Cryo-EM structure of the spinach cytochrome b₆ f complex at 3.6 Å resolution

Abstract

The cytochrome b₆ f (cytb₆ f) complex has a central role in oxygenic photosynthesis, linking electron transfer between photosystems I and II and converting solar energy into a transmembrane proton gradient for ATP synthesis^1–3. Electron transfer within cytb₆ f occurs via the quinol (Q) cycle, which catalyses the oxidation of plastoquinol (PQH₂) and the reduction of both plastocyanin (PC) and plastoquinone (PQ) at two separate sites via electron bifurcation². In higher plants, cytb₆ f also acts as a redox-sensing hub, pivotal to the regulation of light harvesting and cyclic electron transfer that protect against metabolic and environmental stresses³. Here we present a 3.6 Å resolution cryo-electron microscopy (cryo-EM) structure of the dimeric cytb₆ f complex from spinach, which reveals the structural basis for operation of the Q cycle and its redox-sensing function. The complex contains up to three natively bound PQ molecules. The first, PQ1, is located in one cytb₆ f monomer near the PQ oxidation site (Q_p) adjacent to haem b_p and chlorophyll a. Two conformations of the chlorophyll a phytyl tail were resolved, one that prevents access to the Q_p site and another that permits it, supporting a gating function for the chlorophyll a involved in redox sensing. PQ2 straddles the intermonomer cavity, partially obstructing the PQ reduction site (Q_n) on the PQ1 side and committing the electron transfer network to turnover at the occupied Q_n site in the neighbouring monomer. A conformational switch involving the haem c_n propionate promotes two-electron, two-proton reduction at the Q_n site and avoids formation of the reactive intermediate semiquinone. The location of a tentatively assigned third PQ molecule is consistent with a transition between the Q_p and Q_n sites in opposite monomers during the Q cycle. The spinach cytb₆ f structure therefore provides new insights into how the complex fulfils its catalytic and regulatory roles in photosynthesis.

Photosynthesis sustains life on Earth by converting light into chemical energy in the form of ATP and NADPH, producing oxygen as a by-product. Two light-powered electron transfer reactions at photosystems I and II (PSI and PSII) are linked via the cytb₆ f complex to form the ‘Z-scheme’ of photosynthetic linear electron transfer (LET)¹. Cytb₆ f catalyses the rate-limiting step in LET, coupling the oxidation of PQH₂ and reduction of PC and PQ to the generation of a transmembrane proton gradient, which is used by ATP synthase to make ATP^2,3. The cytb₆ f complex is analogous to the cytbc₁ complex found in mitochondria⁴ and anoxygenic photosynthetic bacteria⁵, and both operate via the modified Q cycle^2,6. The cytb₆ f and cytbc₁ complexes are dimeric and have similarly arranged electron transfer cofactors, comprising a 2Fe-2S cluster, two b-type haems and a c-type haem. However, crystallographic structures of cyanobacterial and algal cytb₆ f complexes have revealed additional cofactors that are not found in cytbc₁ complexes, including chlorophyll a, 9-cis β-carotene and an additional c-type high-spin haem^7–9. The Q-cycle involves bifurcated transfer of two electrons, derived from oxidizing a lipophilic PQH₂ molecule at the Q_p-binding site, into the high- (2Fe-2S, cytf) and low- (cytb_p, b_n and c_n) redox potential pathways, whereas the two protons enter the thylakoid lumen^2,6. The high-potential pathway delivers an electron to a membrane-extrinsic soluble acceptor protein, PC, destined for PSI, while the low potential pathway delivers its electron to a PQ molecule bound at the Q_n site near the stromal side of the membrane. Oxidation of a second PQH₂ at the Q_p site culminates in the two-electron reduction of a Q_n site bound PQ, which together with two proton transfers from the stroma, regenerates PQH₂. The Q-cycle thereby doubles the number of protons transferred to the lumen per PQH₂ oxidized^2,6. Yet, full understanding of the Q-cycle mechanism is hindered by a lack of information on the binding of the substrate PQ/ PQH₂ molecules within the complex.

In addition to its role in LET, cytb₆ f also plays a key part as a redox sensing hub involved in the regulation of light harvesting and cyclic electron transfer (CET), which optimize photosynthesis in fluctuating light environments^10,11. Cytb₆ f communicates the redox status of the PQ pool to the loosely associated light harvesting complex II (LHCII) kinase, STN7^12–14. Phosphorylation of LHCII results in a decrease in thylakoid membrane stacking, promoting the exchange of LHCII between PSII and PSI to balance their relative excitation rates¹⁵ and regulate CET¹⁶. CET involves the reinjection of electrons from ferredoxin into the PQ pool, generating a proton gradient for photoprotective downregulation of PSI and PSII or to augment ATP synthesis, without net formation of NADPH¹¹. The cytb₆ f complex has been proposed to fulfil the role of the ferredoxin–PQ oxidoreductase (FQR) during CET, with the stromal-facing haem c_n suggested to channel electrons from ferredoxin NADP⁺ reductase (FNR) bound ferredoxin to the Q_n-site PQ¹⁷. How cytb₆ f performs these central redox-sensing regulatory roles remains unclear.

Genetic manipulation of photosynthetic regulation is now recognized as being key to increasing crop yields to feed a global population projected to approach 10 billion by 2050¹⁸. Indeed, overproduction of the Rieske iron–sulfur protein (ISP) of cytb₆ f in Arabidopsis thaliana led to a 51% increase in yield¹⁹. Further progress in understanding the regulatory roles of cytb₆ f and potentially manipulating them for crop improvement requires knowledge of the structure of the higher plant complex. Here, using a gentle purification procedure to obtain a highly active dimeric complex (Extended Data Fig. 1) and single-particle cryo-EM, we resolve the cytb₆ f complex from Spinacia oleracea (spinach) at 3.6 Å resolution (Extended Data Fig. 2, Extended Data Table 1).

The colour-coded map (Fig. 1a–c) shows the architecture of this dimeric complex surrounded by a disordered density comprising detergent and lipid molecules. The overall organization of this higher plant cytb₆ f complex is similar to crystallographic structures of algal and cyanobacterial complexes from Chlamydomonas reinhardtii⁷ (Protein Data Bank (PDB): 1Q90), Mastigocladus laminosus⁸ (PDB: 1VF5) and Nos-toc sp. PCC 7120⁹ (PDB: 2ZT9) (Extended Data Table 2). Each monomeric unit of the cytb₆ f complex comprises four large polypeptide subunits that contain redox co-factors (cytf, cytb₆, ISP and subunit IV), and four small subunits (PetG, PetL, PetM and PetN). Extended Data Figure 3 shows the density and structural model for each subunit. The extrinsic domains of cytf and the ISP on the lumenal face of the complex flank the membrane-integral cytb₆ subunits (Fig. 1a, b). The organization of the transmembrane integral subunits can be seen on the stromal side of the complex (Fig. 1c), with 13 transmembrane helices visible within each monomer (Fig. 1d–f). Peripheral to the core of cytb₆ (four transmembrane helices) and subunit IV (three transmembrane helices) on the long axis of the complex is the single kinked transmembrane helix of the ISP that crosses over to provide the soluble ISP domain of the neighbouring monomer. The single transmembrane helix belonging to cytf is sandwiched by the transmembrane helices of the four minor subunits PetG, PetL, PetM and PetN, which form a ‘picket fence’ at the edge of the complex.

Fig. 1. Cryo-EM structure of the cytb₆ f complex from spinach.

a–c, Views of the colour-coded cytb₆ f density map showing cytb₆ (green), cytf (magenta), ISP (yellow), subunit IV (cyan), PetG (grey), PetM (pink), PetN (pale orange) and PetL (pale purple). Detergent and other disordered molecules are shown in semi-transparent light grey. a, View in the plane of the membrane. The grey stripe indicates the probable position of the thylakoid membrane bilayer. b, View perpendicular to the membrane plane from the lumenal (p) side. c, View perpendicular to the membrane plane from the stromal (n) side. d–f, Modelled subunits of cytb₆ f shown in a cartoon representation and coloured as in a–c. d, View in the plane of the membrane. e, View perpendicular to the membrane plane from the lumenal side. f, View perpendicular to the membrane plane from the stromal side.

Figure 2a, b shows the organization of the prosthetic groups and lipids, with four c-type haems (f and c_n, dark blue), four b-type haems (b_p and b_n, red), two 2Fe-2S clusters (burnt orange and yellow), two 9-cis β-carotenes (orange), two chlorophyll a molecules (green), three PQ molecules (yellow) and twelve bound lipids (two monogalactosyl diacylglycerol, four phosphatidylglycerol, three sulfoquinovosyl diacylglycerol and three phosphatidylcholine, all shown in white). Extended Data Figure 4 shows the density map and structural model for each prosthetic group. Figure 2c shows all of the bound electron transfer cofactor edge-to-edge distances within the cytb₆ f complex. Electron transfer from the 2Fe-2S cluster is thought to involve movement of the lumenal ISP domain, pivoting between closer association with the Q_p site and the haem f. In comparison to the chicken cytbc₁ complex, in which the two conformations of the ISP were resolved²⁰, the ISP and bound 2Fe-2S cluster in the spinach cytb₆ f structure appear to be in the distal position with respect to haem f, as in the existing algal and cyanobacterial cytb₆ f structures (Extended Data Table 2). PQ locations are generally inferred from crystallographic structures containing tightly bound quinone analogue inhibitors^21–23. The cryo-EM structure was obtained with native PQ molecules (Fig. 2d), clearly defined by their respective densities (Extended Data Fig. 4); their distances from the nearest cofactors are shown in Fig. 2e–g. One PQ molecule (PQ1) is adjacent to the haem b_p and chlorophyll on one side of the dimer (Fig. 2e), and a second (PQ2) binds adjacent to the haem c_n–haem b_n pair on the opposite monomer to PQ1 (Fig. 2f). A third and less clearly defined PQ (PQ3) lies between the haem c_n of one monomer and the haem b_n of the other (Fig. 2g). The density map in this region could also be interpreted as a phospholipid; Extended Data Fig. 6 shows the two possible fits—to a plastoquinone or a lipid.

Fig. 2. The global arrangement of prosthetic groups, lipids and plastoquinone molecules in the spinach cytb₆ f complex.

a, b, The arrangement of molecules in the cytb₆ f complex viewed in the membrane plane (a) and perpendicular to the membrane plane from the stromal side (b). Chl, chlorophyll a; b_n, haem b_n; c_n, haem c_n; b_p, haem b_p; β-car, 9-cis β-carotene; FeS, 2Fe-2S. c, d, Cofactors and edge-to-edge distances (in Å) in the dimeric cytb₆ f complex. e, The location of the 1,4-benzoquinone ring of PQ1 adjacent to haem b_p, the 2Fe-2S centre and two conformations of the chlorophyll molecule, represented in two shades of green. f, Close-up of the 1,4-benzoquinone ring of PQ2 and the nearby haem c_n and haem b_n near the stromal face of the complex. g, The 1,4-benzoquinone ring of PQ3, which sits between the haem c_n and haem b_n from the two cytb₆ f monomers. The cytb₆ f complex is coloured as in Fig. 1, and shows c-type haems (f and c_n, dark blue), b-type haems (b_p and b_n, red), 9-cis-β-carotene (orange), chlorophyll a (green), 2Fe-2S (burnt orange and yellow), lipids (white) and plastoquinones (PQ1–PQ3; yellow).

The 1,4-benzoquinone ring of PQ1 is 16.2 Å from haem b_p and 26.4 Å from the 2Fe-2S cluster (Fig. 2e), and distal to the Q_p quinone oxidizing site defined in the M. laminosus cytb₆ f structure²³ (PDB: 4H13) by the inhibitor tridecylstigmatellin (Fig. 3a, b). The Q_p site is located in a pocket formed by hydrophobic residues from subunit IV (Val84, Leu88, Val98 and Met101) and cytb₆ (Phe81, Val126, Ala129, Val133, Val151 and Val154) (Fig. 3c). Bifurcated electron transfer to the 2Fe-2S cluster and haem b_p involves two deprotonation events mediated by the ISP His128 and subunit IV Glu78 residues^2,3, which are buried inside the Q_p pocket (Fig. 3a, b). The OH group of PQ1 is ∼26 Å from His128, a ligand of the 2Fe-2S cluster (Fig. 3b), so PQ1 is unlikely to be oxidized in its resolved position, which probably represents a snapshot of its approach to the Q_p site. It is notable in this regard that our spinach cytb₆ f structure resolves two conformations of the chlorophyll phytyl tail, one of which permits access to Q_p site and one that restricts it (Fig. 3c, d). There is only one position of the phytyl tail for the chlorophyll on the opposing monomer. The bound chlorophyll adjacent to PQ1 may fulfil a gating function at the Q_p pocket, either controlling access of PQH₂ and/or increasing the retention time of the reactive semiplastoquinone (SPQ) intermediate species formed following electron transfer to the 2Fe-2S cluster. Indeed, spin-coupling between the SPQ and the 2Fe-2S cluster has been detected during enzymatic turnover of cytb₆ f but is absent in cytbc₁ complexes that lack the chlorophyll molecule²⁴. SPQ in the 2Fe-2S-bound state does not react with oxygen, providing a potential mechanism to control the release of superoxide from the Q_p site²⁴ and regulate the activity of the LHCII kinase STN7²⁵, which is proposed to bind to cytb₆ f between transmembrane helices F and H of subunit IV²⁶. Another role for chlorophyll in regulating the activity of STN7 could involve PQH₂ displacing the chlorophyll phytyl tail on binding to the Q_p site; this motion could induce a conformational change in STN7 leading to its activation²⁷.

Fig. 3. Conformational alterations in the chlorophyll phytyl chain at the PQH₂-oxidizing Q_p site.

a, Orientation of the PQ1 in relation to the haem b_p, chlorophyll and 2Fe-2S cofactors. The catalytically essential residue E78 and coordinating residues of the 2Fe-2S cofactor are shown. Tridecylstigmatellin (TDS) is a quinone analogue, superimposed according to its position determined in the cyanobacterial complex (PDB: 4H13)²³, and used here to indicate the probable destination of PQ1 in the Q_p pocket. b, The same cofactors and residues as in a, but in relation to a surface view of cytb₆ (green) and subunit IV (sub IV, cyan). c, The Q_p pocket is highlighted with a red dashed line, showing its position in relation to the chlorophyll and PQ1 molecules; the hydrophobic residues of subunit IV (cyan) and cytb₆ that line the pocket are shown as sticks and coloured cyan and green, respectively. d, The two conformations of the chlorophyll tail (represented in dark green and light green) gate (dashed arrow) the entrance to the Q_p pocket.

PQ2 binds towards the stromal face of the complex, 4.4 Å from the haem c_n–b_n pair at the Q_n reducing site (Fig. 2f). The b_n and c_n haems on each monomer are separated by 4.9 Å, with the b_n haem coordinated by His202 and His100 (cytb₆), whereas the vinyl side-chain of haem c_n is covalently linked to Cys35 (cytb₆) (Fig. 2f). The dimerization interface of the cytb₆ f complex creates a cavity, which is proposed to promote transfer of quinones between the Q_p and Q_n sites on neighbouring monomers⁸ (Fig. 4a, b). It is noteworthy that the three resolved PQ molecules inhabit this cavity and that PQ2 assumes a position ‘diagonally’ opposite to PQ1 (Fig. 4c) on the other monomer, as previously suggested²⁸. PQ2 adopts a bowed conformation that straddles the intermonomer cavity with the distal PQ2 tail appearing to partially obstruct the Q_n site in the neighbouring monomer (Fig. 4d, e). This arrangement may have functional importance in preventing the simultaneous binding of PQ molecules at both Q_n sites, avoiding competition for electrons and favouring faster turnover of the Q cycle. Rapid provision of two electrons for PQ2 bound at a particular Q_n site could be facilitated by the 15.3 Å electron-tunnelling distance between b_p haems (Fig. 2c), which enables rapid inter-monomer electron transfer via the ‘bus-bar’ model from the neighbouring low-potential chain^29,30. Alternatively, the second electron could be provided to the haem c_n directly via an FNR–ferredoxin complex bound at the stromal surface via CET^17,28. The haem c_n propionates on the two halves of the cytb₆ f dimer adopt different conformations (Fig. 4f, g); in the PQ-vacant site on the opposing monomer, the haem c_n propionate is more closely associated with Arg207 (Fig. 4f), whereas in the PQ-occupied site, the haem c_n propionate is rotated towards the 1,4-benzoquinone ring of PQ2 (Fig. 4g). The altered ligation of haem c_n on PQ binding is consistent with the downshift of its redox potential³¹, which would strongly favour PQ reduction. We note that the reduction and oxidation of haem c_n is accompanied by the binding and release of one proton³¹ so only one proton is required from the stromal side via the Arg207 and Asp20 residues (Fig. 4f, g) for PQ2 reduction to proceed rapidly, avoiding SPQ formation. It is also possible to position an oppositely oriented PQ within the density map, albeit with a less satisfactory fit (Extended Data Fig. 5). A third PQ molecule (PQ3) (Fig. 2g) has been assigned to the density between the Q_p and Q_n binding sites (see Extended Data Fig. 6 for an alternative assignment as phosphatidylcholine) with the 1,4-benzoquinone ring near the channel that links the two sides of the intermonomer cavity and the isoprenyl tail at the mouth of the Q_p site. This third PQ may therefore capture a snapshot of the molecule transitioning between the Q_p and Q_n sites in opposite monomers.

Fig. 4. The intermonomer cavity of the spinach cytb₆ f complex.

a, b, Surface representations of the complex, with subunits coloured as in Fig. 1, and cofactors and lipids coloured as in Fig. 2. These two views of the complex are related by a 45° rotation about an axis perpendicular to the membrane to show two views of the cavity and the locations of PQ molecules. c, PQ1–PQ3 are shown in relation to the b_n, c_n and b_p haems in the core of the complex, viewed in the membrane plane. d, The complex viewed from the stromal side of the membrane; peripheral helices of cytb₆ and subunit IV are shown in cartoon representation for clarity, to show PQ2 straddling the intermonomer cavity and sitting between the two c_n haems. e, Close-up of the cavity in d. f, g, The head and tail regions of PQ2 in relation to the c_n haems on both sides of the cavity, highlighting the different orientations of the haem c_n propionates, and the Arg207 and Asp20 side chains. The distances in angstroms between the residues and cofactors are labelled.

The cryo-EM structure of spinach cytb₆ f reveals the positions of natively bound PQ and provides details regarding the conformational switches involved in PQ binding to the Q_n site, chlorophyll gating of the Q_p site and PQ exchange between the sites during Q-cycle operation.

Online content

Any methods, additional references, Nature Research reporting summaries, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-019-1746-6.

Methods

No statistical methods were used to predetermine sample size. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment.

Complex purification

Dimeric cytb₆ f was isolated from dark-adapted market spinach (S. oleracea) in a procedure adapted from Dietrich and Kuhlbrandt³².

In brief, spinach leaves were homogenized in buffer 1 (50 mM Tris-HCl pH 7.5, 200 mM sucrose and 100 mM NaCl). Homogenate was then filtered and centrifuged for 15 min at 4,540g, 4 °C. Following centrifugation, the supernatant containing cell debris was discarded and the pellet resuspended in buffer 2 (10 mM Tricine-NaOH pH 8 and 150 mM NaCl) before centrifugation again for 15 min (4,540g, 4 °C). The resultant pellet was resuspended in buffer 3 (2 M NaBr, 10 mM Tricine-NaOH pH 8 and 300 mM sucrose) and incubated on ice for 15 min before diluting twofold with ice-cold milliQ H₂O and centrifuging (15 min, 4,540g, 4 °C). The resultant pellet was resuspended in buffer 3 and incubated on ice for 15 min before diluting twofold with ice-cold milliQ H₂O and centrifuging again (15 min, 4,540g, 4 °C). The pellet was resuspended in buffer 2 and centrifuged for 15 min, 4,540g at 4 °C. The final pellet was resuspended in a small volume of buffer 4 (40 mM Tricine pH 8.0, 10 mM MgCl₂ and 10 mM KCl). The resultant thylakoid suspension was adjusted to 10 mg ml⁻¹ chlorophyll (chlorophyll concentrations determined as described previously³³).

For selective solubilization of cytb₆ f, the thylakoid suspension (10 mg ml⁻¹ chlorophyll) was diluted with membrane extraction buffer (40 mM Tricine pH 8.0, 10 mM MgCl₂, 10 mM KCl and 1.25% (w/v) Hecameg) to a final concentration of 2 mg ml⁻¹ chlorophyll, 1% (w/v) Hecameg. The resultant solution was mixed thoroughly then incubated for 2 min at room temperature before dilution to 0.75% (w/v) Hecameg with buffer 4. Unsolubilized material was removed by ultracentrifugation at 244,000g at 4 °C for 30 min in a Beckman Ti50.2 rotor.

The solubilization supernatant was concentrated using a Centriprep 100K centrifugal filter (Merck Millipore) before loading onto a 10–40% (w/v) continuous sucrose gradient containing 40 mM Tricine pH 8, 10 mM MgCl₂, 10 mM KCl, 0.8% (w/v) Hecameg and 0.1 mg ml⁻¹ egg yolk l-α-phosphatidylcholine (Sigma). This was ultracentrifuged at 174,587g at 4 °C for 16 h in a Beckman SW32 rotor.

A brown-ish band containing cytb₆ f was collected from a region of the gradient corresponding to ∼16% sucrose. This band was concentrated and loaded onto a ceramic hydroxyapatite column (CHT) (Type I, Bio-Rad) pre-equilibrated in 20 mM Hecameg, 0.1 mg ml⁻¹ phosphatidylcholine and 20 mM Tricine pH 8. The column was washed with 5 column volumes of CHT wash buffer (20 mM Hecameg, 0.1 mg ml⁻¹ phosphatidylcholine and 100 mM ammonium phosphate pH 8) before bound material was eluted with CHT elution buffer (20 mM Hecameg, 0.1 mg ml⁻¹ phosphatidylcholine and 400 mM ammonium phosphate pH 8).

Detergent exchange and gel filtration

Concentrated CHT eluate was loaded onto a 10–35% (w/v) continuous sucrose gradient containing 50 mM HEPES pH 8, 20 mM NaCl and 0.3 mM 4-trans-(4-trans-propylcyclohexyl)-cyclohexyl α-maltoside (tPCCαM) and ultracentrifuged at 175,117g at 4 °C for 16 h in a Beck-man SW41 rotor.

A single brown band containing cytb₆ f was collected from a region of the gradient corresponding to ∼22% sucrose. This band was concentrated and loaded onto HiLoad 16/600 Superdex 200 pg gel filtration column (GE Healthcare) connected to an ÄKTA prime plus purification system (GE Healthcare). The column was run at a rate of 0.2 ml min⁻¹ with 145 ml with gel filtration buffer (50 mM HEPES pH 8, 20 mM NaCl, 0.3 mM tPCCαM). Eluted fractions comprising dimeric cytb₆ f were pooled and concentrated.

SDS–PAGE and BN-PAGE analysis

Samples collected from each purification step were analysed by SDS–PAGE and BN-PAGE. For SDS–PAGE, precast NuPAGE 12% Bis-Tris gels (Invitrogen) were run for 60 min at 180 V before staining with Coomassie blue. For BN-PAGE, precast NativePAGE 3–12% Bis-Tris gels (Invitrogen) were run for 120 min at 160 V before staining with Coomassie blue. Gels were imaged using an Amersham 600 imager (GE Healthcare).

Quantification of purified dimeric cytb₆ f using redox difference spectra

Absorbance spectra were recorded at room temperature on a Cary60 spectrophotometer (Agilent). For redox difference spectra cytochromes were first fully oxidized with a few grains of potassium ferricyanide followed by reduction with a few grains of sodium ascorbate (cytf) then sodium dithionite (cytf and cytb₆). At each stage the sample was mixed thoroughly and incubated for ∼1 min before recording spectra. Redox difference spectra (ascorbate-reduced minus ferricyanide-oxidized and dithionite-reduced minus ascorbate-reduced) were calculated and used to determine the concentrations of c-type haem of cytf and the two b-type haems of cytb₆ using extinction coefficients of 25 mM cm⁻¹ (haem f) and 21 mM cm⁻¹ (cytb₆ haems)³⁴.

Reduction of decylplastoquinone

Approximately 0.1 mg decylplastoquinone (Merck) was dissolved in 100 μl ethanol, mixed with a few grains of sodium dithionite dissolved in 100 μl milliQ H₂O and vortexed until the solution became colourless. Decylplastoquinol was extracted by mixing with 0.5 ml hexane, vortexing and centrifuging at 16,000g for 2 min. The hexane layer was carefully removed ensuring none of the aqueous phase was collected. Hexane extraction was repeated on the aqueous phase twice more, then the hexane solutions were pooled and dried in a rotary evaporator at 30 °C for 1 h before re-dissolving in ∼100 μl DMSO. Decylplastoquinol concentration was determined by diluting 10 μl of the DMSO solution into 795 μl ethanol, recording the absorbance spectrum between 250 and 350 nm and using an extinction coefficient³⁵ of 3,540 M⁻¹ cm⁻¹ at 290 nm.

Purification of PC

PC was purified in its oxidized form from market spinach. In brief, spinach leaves were homogenized in buffer containing 50 mM sodium phosphate pH 7.4, 5 mM MgCl₂ and 300 mM sucrose. Homogenate was then filtered and centrifuged for 15 min at 4,000g. Following centrifugation, the supernatant containing cell debris was discarded and the pellet was resuspended in buffer containing 10 mM Tricine pH 7.4 and 5 mM MgCl₂. The solution was incubated on ice for 1 min before diluting twofold with buffer containing 10 mM Tricine pH 7.4, 5 mM MgCl₂, 400 mM sucrose and centrifuging for 15 min at 4,000g. Following centrifugation, the pellet was resuspended to a chlorophyll concentration of 2 mg ml⁻¹ in buffer containing 10 mM HEPES pH 7.6, 5 mM NaCl and 5 mM EDTA, and sonicated for 10 min, at 30 s intervals. The solution was centrifuged at 200,000g for 1 h to pellet any large unbroken material. The supernatant was applied to four 5-ml GE Healthcare Hi-TRAP Q FF anion-exchange columns in series, equilibrated in HEPES pH 8, 5 mM NaCl. A gradient of 0.005–1 M NaCl was used for elution, with PC eluting at around 200 mM. PC-containing fractions were identified by the blue colour on addition of potassium ferricyanide. These fractions were pooled, concentrated in a Vivaspin 3-kDa molecular-weight cut-off spin concentrator and loaded onto a Superdex 200 16/600 FPLC column, equilibrated with 20 mM HEPES pH 8 and 20 mM NaCl. The resulting PC fractions were pooled, concentrated and frozen at −80 °C until use.

Activity assays

Reduction of PC by cytb₆ f was monitored by stopped-flow absorbance spectroscopy using an Olis RSM 1000 rapid-scanning spectrophotometer equipped with a USA-SF stopped flow cell at 20 °C. Solution A (231.25 nM cytb₆ f and 62.5 μM PC in 50 mM HEPES pH 8, 20 mM NaCl and 0.3 mM tPCCαM) and solution B (1.25 mM decylplasto-quinol in the same buffer) were prepared and the reaction was initiated by mixing the solutions in a 4:1 volumetric ratio (final concentrations: 185 nM cytb₆ f, 50 μM PC and 250 μM decylplastoquinol). PC reduction was monitored by recording absorbance spectra between 420 and 750 nm at a rate of 62 scans s⁻¹ and plotting the change in absorbance³⁶ at 597 nm. In a control reaction, cytb₆ f was omitted to record the uncatalysed reduction of PC by decylplastoquinol. Fitting of the initial reaction rates was performed in Origin. All measurements were carried out in triplicate.

CryoEM specimen preparation and data acquisition

In brief, 3 μl of purified cytb₆ f (∼17 μM) was applied to freshly glow-discharged holey carbon grids (Quantifoil R1.2/1.3, 400 mesh Cu). The grids were blotted for 2 s at 8 °C then plunge frozen into liquid ethane using a Leica EM GP at 90% relative humidity. Data acquisition was carried out on a Titan Krios microscope operated at 300 kV (Thermo Fisher) equipped with an energy filtered (slit width 20 eV) K2 summit direct electron detector. A total of 6,035 movies were collected in counting mode at a nominal magnification of 130,000× (pixel size of 1.065 Å) and a dose of 4.6 e⁻ Å⁻² s⁻¹ (see Extended Data Table 1). An exposure time of 12 s was used and the resulting movies were dose-fractionated into 48 fractions. A defocus range of −1.5 to −2.5 μm was used.

Image processing and 3D reconstruction

Beam-induced motion correction and dose-fractionation were carried out using MotionCor2. Contrast transfer function (CTF) parameters of the dose-weighted motion-corrected images were then estimated using GCTF³⁷. All subsequent processing steps were performed using RELION 2.1³⁸ or 3.0³⁹ unless otherwise stated.

In total, 422,660 particles were manually picked from 6,035 micrographs. These particles were extracted using a box size of 220 × 220 pixels and subjected to reference-free 2D classification. A typical micro-graph showing picked particles is shown in Extended Data Fig. 2a, b. Particles that categorized into poorly defined classes were rejected, while the remaining 292,242 (69.2%) particles were used for further processing. A subset of 30,000 particles was used to generate a de novo initial model using the ‘3D initial model’ subroutine. The initial model low-pass filtered to 20 Å was used as a reference map for subsequent 3D classification into 10 3D classes. One stable 3D class at a resolution of 5.38 Å was selected for high-resolution 3D auto-refinement; this class accounted for a subset of 108,560 particles (25.6%). This subset of refined particles was then re-extracted and re-centred before another round of 3D auto-refinement was carried out. The resultant 4.85 Å density map was corrected for the modulation transfer function (MTF) of the Gatan K2 summit camera then further sharpened using the post-processing procedure to 4.02 Å. Per-particle CTF-refinement was carried out and a soft mask was created which included the detergent shell. The final global resolution estimate of 3.58 Å was based on the gold-standard Fourier shell correlation (FSC) cut-off of 0.143.

Local resolution was determined using one of two unfiltered half-maps as an input, a calibrated pixel size of 1.065 and a B-factor of −103. The output local resolution map is shown in Extended Data Fig. 2d, e.

Model building

Initially, a homology-based approach was performed using the crystal-lographic structure of Nostoc sp. PCC 7120 cytb₆ f (PDB: 4OGQ)⁴⁰ as a template. Sequence alignments of the eight polypeptide subunits of cytb₆ f were carried out using Clustal Omega (Extended Data Figs. 7, 8). The model was rigid-body docked into the density using the ‘fit in map’ tool in Chimera⁴¹. This was then followed by manual adjustment and real-space refinement using COOT⁴². Sequence assignment and fitting was guided by bulky residues such as Arg, Trp, Tyr and Phe. After fitting of the polypeptide chains and cofactors in one half of the dimeric complex, the other half of the complex was then independently fitted into the C1 density map. Once both halves of the complex were fitted, cofactors, lipids and plastoquinone-9 molecules were fitted into regions of unassigned density. The final model underwent global refinement and minimization using the real space refinement tool in PHENIX⁴³. The final refinement statistics are summarized in Extended Data Table 1.

Pigment analysis by reversed-phase HPLC

Pigments were extracted from purified cytb₆ f with 7:2 acetone:methanol (v/v) and clarified extracts were separated by reversed-phase HPLC at a flow rate of 1 ml min⁻¹ at 40 °C using a Supelco Discovery HS C18 column (5 μm particle size, 120 Å pore size, 25 cm × 4.6 mm) on an Agilent 1200 HPLC system. The column was equilibrated in acetonitrile: water:trimethylamine (9:1:0.01 v/v/v) and pigments were eluted by applying a linear gradient of 0–100% ethyl acetate over 15 min followed by isocratic elution with 100% ethyl acetate for a further 5 min. Elution of carotenoid and chlorophyll species was monitored by absorbance at 400, 450, 490 and 665 nm. Chlorophyll a was identified by its absorption spectra and known retention time⁴⁴. The major carotenoid species was confirmed as 9-cis β-carotene using a standard obtained from Sigma-Aldrich (product no. 52824).

Extended Data

Extended Data Fig. 1. Purification of cytb₆ f from spinach.

a, Absorption spectrum of ascorbate-reduced purified b₆ f complex. The peak at 421 nm corresponds to the Soret band of bound pigments (chlorophyll a and haems). The peaks at 554 and 668 nm correspond to c-type haems of cytf and chlorophyll a, respectively. The inset panel shows redox difference spectra of ascorbate-reduced minus ferricyanide-oxidized b₆ f (dashed line) and dithionite-reduced minus ascorbate-reduced (dotted line) cytb₆ f. Redox difference spectra show haem f absorption peaks at 523 and 554 nm as well as absorption peaks at 534 and 563 nm corresponding to the b-type haems of cytb₆. The calculated ratio of cytb₆ b-type haems to the c-type haem of cytf was ∼2 using extinction coefficients of 25 mM cm⁻¹ (f) and 21 mM cm⁻¹ (b₆)³⁴. The spectra exhibit the absorption properties characteristic of intact cytb₆ f. Spectra were recorded at room temperature. b, SDS–PAGE analysis of purified cytb₆ f indicates that the sample is highly pure, with the four large subunits of the complex (cytf, cytb₆, the Rieske ISP and subunit IV) running at ∼31 kDa, ∼24 kDa, ∼20 kDa and ∼17 kDa, respectively and the four small subunits (PetG, PetL, PetM and PetN) running at around 4 kDa (not shown). c, d, Negative-stain and BN-PAGE analysis of purified cytb₆ f demonstrates the sample is dimeric and highly homogenous, with a single band corresponding to dimeric cytb₆ f shown in lane 1. Lane 2 shows a sample that has been deliberately monomerized following incubation with 1% Triton X-100 for 1 h. For gel source data see Supplementary Fig. 1. e, The catalytic rate of plastocyanin reduction by the purified dimeric cytb₆ f complex as determined by stopped-flow absorbance spectroscopy. A rate of 200 e⁻ s⁻¹ was determined by taking the initial linear region from the enzyme-catalysed reaction (solid line) and subtracting the background rate measured in the absence of enzyme (long-dashed line). Plastocyanin reduction was not observed in the absence of decylplastoquinol (short-dashed line). Reactions were initiated upon addition of decylplastoquinol to the solution containing plastocyanin and b₆ f while monitoring the loss of absorbance at 597 nm. Final concentrations were 50 μM plastocyanin, 185 nm b₆ f and 250 μM decylplastoquinol. All experiments were performed in triplicate and controls were performed in the absence of b₆ f or decylplastoquinol.

Extended Data Fig. 2. Cryo-EM micrographs of the spinach cytb₆ f complex and calculation of the cryo-EM map global and local resolution.

a, Cytb₆ f particles covered by a thin layer of vitreous ice on a supported carbon film. b, Examples of dimeric cytb₆ f particles are circled in green. We recorded 6,035 cryo-EM movies, from which 422,660 particles were picked manually for reference-free 2D classification. The final density map was calculated from 108,560 particles. c, Gold-standard refinement was used for estimation of the final map resolution (solid black line). The global resolution of 3.58 Å was calculated using a FSC cut-off at 0.143. A model-to-map FSC curve (solid grey line) was also calculated. d, e, A C1 density map of the cytb₆ f complex both with (d) and without (e) the detergent shell. The map is coloured according to local resolution estimated by RELION and viewed from within the plane of the membrane. The colour key on the right shows the local structural resolution in angstroms (Å).

Extended Data Fig. 3. Cryo-EM densities and structural models of polypeptides in the cytb₆ f complex.

Polypeptides are coloured as in Fig. 1. The contour levels of the density maps were adjusted to 0.0144.

Extended Data Fig. 4. Cryo-EM densities and structural models of prosthetic groups, lipids and plastoquinone molecules in the cytb₆ f complex.

c-type haems (f, c_n; dark blue), b-type haems (b_p, b_n; red), 9-cis β-carotene (orange), chlorophyll a (major conformation, dark green; minor conformation, light green), 2Fe-2S (burnt orange and yellow), plastoquinones (yellow), monogalactosyl diacylglycerol (light pink), phosphatidylcholine (light cyan), sulfoquinovosyl diacylglycerol (light green) and phosphatidylglycerol (light purple). The contour levels of the density maps were adjusted to 0.0068.

Extended Data Fig. 5. Alternative interpretation of the region assigned as PQ2.

a, b, The density map showing two possible alternative conformations for PQ2, the major conformation (a) and the alternative conformation (b). Cofactors are coloured as in Extended Data Fig. 4 with b-type haems (b_p and b_n) coloured red, c-type haems (c_n) coloured dark blue, chlorophyll a (major conformation) coloured dark green, plastoquinones coloured yellow and the cytb₆ subunit coloured light green. The contour level of the density map was adjusted to 0.0089.

Extended Data Fig. 6. Alternative interpretations of the density map in the region assigned as PQ3.

a, b, The density map modelled with a plastoquinone molecule (a) and a phosphatidylcholine molecule (b). Top, the protein-free density map; bottom, the map including cytb₆ (green). The 2.9 Å distance indicates a close contact between the PQ3 head group and the conserved Lys208. Cofactors are coloured as in Extended Data Fig. 4 with b-type haems (b_p and b_n) coloured red, chlorophyll a (major conformation) in dark green, plastoquinones in yellow, phosphatidylcholine in light cyan, sulfoquinovosyl diacylglycerol in mint green and the cytb₆ subunit in light green. The contour level of the density map was adjusted to 0.0127.

Extended Data Fig. 7. Multiple sequence alignment of cytb₆ f subunits cytf and cytb₆.

a, b, Sequences of cytf (a) and cytb₆ (b) from cyanobacterial (M. laminosus and Nostoc sp. PCC7120), algal (C. reinhardtii) and plant (S. oleracea) subunits were aligned in Clustal Omega v.1.2.4. Conserved identities are indicated by asterisks, and similarities by double or single dots. Polar residues are coloured in green, positively charged residues are pink, hydrophobic residues are red and negatively charged residues are blue. The sequences omit signal peptides.

Extended Data Fig. 8. Multiple sequence alignment of the Rieske ISP, subunit IV, PetG, PetL, PetM and PetN.

a–f, Sequences of Rieske ISP (a), subunit IV (b), PetG (c), PetL (d), PetM (e) and PetN (f) from cyanobacterial (M. laminosus and Nostoc sp. PCC7120), algal (C. reinhardtii) and plant (S. oleracea) subunits were aligned in Clustal Omega v.1.2.4. Conserved identities are indicated by asterisks, and similarities by double or single dots. Polar residues are coloured in green, positively charged residues are pink, hydrophobic residues are red and negatively charged residues are blue. The sequences omit signal peptides.

Extended Data Table 1. Cryo-EM data collection, refinement and validation statistics.

	S. oleracea cytb6f (EMD-4981) (PDB 6RQF)
Data collection and processing
Magnification	130,000 X
Voltage (kV)	300
Electron exposure (e–/Å2)	1.15 (55.2 e- on 48 frames)
Defocus range (μm)	-1.5 to-2.5
Pixel size (Å)	1.065
Symmetry imposed	Cl
Initial particle images (no.)	422,660
Final particle images (no.)	108,560
Map resolution (Å)	3.58
FSC threshold	0.143
Map resolution range (Å)	~3.3-8.3
Refinement
Initial model used (PDB code)	RELION de novo model from 30,000 particles
Model resolution (Å)	3.58
FSC threshold	0.143
Model resolution range (Å)	~3.3-8.7
Map sharpening B factor (Å2)	Estimated automatically using RELION*
Model composition
Non-hydrogen atoms	16,359
Protein residues	1,944
Ligands	29
B factors (Å2)
Protein	RELION auto-estimated
Ligand	RELION auto-estimated
R.m.s. deviations (PHENIX)
Bond lengths (Å)	0.009
Bond angles (°)	1.182
Validation
MolProbity score	1.83
Clashscore	9.67
Poor rotamers (%)	0.18
Ramachandran plot
Favored (%)	95.40
Allowed (%)	4.24
Disallowed (%)	0.37

^*Data from ref. ⁴⁵.

Extended Data Table 2. Comparison of cofactor distances in b₆f and bc₁ dimers from different species.

	(PDB 6RQF)	(PDB 1Q90)	(PDB 2E74)	(PDB 4OGQ)
Source	S. oleracea	C. reinhardtii	M. laminosas	Nostoc sp. PCC 7120
Resolution (Å)	3.6	3.1	3.0	2.5
Inhibitors *	-	TDS (Qp)	-	-
Distances:
bn- cn (Å)	4.7, 4.7	4.7, 4.7	4.7, 4.7	4.6, 4.6
bn- bp (Å)	12.1, 12.0	12.2, 12.2	12.2, 12.2	12.1, 12.1
bp- bp (Å)	15.3	15.1	15.2	15.3
bp - [2Fe-2S] (Å)	25.6, 25.5	22.9, 22.9	25.5, 25.5	25.3, 25.3
[2Fe-2S]-f (Å)	25.9, 26.1	27.8, 27.8	26.2, 26.2	26.2, 26.2
	(PDB1BCC) (distal)	(PDB 3BCC) (proximal)
Source	G. gallus	G. gallus
Resolution (Å)	3.2	3.7
Inhibitors *	-	STG (Qp), AMY (Qn)
Distances:
bn- bp (Å)	12.4, 12.4	12.3, 12.3
bp- bp (Å)	14.4	14.5
bp- [2Fe-2S] (Å)	30.3, 30.3	23.0, 23.1
[2Fe-2S]-c1 (Å)	16.8, 16.8	27.3, 27.5

The distances are edge-to-edge (Å), for each half of the b₆f dimer from different species (PDB: 6RQF, 1Q90, 2E74, 4OGQ) and the bc₁ dimer from Gallus gallus with the Rieske ISP in its distal (PDB: 1BCC) and proximal (PDB: 3BCC) positions. AMY, antimycin; STG, stigmatellin.

Reporting summary.

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Data availability

All relevant data are available from the authors and/or are included with the manuscript or in the Supplementary Information. Atomic coordinates and the cryo-EM density map have been deposited in the Protein Data Bank under accession number 6RQF and the Electron Microscopy Data Bank (EMDB) under accession number EMD-4981.