The role of BST4 in the pyrenoid of Chlamydomonas reinhardtii

In many eukaryotic algae, CO2 fixation by Rubisco is enhanced by a CO2-concentrating mechanism, which utilizes a Rubisco-rich organelle called the pyrenoid. The pyrenoid is traversed by a network of thylakoid-membranes called pyrenoid tubules, proposed to deliver CO2. In the model alga Chlamydomonas reinhardtii (Chlamydomonas), the pyrenoid tubules have been proposed to be tethered to the Rubisco matrix by a bestrophin-like transmembrane protein, BST4. Here, we show that BST4 forms a complex that localizes to the pyrenoid tubules. A Chlamydomonas mutant impaired in the accumulation of BST4 (bst4) formed normal pyrenoid tubules and heterologous expression of BST4 in Arabidopsis thaliana did not lead to the incorporation of thylakoids into a reconstituted Rubisco condensate. Chlamydomonas bst4 mutant did not show impaired growth at air level CO2. By quantifying the non-photochemical quenching (NPQ) of chlorophyll fluorescence, we show that bst4 displays a transiently lower thylakoid lumenal pH during dark to light transition compared to control strains. When acclimated to high light, bst4 had sustained higher NPQ and elevated levels of light-induced H2O2 production. We conclude that BST4 is not a tethering protein, but rather is an ion channel involved in lumenal pH regulation possibly by mediating bicarbonate transport across the pyrenoid tubules.


INTRODUCTION 70
Maintaining improvement in crop yields to keep pace with the rising demands for food 71  The protein BST4 (bestrophin-like protein 4, also known as Rubisco binding 95 membrane protein 1, RBMP1; Cre06.g261750) localizes exclusively to the pyrenoid 96 tubules and has been proposed to function as a tether protein, linking the Rubisco 97 matrix to the tubules (Meyer et al., 2020). BST4 is a predicted transmembrane protein 98 domain and N-terminal (Fig. 1D), both of which found BST4 to resolve in a distinct 130 clade from BST1-3 but within the wider green algae group.

143
As well as being distinct at a sequence level, BST4 also localizes differently from 144 BST1-3. While BST1-3 localize throughout the thylakoid membrane and are enriched 145 at the pyrenoid periphery (Mukherjee et al., 2019), BST4 localizes to the centre of the 146 pyrenoid in a pattern that resembles the pyrenoid thylakoid tubule system ( Fig. 2A was depleted (Fig. 2C, D), indicating that BST4 is located in the tubules and not the 151 Rubisco-enriched pyrenoid matrix. Nevertheless, previous work has shown that the C-152 terminal RBMs of BST4 enable the protein to interact with CrRBCS1 (Meyer et al. 153 2020). We confirmed using a yeast-2-hybrid approach that the C-terminus of BST4 154 interacts with CrRBCS1 (Fig. S2A). We also measured the efficiency of Förster 155 resonance energy transfer (FRET) from Venus to mScarlet-I, and found that the FRET 156 efficiency was ~35%, supporting the proximity of CrBST4 to CrRBCS1 (Fig. S2B). 2022). To investigate if this is the case for BST4, we initially used an AlphaFold-159 multimer modelling approach. When five chains of BST4 were inputted to AlphaFold 160 a pentameric structure was predicted (Fig. 2E). To test the complex assembly of BST4 161 in vivo, we utilized a Slimfield microscopy molecular tracking method (Plank et al., 162 2009). We first developed a Chlamydomonas line that only expressed fluorescently 163 tagged BST4 by expressing BST4-mScarlet-I in a bst4 mutant background (Fig. S3). 164 We then tracked individual fluorescent molecules of BST4-mScarlet-I and quantified 165 the number of BST4 monomers per complex. The resulting probability distribution 166 revealed that the most common BST4 complex is made up of five molecules (Fig. 2F). 167 Other peaks showed complexes with numbers of molecules divisible by five, which 168 may be multiple pentameric channels grouping together. The interval between the 169 probability peaks was also five (Fig. 2G). To further support higher order complex 170 assembly of BST4 we ran purified Chlamydomonas thylakoid membranes on a Blue 171 Native-PAGE gel (BN-PAGE) and immunoblotted for BST4 (Fig. 2H). BST4 formed a 172 smear at ~1000 kDa, which is considerably larger than a pentamer (~330kDa). This 173 could be due to higher order assemblies of BST4, pentameric BST4 in a complex with 174 other proteins and/or aberrant migration during BN-PAGE due to influences by 175 complex shape. Collectively, our data support in vivo higher order assembly of BST4 176 potentially as a pentamer. 177

BST4 localizes to the stroma lamellae thylakoid membrane in Arabidopsis 191
We used Arabidopsis as S2Cr line was therefore used as a platform to test whether BST4 acts as a tether 199

protein. 200
We initially localized a BST4-mNeon fusion protein following stable expression in S2Cr 201 (Fig. 3). BST4-mNeon was observed in chloroplasts, demonstrating that the native 202 chloroplast signal peptide was compatible with the chloroplast targeting mechanism in 203 land plants (Fig. 3A), as seen previously for other Chlamydomonas proteins (Atkinson and not in the stromal fraction (Fig. 3D). The thylakoids were then further fractionated 212 into grana stacks and stroma lamellae sub-fractions. BST4 was found in the stroma 213 lamellae fraction (Fig. 3E)

222
Immunoblots of sub-chloroplast fractions isolated from Arabidopsis line S2Cr expressing BST4. RbcL and CP43 223 were probed for as stromal and thylakoid controls, respectively. E. Immunoblots of fractionation thylakoids from 224 Arabidopsis line S2Cr expressing BST4. CP43 and PsaB were used for grana stack and stroma lamellae controls, 225 respectively F. Trypsin protease protection assay. Intact thylakoids containing BST4 subjected to 0 or 100 μg/ml 226 trypsin with or without the addition of 1% (v/v) Triton. AtpB and PsbO used as controls for stromal facing (exposed) 227 and lumen facing (protected), respectively. G. Immunoblot of proteins from thylakoids separated by Blue Native-

232
As BST4 has two RBMs on its C-terminus, we hypothesized that BST4 should be 233 orientated with the C-terminus facing the stroma so that the RBMs are available to 234 interact with CrRBCS. To determine the orientation of BST4, we performed a protease 235 protection assay on thylakoids isolated from the untagged BST4 transgenic plants. 236 Our antibody was raised against the C-terminal end of BST4, so it could be used to 237 assess whether the C-terminus was exposed for degradation in the stroma or if it was 238 protected in the lumen. We found that the BST4 C-terminus was fully degraded after 239 a 60 min treatment of trypsin, indicating that it faced the stroma (Fig. 3F). There was 240 some degradation of the lumenal control, PsbO, which we attribute to a portion of the 241 thylakoid membrane preparation not being fully intact. However, PsbO was fully 242 degraded when the membranes were solubilized, indicating that they were sufficiently 243 intact to differentiate between lumenal-and stromal-facing peptides. Therefore, BST4 244 was observed in the expected location and orientation in plant thylakoid membranes. 245 Finally, we also tested whether BST4 forms a complex in the Arabidopsis thylakoid 246 membrane. We subjected thylakoids from the untagged BST4 transgenic plants to  PAGE and detected a single band of ~850 kDa (Fig. 3G). Thus, BST4 forms a similar 248 order complex in Arabidopsis to that in Chlamydomonas but may be lacking additional 249 interaction partners present in Chlamydomonas.

294
Although BST4 partially co-localized with the Rubisco condensate, we found no 295 evidence to suggest that that BST4 could facilitate the inclusion of thylakoid 296 membranes. Confocal microscopy was not sufficient to determine if chlorophyll 297 autofluorescence was in the proto-pyrenoids of S2Cr lines expressing BST4-mCherry 298 and EPYC1-tGFP. However, transmission electron microscopy (TEM) revealed no 299 visible indication of thylakoid membranes in the condensates, which were structurally 300 similar to condensates in S2Cr-EPYC1 lines lacking BST4 (Fig. 4F)

BST4 is not necessary for tubule formation in Chlamydomonas 304
To investigate whether BST4 was necessary for normal formation of the thylakoid 305 tubule-structure in Chlamydomonas we compared the structure of pyrenoids of WT 306 and a BST4 knock-out line (bst4) (Fig. S3). TEM images showed that pyrenoids from 307 bst4 were structurally comparable to the WT control, including the presence of 308 pyrenoid tubules (Fig. 4G). There were also no differences in pyrenoid size or shape 309 between the two lines when comparing pyrenoids from 40-50 cells from each genotype 310 (Fig. S4). As a result, we conclude that BST4 is not necessary for the pyrenoid tubule-311 Rubisco matrix interface in Chlamydomonas. We speculate that BST4 may require a 312 pre-existing pyrenoid tubule network to be localized in the pyrenoid rather than driving 313 the inclusion of thylakoid membranes into the Rubisco matrix, or is redundant as a 314 tether protein. RBMs in BST4 localization in Chlamydomonas, we generated a truncated version of 321 BST4 (residues 1 to 386) that lacked the C-terminus containing the two RBMs 322 (BST4ΔC-term) and compared localisation with full-length BST4 expressed in WT and 323 the bst4 mutant ( Fig. 5 and Fig. S5). BST4ΔC-term-mScarlet expressed in the bst4 324 mutant line did not localize to the pyrenoid tubules but was found throughout the 325 thylakoid membrane (Fig. 5). This is consistent with our findings in Arabidopsis with 326 BST4WR/EE (Fig. 4C), demonstrating that localization of BST4 is driven via RBM-327 Rubisco interaction. Thus, the C-terminus is necessary for BST4 localisation to the 328 pyrenoid tubules in the presence of the matrix. 329 When BST4ΔC-term-Venus was expressed in WT Chlamydomonas (i.e., that still 330 produced non-truncated BST4), we observed fluorescence throughout the thylakoids, 331 but with the majority of the signal still localized to the pyrenoid tubules (Fig. S5). We 332 conclude that BST4ΔC-term-Venus is recruited to the pyrenoid through an interaction 333 with the native full-length BST4, which is further evidence that BST4 oligomerizes.

337
We next investigated whether BST4 localizes to the pyrenoid tubules, or whether 338 BST4 localizes to the tubules through an interaction with Rubisco. To do so, we utilized 339

Chlamydomonas KO mutant bst4 does not have a defective CCM but is under 356 oxidative stress 357
To test whether BST4 has a role in the operation of the CCM, we measured the growth 358 of bst4 compared to WT under various CO2 conditions (Fig. 6). Semi-quantitative spot 359 assays indicated no reduction in growth under CO2-limiting conditions (Fig. S7). The 360 liquid growth assay also demonstrated that bst4 grew well in both CO2 replete and 361 limiting conditions, 3% and 0.04%, respectively (Fig. 6A). The bst4 mutant had a 362 slightly increased cell density compared to WT (Fig. 6A) in the 0.04% CO2 treatment. 363 However, when comparing the calculated specific growth rates (μ h -1 ) for both the 364 exponential growth phase (days 0-3, bst4 0.0402 ± 0.0003 μ h -1 and WT 0.0389 ± 365 0.0007 μ h -1 ) or the full growth assay (days 0-5, bst4 0.0257 ± 0.0001 μ h -1 and WT 366 0.0241 ± 0.001 μ h -1 ), there was no statistically significant increase in specific growth 367 rates between bst4 and WT (two-tailed t-test, p = 0.17 and 0.20, respectively, n = 3, 368 full results Table S1). In combination, we concluded that BST4 is not an essential 369 component of the CCM and may therefore have a redundant or regulatory role. We 370 also included the complemented bst4::BST4 and bst4::BST4ΔC-term-mScarlet-I 371 (hereafter bst4::BST4ΔC-term) lines in the spot and liquid growth assays. While all lines 372 grew well in the spot assay, in the liquid growth bst4::BST4ΔC-term grew comparably to 373 WT and bst4 whereas bst4::BST4 exhibited a slightly reduced growth than the other 374 lines in both CO2 conditions. 375  diffuse cells on the periphery of the colony. We used a range of CO2 and light 387 conditions to investigate the diffuse colony phenotype (Fig. 6B) and found it was most 388 apparent under high light (300 μmol photons m −2 s −1 ) and low or very low CO2 389 conditions (0.04% CO2, 0.01% CO2, respectively). WT, bst4::BST4 and bst4::BST4-390 mScarlet-I complemented lines had little or no diffusivity, although WT did display a 391 slightly diffuse colony phenotype at 300 μmol photons m −2 s −1 . Interestingly, BST4ΔC-392 term-mScarlet-I was unable to rescue the diffuse colony phenotype suggesting that 393 either the presence of the C-terminus or localization to the tubules is essential for the 394 function of BST4. indicating that bst4 cells may be experiencing oxidative stress. 407 To confirm that the phototactic response of bst4 was due to oxidative stress, we 408 recorded the direction of phototaxis in cells exposed to either ROS or a ROS quencher. to WT under low CO2 conditions, we hypothesized that CO2 fixation is not limiting in 418 this case. Rather, the absence of BST4 may disrupt regulation of the light reactions. 419

BST4 is associated with regulation of the lumenal pH in Chlamydomonas. 420
To assess the impact of BST4 on the light reactions, we measured photosynthetic 421 parameters of WT, bst4 and bst4 complemented with either full length BST4 or 422 BST4ΔC-term. Chlorophyll and carotenoid levels were slightly lower in bst4, however this 423 was not significantly different. Complementation with BST4 recovered chlorophyll and 424 carotenoid levels to a higher level than WT whilst in bst4::BST4ΔC-term lines levels were 425 comparable to the bst4 mutant (Fig. S9A). 426 Next, we measured the slow kinetics of chlorophyll fluorescence upon transition from 427 dark to high light to determine the impact of BST4 on the distribution of absorbed light 428 between non-photochemical quenching (NPQ) and electron transport through 429 photosystem II (Y(II)) ( Fig. 7). Fv/Fm was slightly higher in bst4, bst4::BST4 and 430 bst4::BST4ΔC-term compared to WT (Fig. 7A). NPQ was induced faster and reached a 431 significantly higher steady state in bst4 than in WT (Fig. 7B) were observed during the light phase (Fig. 7C). However, in the subsequent dark 435 phase, Y(II) in bst4 recovered faster in line with the higher Fv/Fm (Fig. 7C). 436 The increase in NPQ suggested that BST4 influences the reactions occurring in the 437 thylakoid membrane. A higher steady-state NPQ indicates that the bst4 cells 438 experience high light stress and respond by accelerating photoprotective excess light 439 dissipation to maintain optimal electron transport. This observation aligns with 440 increased ROS production observed in bst4 (Fig. 6C), which would result in elevated 441 NPQ to compensate for the increase in oxidative stress. In the case of bst4::BST4ΔC-442 term, it is possible that the truncation directly affects BST4 activity. Alternatively, the 443 mis-localization of BST4 activity throughout the thylakoid membranes, instead of solely 444 localized to the thylakoid tubules, results in the disruption of optimal electron transport 445 and the need for increased photoprotection. 446 NPQ is induced in response to a drop in lumenal pH, a process that is mediated by 447 the quencher protein LHCSR3. Thylakoid ion channels can influence NPQ through 448 altering the partitioning of the proton motive force (PMF) into pH gradient (ΔpH) and 449 electrical gradient (ΔѰ). To investigate whether BST4 was altering lumenal pH we 450 measured the slow decay kinetics of electrochromic shift (ECS), which allows the 451 determination of PMF partitioning. WT and bst4::BST4 displayed similar standard 452 kinetics. However, both bst4 and bst4::BST4ΔC-term had abnormal kinetics and did not 453 reach steady state (Fig. S8A, B). Therefore, we could not derive PMF partitioning for 454 all the genotypes. The abnormal ECS kinetics in bst4 and bst4::BST4ΔC-term may 455 indicate altered proton-related processes, such as over acidification of the thylakoid 456 lumen. We also measured the accumulation of protein LHCSR3 after exposing cells 457 to 150 μmol m -2 s -1 for 3 h and found that bst4 accumulated more LHCSR3 as

469
and C. Photosystem II quantum yield (Y(II)). D to F. ECS decay kinetics were performed on cells pre-exposed for 470 10 min to high light and the D. total PMF, E. gH + , and F. total H + flux (vH + ) were determined as described in Methods.

471
Data are the means ± SEM (n=3 replicates). Different letters indicate statistically significant difference among the 472 genotypes (one-way ANOVA test, followed by Turkey's post hoc test, P < 0.05). Statistical analysis was performed 473 separately for measurements made with and without HCO3 -(black and green letters, respectively). G. Immunoblot  To further explore the proton related processes, we recorded fast decay ECS kinetics 479 which allows for the determination of the total PMF size, as well as H + conductivity of 480 ATP synthase (gH + ) (Fig. S9E). The PMF after 10 min of illumination did not 481 significantly differ among any of the genotypes (Fig. 7D). At the same timepoint, the 482 gH + was significantly higher in bst4 and bst4::BST4ΔC-term than in WT and bst4::BST4 483 (Fig. 7E). We also calculated the total proton flux (νH + ), which showed a similar trend 484 although νH + was also significantly lower in bst4::BST4, compared to WT (Fig. 7F). 485 Based on the proton flux data, we postulate that more ATP is produced in bst4 and (2022) proposed a model whereby RuBP acts as proton carrier to increase H + 500 concentration in the pyrenoid tubules. It is possible that BST4 is the channel that 501 facilitates RuBP translocation in the tubules in this model. To test this hypothesis, we 502 used small molecule analogues K-PEP and K-Gluconate but no currents were 503 detected for these either (Fig. S9A). Therefore, we were unable to draw conclusions 504 as to what BST4 is permeable to. BST4 may require certain conditions to be open that 505 are not met in the oocyte system, such as post-translational modifications, a specific 506 pH or an interaction partner. BST4 was found to be phosphorylated and had an 507

BST4 has no impact on growth in Arabidopsis S2Cr line 512
Although the permeability is undefined, if BST4 is a functional anion channel, it is 513 possible that its presence in plant thylakoid membranes might have an effect. We 514 generated three independent BST4 no tag lines in the S2Cr background and used them 515 to assess the impact of BST4 on Arabidopsis physiology (Fig. S10A). We found there 516 was no difference in growth between plant expressing BST4 and their azygous 517 segregants, as determined by the rosette area (Fig. S10B). We also found that BST4 518 expressing plants tended to have slightly lower Fv/Fm as compared to azygous 519 segregants and the parent line, although this was not significant (Fig. S11C). Further 520 measurements were made for lines 2 and 3 on the kinetics of NPQ, Y(II) and for PMF 521 size and partitioning but no consistent differences were observed (Fig. S11). 522

Proposed model for the role of BST4 in the Chlamydomonas pyrenoid 523
Based on these findings, we speculate that BST4 may have a regulatory role in the 524 CCM through its function as a bestrophin channel. We propose two possible models 525 to explain our data. The first model is that BST4 is functioning as an anion channel to 526 import anions from the pyrenoid into the thylakoid lumen. This would counterbalance 527 the charge of protons in the tubules, possibly contributing to pH regulation or 528 maintaining a high flux of protons towards the pyrenoid to power the CCM (Figure 8). 529 The absence of BST4 would cause disruption of lumen pH or cause a drop in CCM 530 efficiency that causes stress under high light and low CO2. The second model is that 531 BST4 is part of a HCO3recovery system to takes up HCO3from within the pyrenoid, 532 allowing HCO3to be transported back into the tubules where it can be dehydrated to 533 CO2 and thus available to be fixed by Rubisco. In this scenario, the absence of BST4 534 would lead to a slight reduction in CCM operation efficiency, perhaps not sufficient to 535 cause a growth penalty but enough to cause a build-up of protons in the tubule lumen 536 and alter the homeostasis of the light reactions. Further work is required to elucidate 537 the function of BST4, primarily, determining what species BST4 is permeable to, any 538 modes of channel regulation and possible interaction partners. 539 knock-out line there is increased ROS production. There is also likely an over acidified thylakoid lumen, which 542 causes elevated NPQ and increased ATP generated through ATP Synthase activity. We suggest that BST4 543 functions as a thylakoid membrane channel, which plays a role in balancing the light reactions with carbon fixation 544 while the CCM is active. In this role, we propose BST4 is a channel that facilitates the transport of anions from the 545 pyrenoid matrix into the thylakoid lumen whereby they act as a sink for available protons (Model 1). It is possible 546 that BST4 is specifically a HCO3channel (Model 2) that facilitates the recovery of HCO3from the pyrenoid matrix.

547
This would subsequently enable CAH3 to catalyse the dehydration of HCO3into CO2, which is then able to diffuse 548 back into the pyrenoid and be fixed by Rubisco.

Phylogenetic Analysis 551
Sequences for phylogenetic analysis were compiled by blasting of BST4 552 terminal (sequence begins R35), any subsequent mutations were made via PCR. 582

Arabidopsis transformation 583
Arabidopsis was transformed by floral dip as previously described in (Atkinson et al., 584 2016). BST4-mNeon primary transformants were screened for transgene insertion by 585 seed fluorescence from pFAST-R and BST4 expression was confirmed by checking 586 for mNeon fluorescence and by immunoblot. BST4 no tag primary transformants were 587 screened using kanamycin resistance and immunoblot. Zygosity was checked via 588 seed fluorescence from pFAST-R or kanamycin resistance. 589

Chlamydomonas cell culture conditions and strain details 590
Chlamydomonas reinhardtii cultures were maintained as previously described (Ma et 591 al., 2011). Tris-Acetate-Phosphate (TAP) and minimal (TP) media (acetate free) were 592 prepared according to Sueoka (1960). TAP and TP agar plates for growth were made ATGCTTCTCTGCATCCGTCT and reverse ATGTTTTACGTCCAGTCCGC). The bst4 600 knock-out was complemented with BST4 constructs described herein. All 601 complemented lines were validated by western blotting of BST4 and specified epitope 602 tags, described below (Fig. S3B-E). 603

Chlamydomonas transformation 604
For each Chlamydomonas transformation, 28 ng kbp −1 of plasmid was linearized by 605 restriction digest. Cells were grown to 2-4 x 10 6 cell mL

Phototaxis Assays 636
Chlamydomonas cells were grown heterotrophically in TAP media until they reached 637 2-4 x 10 6 cell mL -1 and harvested by centrifugation at 1000 xg for 10 min. Pelleted cells 638 were either re-suspended in TP media or, for ROS manipulation assays, a phototaxis respectively and 70% relative humidity. 651

Chlamydomonas Confocal Microscopy 652
Transgenic fluorescent strains were initially grown heterotrophically in TAP media until 653 reaching 2 -4 ×10 6 cells mL -1 and resuspended in TP media overnight prior to 654 imaging. Cells were mounted on 8-well chamber slides and overlayed with 1.5% low 655 melting point agarose made with TP-medium. Images were collected on a LSM880 656 (Zeiss) equipped with an Airyscan module using a 63× objective. Laser excitation and 657 Emission setting for each channel used are set as below: Venus (Excitation: 514 nm; 658 Emission 525 -500 nm); mScarlet-I (Excitation: 561 nm; Emission 570 -620 nm); 659 Chlorophyll (Excitation: 633 nm; Emission 670 -700 nm). The intervals between all peaks for each acquisition were aggregated across the 711 pyrenoid population, weighted by inverse square-root distance (thereby accounting for 712 shot-noise in broader intervals). A second distribution (Fig. 2G) was then generated 713 from this weighted population of intervals. The kernel width in this estimate was 0.7 714 molecules multiplied by the square root of the mean stoichiometry divided by the root 715 number of intervals (thereby accounting for shot-noise in intervals between peaks of 716 higher stoichiometry). The periodicity was then reported as the mode of this 717 distribution and its 95% confidence interval. 718

Arabidopsis Confocal microscopy 719
Small sections of 3-4-week-old leaf tissue (~5-10 mm 2 ) were adhered to slides using 720 double-sided tape with basal side up. A x40 water immersion objective lens was used. In order to quantify BST4 protein in Chlamydomonas lines, cells were grown in TP 747 media at ambient CO2 until reaching 2-4 x 10 6 cells mL -1 . Cells were harvested by 748 centrifugation at 1000 x g for 10 mins, normalized to Chl content and resuspended in 749 the extraction buffer described above. Samples were freeze-thawed three times and 750 spun at 20, 000 x g for 20 min at 4℃. Protein extractions containing 5 µg of Chl with 1 751 x SDS loading buffer were boiled at 100℃ for 5 min and loaded onto a 4-20% 752 polyacrylamide gel (Mini Protean TGX, Biorad Laboratories). Proteins were 753 transferred to a PVDF-FL membrane on a Bio-rad semidry blotting system. BST4 754 primary antibody was used as described above alongside alpha-tubulin primary 755 antibody raised in mouse (Agrisera), as a loading control. Anti-rabbit and anti-mouse respectively, according to Genty et al., (1989). 802

Electrochromic shift (ECS) in Chlamydomonas 803
ECS measurements in Chlamydomonas were carried out using a the Dual-PAM-100 804 equipped with a P515/535 module (Walz). Cells grown and prepared as for Chl 805 fluorescence experiments described previously were layered on a glass slide and 806 Samples were washed twice with dH2O and twice with 50 mM sodium cacodylate. 868 Fixed samples were dehydrated in an acetone series (25%, 50%, 75%, 90% and 869 100%) ~20 minutes each step. Dehydrated samples were infiltrated with Spurr's resin 870 by incubating in 25% then 50% Spurr resin in acetone for 30 minutes and transferred 871 to 75% for 45 minutes at room temperature. They were then incubated in 100% Spurr 872 resin overnight before polymerising at 70 °C for 24 hours. Sections ~70 nm thick were 873 collected on Copper grids and stained with saturated uranyl acetate and lead citrate. 874 Images were collected with a FEI Tecnai 12 BT at 120kV using a Ceta camera. 875

Electron microscopy of Arabidopsis 876
Leaves were cut into small 5 mm strips and fixed in 4% (v/v) Paraformaldehyde/0.5% oocytes was validated by western blotting (Fig S10B)    photosystem II quantum yield (Y(II)) were recorded during 10 min of illumination at 830 µmol photons m −2 s −1 followed by a 5-minute dark period. Data are presented as means ± SEM (n=4-6). Asterisks indicate statistical difference between plants expressing BST4 and their Azygous (Az) segregants according to unpaired t-test (P ≤ 0.05). D. Proton motive force (PMF) size and E. partitioning to pH gradient (∆pH) after 3 min illumination at 830 µmol photons m −2 s −1 . Data are means ± SEM (n=5-6). The letters in E and F indicate non-significant differences between plants expressing BST4 and their azygous segregants according to unpaired t-test (P > 0.05).