disulfide promotion (70). (E) The b6f redox kinetics in anoxia showed slightly slower cytochrome f oxidation whereas reduction was like in oxic cells. Net reduction of b-hemes in the first 10-ms was of negligible amplitude and a large oxidation phase followed. (F) ECS and b-phase kinetics in anoxia resembled oxic panel B. (G) Cytochrome f reduction rates kf-red were calculated from the fitted decay functions and showed significant slowdown in the presence of MV. (H) The electrogenic b-phase also evolved at slower rates (k) in MV samples. Compared to WT, the pgr5 mutant showed slower k in oxic and anoxic conditions. (I) After the apparent b-heme reduction phase ceased, the slow oxidation rates were expressed as kb-ox. MV slowed down kb-ox and the WT showed faster kb-ox in anoxia whereas the anoxic mutant kb-ox remained unchanged. (J) The relative b-phase amplitudes, compared to the a-phase, were comparable and only larger in pgr5 MV samples. The mutant had a tendency, though, to produce larger b-phase amplitudes. (N = 3 ± SD; Student's t-test *P < 0.05, **P < 0.005 and ***P < 0.0005) Figure 7. Model summarizing the multiple-turnover measurements under anoxic conditions. Except for PC and NADPH2, the redox levels were measured (blue and red stand for oxidized and reduced, respectively). PSI acceptor side limitation (YNA) and photosynthetic control (pH) were established between weakly (transparent) and strongly contributing levels (bold). Modified forward reaction efficiencies are highlighted (Qo: PQH2 oxidation, Rieske ISP/cyt.f interaction; Qi: PQ reduction, bh/ci electron sharing, Fd-dependent ci reduction; NADPH2 formation). Refer to Figures 8 and S1 for details on bh/ci interaction and stromal electron utilization. (A) When anoxic WT b6f operates in the Fd-assisted Q cycle mode, b6f electrogenicity is maintained and pH protects PSI. (B) Inefficient utilisation of excess stromal electrons at the Qi-site stalls the pgr5 b6f in anoxic conditions, thus weakening YNA. The effect could be linked to aggravated, incomplete FNR tethering (20) and a disturbed lowpotential chain oxidation influences ISP/cyt.f interaction (12). Refer to Figure S8C for a similar scenario in the presence of MV. . The dark-equilibrated pre-flash bl ox /bh red re-equilibrated again to bl ox /bh red after the Qo-site turnover and electrogenic charge transfer (unaltered b-heme net reduction). During this initial phase, one PQH2 is formed at the Qi-site by making use of the pre-reduced bh red /ci red redox couple. Although not indicated here, this reaction is slightly slower in pgr5 when taking slower b6f electrogenicity as a proxy (cf. Figure 6H). (A) The following b-heme oxidation phase is slower in the mutant and may be attributed to poor Qi-site substrate availability which is promoted upon ci red formation (7). In presence of a membrane potential (), the bh red /ci ox redox couple equilibrates (2) and a relatively slow transition to bh ox /ci red follows the  decay in pgr5 (grey and violet curve). (B) Unlike the mutant, WT retains sufficiently more FNR to the thylakoid membrane (20) which might be an allosteric modulator by interacting with the b6f (23-25). Thus, Fd bound to FNR might drive the transient generation of ci red in the presence of , which then creates the 'PQ-accessible' Qi-site specifically in WT and, due to the bh red /ci red redox couple, produces faster b-heme oxidation. Compared to pgr5, the faster b-heme oxidation also accelerates the Rieske FeS domain to swap closer to cytochrome f (reviewed in 40), which could be important during multiple turnover operation as indicated in Figure 7.