Structural Basis of Metabolite Transport by the Chloroplast Outer Envelope Channel OEP21 PDF

nature structural & molecular biology Article https://doi.org/10.1038/s41594-023-00984-y Structural basis of metabolite transport by the chloroplast outer envelope channel OEP21 Received: 16 February 2023...

nature structural & molecular biology Article https://doi.org/10.1038/s41594-023-00984-y Structural basis of metabolite transport by the chloroplast outer envelope channel OEP21 Received: 16 February 2023 Umut Günsel 1,2,5, Kai Klöpfer1,5, Elisabeth Häusler2,5, Manuel Hitzenberger3, Bettina Bölter4, Laura E. Sperl1, Martin Zacharias 3, Jürgen Soll4 Accepted: 31 March 2023 & Franz Hagn 1,2 Published online: 8 May 2023 Check for updates Triose phosphates (TPs) are the primary products of photosynthetic CO2 fixation in chloroplasts, which need to be exported into the cytosol across the chloroplast inner envelope (IE) and outer envelope (OE) membranes to sustain plant growth. While transport across the IE is well understood, the mode of action of the transporters in the OE remains unclear. Here we present the high-resolution nuclear magnetic resonance (NMR) structure of the outer envelope protein 21 (OEP21) from garden pea, the main exit pore for TPs in C3 plants. OEP21 is a cone-shaped β-barrel pore with a highly positively charged interior that enables binding and translocation of negatively charged metabolites in a competitive manner, up to a size of ~1 kDa. ATP stabilizes the channel and keeps it in an open state. Despite the broad substrate selectivity of OEP21, these results suggest that control of metabolite transport across the OE might be possible. Chloroplasts are the primary sites of photosynthesis in plant cells. of 75 kDa), which is required for pre-protein import into chloroplasts, During photosynthesis, ATP and NADPH are generated and used for fixa- more channels for the transport of metabolites and solutes have tion of CO2 into the TP d-glyceraldehyde-3-phosphate (GAP) through been discovered6,7. This includes the amine and amino acid channel the Calvin–Benson–Bassham cycle in the chloroplast stroma. These OEP16 (ref. 8), OEP21 (ref. 9) and the cation-selective channels OEP23 primary products must be transported across the IE and OE mem- (ref. 10), OEP24 (ref. 11) and OEP37 (ref. 12). TP translocation across the branes into the cytosol for further metabolic processing and energy OE membrane is mainly conducted by OEP21 and OEP24 (refs. 6,7,13) supply1. Each of these membranes is equipped with a distinct set of ion (Fig. 1a). To account for the different levels of metabolite flux in distinct channels and transporters that enable transport of nutrients, solutes plant species, the abundance of individual OEPs can vary strongly. The and metabolites (Fig. 1a). Transport across the IE membrane is medi- larger and less selective OEP24 pore is abundant in C4 plants, in which a ated by a diverse set of proteins (for example, TGD-1, PPT, TPT, NNT)2, higher rate of carbon fixation and metabolite flux occurs13. By contrast, whereas the OE membrane was initially considered to be a permeable the outward-rectifying OEP21 channel is more prominent in C3 plants sieve that cannot form a barrier for small molecules3. This view was and has been shown to interact with ATP, TPs and phosphate in a com- questioned by the finding that the OE contains multiple substrate and petitive manner9,14. However, the molecular and structural features of charge-selective channels4, suggesting a more specific transport mech- metabolite transport and selectivity remain elusive for OEP21, as well anism5. Beside TOC75 (translocase of the outer chloroplast membrane as the entire OEP channel family. 1 Bavarian NMR Center (BNMRZ), Department of Bioscience, School of Natural Sciences, Technical University of Munich, Garching, Germany. 2Institute of Structural Biology, Helmholtz Munich, Neuherberg, Germany. 3Lehrstuhl für Theoretische Biophysik (T38), Department of Bioscience, School of Natural Sciences, Technical University of Munich, Garching, Germany. 4Biozentrum, LMU München, Department of Biology, Planegg-Martinsried, Germany. 5 These authors contributed equally: Umut Günsel, Kai Klöpfer, Elisabeth Häusler. e-mail: [email protected] Nature Structural & Molecular Biology | Volume 30 | June 2023 | 761–769 761 Article https://doi.org/10.1038/s41594-023-00984-y a Cytosol b × 104 c 1.0 1.0 Apo: Normalized CD signal 58.5 °C Θ MRW (deg cm2 dmol–1) 0.5 0.8 GAP: 0.6 59.2 °C OEM OEP21 OEP24 OEP37 0 0.4 ATP: –0.5 63.1 °C 0.2 –1.0 0 –1.5 Pi TP IMS 200 220 240 260 20 40 60 80 100 ATP Wavelength (nm) Temperature (°C) d L5 L8 4 β1 β2 β3 β4 β5 β6 β7 β8 β9 β10 β11 β12 Sec. C.S. (ppm) IEM TPT 0 –4 –8 Stroma 0 20 40 60 80 100 120 140 160 180 TP Residue number e Loop 5 f (aa 70–80) 11 Å d = 11.6 Å 30 Pore z-coordinate (Å) +20 Loop 8 20 (aa 129–139) 0 d = 7.8 Å –20 10 Electrostatic potential (kT e–1) N d = 11.4 Å C 0 12.0 10.4 8.8 7.2 22 Å Pore diameter (Å) Fig. 1 | OEP21 is a highly positively charged β-barrel porin in the OE of 12 β-strand regions of OEP21. e, NMR structural bundle of the OEP21 β-barrel membrane. a, TP transport across the IE and OE membranes by the TP/Pi pore showing well-defined secondary structure elements (r.m.s.d. of 0.5 Å) and translocator (TPT, PDB:5Y78)40 and OE proteins (OEPs) of various sizes, such a funnel-like shape. f, Analysis of the pore geometry, indicating a 7.8-Å-wide as OEP21, OEP24 and OEP37. Pi, inorganic phosphate. b, Far-UV CD spectrum of constriction site on the electrostatic potential surface map of the pore interior recombinant OEP21 in LDAO micelles. ΘMRW, ellipticity mean residue weight. (blue indicates positively charged regions, and negatively charged regions are c, Thermal stability of apo-OEP21 and in presence of 0.5 mM GAP or ATP. shown in red). d, NMR secondary chemical shift (Sec. C.S.) information indicating the presence Here, we used NMR spectroscopy to determine the high-resolution Results structure of OEP21 from garden pea. We show that this channel consists OEP21 is a β-barrel membrane pore with a highly positively of 12 β-strands that form a cone-shaped β-barrel pore, with the wider charged interior surface opening oriented toward the chloroplast intermembrane space (IMS). We first optimized protein refolding conditions for OEP21 in detergent The inside of the pore is highly positively charged, suggesting specific micelles and identified the detergent lauryldimethylamine-N-oxide binding and translocation of negatively charged metabolites. Metabo- (LDAO), which has been previously reported to be suitable for the refold- lites interact with OEP21 in a charge-dependent and competitive man- ing of β-barrel membrane proteins15,16. An analysis of OEP21 in LDAO ner. Interestingly, binding of ATP stabilizes the channel and affects the micelles by far-ultraviolet (UV) circular dichroism (CD) spectroscopy oligomer-to-monomer equilibrium. Using NMR and molecular dynamics indicated that the protein had β-sheet secondary structure (Fig. 1b). (MD) simulations, we show that the translocation trajectory of GAP is Because OEP21 has been reported to bind or transport GAP and ATP9, we guided by patches of positive charges in the channel. Finally, we show that next performed CD-detected thermal melting experiments (Fig. 1c) in not only TPs, but also larger molecules up to a molecular weight of ~1 kDa the presence of these molecules. GAP, and especially ATP, led to stabi- can pass OEP21. Taken together, these data provide detailed mechanistic lization of OEP21. ATP is present at a concentration of ~1–2 mM in plant insights into the functionality of an important member of the OEP family cells17. Thus, we added ATP to the OEP21 sample for the subsequent NMR and suggest that these pores show a distinct level of selectivity and might structure determination. With this setup, we obtained high-quality be able to respond to changes in the cellular milieu in plants. two-dimensional (2D) and multidimensional NMR spectra (Extended Nature Structural & Molecular Biology | Volume 30 | June 2023 | 761–769 762 Article https://doi.org/10.1038/s41594-023-00984-y Table 1 | NMR and refinement statistics for ATP-bound by NMR paramagnetic relaxation enhancement (PRE) experiments OEP21 in LDAO micelles using a spin-labeled fatty acid (12-doxyl-stearic acid) that led to signal attenuation of membrane-incorporated parts of OEP21 (Extended Data OEP21 Fig. 1e,f), but left L5 unaffected. An analysis of the inner pore diameter NMR distance and dihedral constraints (Fig. 1f) indicates that the most constricted position of the channel is Distance constraints 7.8 Å wide. The dimensions of the OEP21 pore are in good agreement with previously published electrophysiology data14, which led to esti- Total NOE 623 mations of a wider vestibule of 2.4 nm and a restriction zone of ~1 nm. Intra-residue 106 The determined structure reveals that the pore has a highly positively Inter-residue 517 charged inner surface (Fig. 1f), which suggests that it binds to metabo- lites in a charge-dependent manner, supported by the increased ther-   Sequential (|i – j| = 1) 154 mal stability when bound to GAP and ATP (Fig. 1c).   Medium-range (|i – j| < 4) 57   Long-range (|i – j| > 5) 306 Orientation and oligomeric state of OEP21 in the chloroplast OE   Intermolecular 0 The funnel-like shape of OEP21 is in agreement with its proposed outward-rectifying properties14. Because metabolite flow is mainly Hydrogen bonds 110 directed from the IMS into the cytosol under light conditions of pho- Total dihedral angle restraints 255 tosynthesis, an orientation with the wider opening in the IMS is plau- ϕ 128 sible. To probe the orientation of OEP21 in the OE, we next applied ψ 127 limited proteolysis experiments with isolated right-side-out OE vesicles (OEVs)21,22. In agreement with published data23, treatment of OEVs with Structure statistics trypsin resulted in two specific fragments (Fig. 2a, left). Decoration Violations (mean and s.d.) of the same samples with antibodies against Toc64 and Toc75 (refs. Distance constraints (Å) 0.035 ± 0.002 24,25) confirmed the efficiency of the trypsin treatment (Extended Data Fig. 2a), i.e. the large cytosolic domain of Toc64 is digested25 while the Dihedral angle constraints (°) 0.11 ± 0.04 membrane-inserted part remains intact. Toc75 remains mostly intact Max. dihedral angle violation (°) 1.211 as it is deeply embedded within the bilayer, exposing only short loops Max. distance constraint violation (Å) 0.203 to the cytosol26,27. N-terminal Edman sequencing of the immunopre- Deviations from idealized geometry cipitated larger fragment of OEP21 (Fig. 2b and Extended Data Fig. 2b) unambiguously identified a trypsin cleavage site in L5 (Fig. 2c)5, indicat- Bond lengths (Å) 0.00384 ± 0.00008 ing that L5 is oriented toward the cytosol, and revealed that the N- and Bond angles (°) 0.87 ± 0.01 the C-termini are both located in the IMS. Trypsin-digestion experi- Impropers (°) 2.4 ± 0.1 ments with recombinant OEP21 reconstituted in liposomes showed an identical cleavage pattern (Fig. 2a, right), suggesting that refolded Average pairwise r.m.s. deviationa (Å) OEP21 adopts a native topology. The addition of ATP or GAP to the Heavy 1.1 ± 0.1 liposome preparations slightly protected L5 from proteolytic diges- Backbone 0.5 ± 0.1 tion (Extended Data Fig. 2c), implying that the molecules interact with a Pairwise r.m.s. deviation was calculated among 10 refined structures within ordered the loop. However, the orientation of OEP21 in liposomes could not secondary structure elements (residues 1–9, 14–25, 28–36, 44–54, 58–68, 81–94, 102–113, be controlled, giving rise to only partial cleavage, whereas complete 117–128, 140–146, 151–158, 161–166, 169–176). cleavage was observed after liposome disruption by the addition of detergent (Extended Data Fig. 2d). Data Fig. 1a,b), which enabled us to obtain sequence-specific backbone Crosslinking experiments have shown that OEP21 forms larger resonance assignments. An analysis of the secondary 13C chemical shift oligomers in the OE9. To corroborate this finding, we performed Blue information indicated that there were 12 β-strand regions of varying Native (BN)-PAGE (Extended Data Fig. 2e,f) of OEVs extracted with the lengths (Fig. 1d), in contrast to the primary-sequence-based predicted mild detergent DDM, in which OEP21 oligomers could be detected in secondary structure content (8 β-strands)9, but in good agreement with a molecular weight range of 40 to approximately 200 kDa, confirm- the prediction of the program AlphaFold18. High-resolution NMR struc- ing the presence of dimers and higher-order oligomers in a native ture determination was conducted with a uniformly 2H- and 15N-labeled membrane environment. Chemical crosslinking experiments with and selectively methyl-group-labeled (Ile-δ1, Leu-δ2, Val-γ2, Ala-β) recombinant OEP21 in liposomes (Fig. 2d) or LDAO detergent micelles OEP21 sample bound to ATP for the acquisition of a set of heteronuclear (Extended Data Fig. 2g) also showed the presence of higher-order three-dimensional (3D)-NOESY NMR experiments19. These data were oligomers. ATP, which stabilizes OEP21, led to slightly reduced oli- used to assign the side chain methyl resonances (Extended Data Fig. 1c) gomer formation. Furthermore, the oligomeric state of OEP21 can be and extract NOE distance restraints for structure calculation20 (Table 1 controlled by adjusting the detergent concentration (Extended Data and Extended Data Fig. 1d), which resulted in a well-defined structural Fig. 2h). To estimate the impact of oligomerization on the OEP21 struc- bundle of OEP21 showing a root mean square deviation (r.m.s.d.) of ture, we recorded 2D-[15N,1H]-TROSY NMR experiments at a 100- or 0.5 Å in ordered secondary structure elements (Fig. 1e and Table 1). 300-mM detergent concentration (Fig. 2e). At the lower concentration, The overall shape of OEP21 resembles a funnel with a diameter of 22 Å we observed strong line broadening of the backbone amide resonances on one side and 11 Å on the opposite side. Outside the membrane, most in the folded β-barrel, as expected for a larger assembly. However, connecting loops are structurally well-defined, with a length of 3–7 strong NMR signals originating from L5 were visible in the oligomeric amino acids (aa), except for the two longer loops (L5: aa 70–80 and state, suggesting that the cytosolic loop is affected by oligomerization. L8: aa 129–139) that are located on each side of the β-barrel (Fig. 1e). In addition, signals in the random coil chemical shift region appeared L8 is not visible in the NMR spectra to a large extent, presumably in the spectrum, possibly caused by squeezing of the β-barrel in the owing to µs-to-ms motions. By contrast, L5 gives rise to strong NMR oligomer. Interestingly, the addition of ATP improved the NMR spectral signals but does not show NOE contacts to other parts of the protein, quality and resulted in the disappearance of the random coil peaks, con- indicating intrinsic flexibility. Solvent accessibility of L5 was probed firming the assumption that ATP can stabilize the shape of the OEP21 Nature Structural & Molecular Biology | Volume 30 | June 2023 | 761–769 763 Article https://doi.org/10.1038/s41594-023-00984-y a Trypsin Trypsin b c d V R Y Trypsin MW – + MW – + MW G A cleavage + GAP + ATP (kDa) (kDa) (kDa) V Apo K MW 100 S L5 (kDa) Higher 66 70 70 D N Flexible L5 Pentamer 50 Q D 100 Tetramer 55 82 70 45 35 IgGH K Trimer 68 K 50 35 β6 Dimer 35 25 35 25 β5 Cytosol 20 25 IgGL 25 Monomer 18 15 15 20 14 7 YAKNDK OEP21 15 10 IMS 7 Lipids α-OEP21 e f g WT C109A 15 200 2+ δ1( N) 300 mM LDAO 100 mM LDAO 100 mM LDAO Cu [ppm] L5 + ATP MW – + + 110 G74 (kDa) ATP 160 L5 ATP 75 S72 ATP 63 KD for ATP (µM) 115 ATP N80 K79 120 48 d V75 L5 120 Y77 35 R76 V73 80 25 125 m A78 20 Unfolded Unfolded 40 130 17 0 11 11 10 9 8 7 11 10 9 8 7 11 10 9 8 7 10 20 30 50 100 1 δ2( H) [ppm] [LDAO] (mM) Fig. 2 | Oligomeric state and orientation of OEP21 in the chloroplast outer to 5 mM GAP or ATP by BS crosslinking. e, 2D-[15N,1H]-TROSY NMR experiments 3 envelope. a, Analysis of trypsin-treated (+) or untreated (–) isolated OE vesicles with 2H,15N-labeled OEP21 at high (light blue) and low (red) detergent conditions (left) or recombinant OEP21 reconstituted liposomes (right) by immunoblotting and in the presence of ATP (magenta). Encircled NMR signals in the random coil against OEP21 or Coomassie staining. b, Coomassie-stained PVDF membrane region are visible only in the oligomeric apo state. f, Affinity between OEP21 and from the immunoprecipitated trypsin-treated OEP21 fragment used for ATP at the indicated LDAO concentrations, derived from ITC experiments. Bars Edman sequencing. The immunoprecipitated fragment is indicated by the represent the mean value of n ≥ 1 individual measurements. g, Non-reducing one-letter amino acid code, as are the positions of the light and heavy chains SDS–PAGE of WT OEP21 and OEP21-C109A (20 μM) in the absence (–) or presence from the antiserum. c, Topology of OEP21 in the OE. The position of the trypsin (+) of oxidizing 1 mM Cu2+. d, dimer; m, monomer. a, d and g are representative of cleavage site suggests that L5 is oriented toward the cytosol. d, Analysis of the n ≥ 2 independent experiments. oligomerization state of OEP21 in liposomes (20 μM) in the apo form or bound β-barrel and reduce its oligomeric state. Using isothermal titration with metabolites, we determined the binding affinities of relevant nega- calorimetry (ITC) experiments, we showed that OEP21’s affinity for ATP tively charged molecules by a combination of ITC and NMR to character- decreases at detergent conditions that favor the oligomer, presumably ize high- and low-affinity interactions, respectively. ATP, carrying four owing to decreased accessibility of the binding site (Fig. 2f). A com- negative charges at physiological pH, shows an affinity in the low µM parison of the 2D-[15N,1H]-TROSY spectra of OEP21 at 300 and 500 mM range, with a 1:1 binding stoichiometry (Fig. 3a). By contrast, as probed LDAO (Extended Data Fig. 3a,b) showed spectral changes at one side of by NMR, the monophosphorylated substrate GAP binds with a much the β-barrel, likely indicating the dimerization surface. Interestingly, a lower affinity of ~150 µM (Fig. 3b). To identify the molecular features cysteine residue (Cys109) in the center of this region is well-positioned that are required for binding to OEP21, we determined the affinities to engage in a disulfide bridge. An analysis of Cu2+-oxidized OEP21 by of a larger pool of metabolites (Fig. 3c). The high affinity for ATP was SDS–PAGE clearly showed that cysteine-mediated crosslinking is pos- not altered by changes in the pH, whereas the number of phosphate sible (Fig. 2g). In addition, NMR 13Cα and 13Cβ chemical shifts of Cys109 moieties in the metabolite appeared to be more critical: the affinity in OEP21 indicate the presence of a disulfide bridge28 (Extended Data for ADP3− was markedly reduced, and no binding could be detected for Fig. 3c), which enabled the assembly of a dimeric disulfide-bridged AMP2− by ITC, suggesting an equally weak interaction as that observed structural model of OEP21 (Extended Data Fig. 3d and Supplemen- for GAP (Fig. 3b). Other triphosphate nucleosides (GTP, UTP, CTP) tary Data 1). Changing Cys109 to alanine does not alter the second- showed a similar binding affinity to that of ATP, with a slightly lower ary structure content of OEP21 but leads to a reduction in its thermal affinity for the smaller pyrimidine nucleotides (UTP and CTP). These stability (Extended Data Fig. 3e,f), demonstrating the importance of data suggest that binding of metabolites to OEP21 is dominated by their the cysteine residue. Strikingly, we observed a severe enhancement negative charge density. However, in addition to the phosphate moiety, in oligomerization when Cys109 is oxidized, whereas GAP or, more other parts of the metabolite contribute to the interaction, as is evident prominently, ATP reduced this process (Extended Data Fig. 3g). from the large difference in affinity of GAP and phosphate (Fig. 3b,c). In line with the binding data, the stabilization of OEP21 was stronger with OEP21 binds to metabolites by a promiscuous electrostatic molecules carrying a higher number of negative charges (Extended mechanism Data Fig. 4a). To obtain higher-resolution insights on the ATP-binding It has been suggested that ATP can inhibit TP binding to OEP21 (ref. 9). mode, we next conducted NMR titration experiments and analyzed To reveal the molecular and structural basis for the interaction of OEP21 the ligand-induced chemical shift perturbations in OEP21 (Fig. 3d and Nature Structural & Molecular Biology | Volume 30 | June 2023 | 761–769 764 Article https://doi.org/10.1038/s41594-023-00984-y a –5 b 1 c 24 Fold increase in KD 20 –10 20 line broadening ∆H (kJ mol–1) 16 Rel. NMR KD (mM) –15 15 0.5 12 –20 ATP: GAP: 10 KD = 7.7 ± 2.8 µM 8 KD = 152 ± 19 µM –25 n = 1.14 ± 0.16 5 n.b. 4 –30 0 0 0 0 1 2 3 4 5 0 0.5 1.0 1.5 2.0 Molar ratio (ATP:OEP21) [GAP] (mM) 7 6 P P TP TP P AP e pH at AD AM UT pH G C G ph P P os AT AT Ph d L5 e L5 R142 K104 K90 R174 ATP2 R84 R51 N164 180° ATP1 CSPs (ppm) R140 > m.v. + s.d. R157 3 R66 R142 > m.v. R174 R33 C109 K104 K90 Fold increase in KD K19 2 1:1 molar ratio Excess ATP 1 f 120 50 0 T K1 5 R3 A R5 A 6A 6 A N 6A R6 R 6W K9 A K1 A C 4A R1 9A R1 A R1 2A R1 7A A Fold increase in KD for ATP W ∆L 4 9 3 1 4 0 40 74 5 4 16 40 R8 0 10 R 80 g R6 30 L5 20 40 >1 90° Rel. NMR 0.37 10 intensity 0.27 0 0 0 L5 0 0.5 1.0 2.0 0 1 10 50 [NaCl] (M) [MgCl2] (mM) h i j 1.0 GAP GAP + Mg2+ Phosphate 30 1.0 4 OEP21:ATP complex 0.8 0.8 KD (mM) for GAP IC50 (mM) 3 20 0.6 0.6 2 0.4 0.4 10 0.2 OEP21-WT 0.2 1 OEP21∆L5 0 0 0 0 0 0.4 0.8 1.2 –1.0 0 1.0 –2 –1 0 1 2 WT ∆L5 WT ∆L5 WT ∆L5 + + T O Mg 2 5 5 W 1-W log([GAP]) log([GAP]) log([Pi]) -L ∆L 21 21 2 EP EP EP AP 2+ te T O O a g G ph M - 21 os + EP AP Ph O G Fig. 3 | Metabolites bind to OEP21 in a charge-dependent and competitive residues obtained from n = 3 fluorescence polarization (FP) measurements with manner. a,b, ITC (a) and NMR (b) binding experiments with OEP21 and ATP or GAP. MANT-ATP. f, Effect of NaCl and MgCl2 concentration on the affinity between ΔH, binding enthalphy. c, Affinities of OEP21 for negatively charged metabolites. OEP21 and ATP, measured by FP. Bars represent mean value of n = 3 individual Bars indicate mean value of individual measurements, which are multiple ITC or measurements. g, Relative NMR signal intensities of OEP21 upon the addition of FP experiments (n ≥ 2) or individual residues from the NMR titration experiment 5 mM GAP mapped onto the OEP21 structure. h, The affinity of OEP21 for GAP in the (n = 12). n.b., no binding could be detected by ITC for AMP. d, NMR chemical shift presence of Mg2+or with deletion of L5. OEP21 L5 alone weakly interacts with GAP, perturbations mapped onto an MD-based complex structural model of OEP21 and as probed by NMR. i, Competition experiments with a complex of MANT-ATP and ATP at an internal, high-affinity (ATP1) and a peripheral, low-affinity (ATP2) binding OEP21-WT or OEP21ΔL5 upon stepwise addition of GAP, GAP + Mg2+, or phosphate. site. m.v., mean value; s.d., standard deviation. e, Relative KD values of OEP21 j, IC50 values derived from the experiments shown in i. Bars in h and j represent variants without L5 or containing single point mutations of positively charged mean values of n ≥ 2 measurements. Extended Data Fig. 4b). We identified a binding site located within the the low µM range, in good agreement with the ITC data, whereas the β-barrel pore that was fully occupied at a 1:1 protein-to-ATP molar ratio. peripheral binding site exhibits an average dissociation constant (KD) At excess ATP, CSPs and an increase in NMR signal intensity were visible value of ~500 µM (Extended Data Fig. 4e), which could not be detected at the cytosolic entry of the pore, involving L5 (Fig. 3d and Extended by ITC. Furthermore, deletion of L5 in OEP21 (ΔL5) led to a reduced Data Fig. 4c,d). The affinity of the binding site inside the pore was in affinity for ATP at the exterior binding site (Extended Data Fig. 4f). Nature Structural & Molecular Biology | Volume 30 | June 2023 | 761–769 765 Article https://doi.org/10.1038/s41594-023-00984-y To obtain additional details on the binding mode of ATP, we con- GAP led to a stepwise dissociation of MANT-ATP with a half-maximal ducted unrestrained MD simulations of up to 2 µs in duration, until inhibitory concentration (IC50) value of ~10 mM for OEP21-WT and a stable binding pose was reached (Extended Data Fig. 5a, left, and ~26 mM for OEP21ΔL5 (Fig. 3j). Because Mg2+ is released from the Supplementary Methods). This unbiased MD approach showed that chloroplast thylakoid lumen into the stroma under light conditions29, two basic binding poses can be adopted, one inside the β-barrel and and Mg2+ transporters have been reported in the chloroplast IE30, it is the second at the peripheral binding site (Extended Data Fig. 5a, right). very likely that elevated concentrations of both GAP and Mg2+ in the In the internal binding pose, ATP has a multitude of bonding interac- IMS are present. The simultaneous addition of GAP and Mg2+ to the tions with OEP21, whereas fewer contacts are formed in the peripheral OEP21–MANT-ATP complex decreased the IC50 values for both OEP21 location, corroborating the observed differences in affinity (Extended constructs to ~1 mM. This behavior, together with the weaker binding Data Fig. 5b). The two binding poses are in excellent agreement with affinity of ATP in presence of Mg2+, indicates that GAP and Mg2+ syner- the experimental NMR CSP pattern of OEP21 with ATP (Fig. 3d). How- gistically facilitate ATP dissociation from OEP21. Inorganic phosphate ever, the MD data and the observed global NMR effects suggest that also dissociated the OEP21–ATP complex with an IC50 value of ~10 mM, more than just a single binding pose might be possible for each site. To again with a lower value for OEP21-WT than for the ΔL5 variant (Fig. 3i,j). investigate the binding specificity and the impact of positive charges Such a high phosphate concentration lies within what has been in OEP21, we produced 14 point variants in regions identified by the observed in plant cells (1 to 10 mM)31. In comparison, the relative high NMR titrations, as well as the ΔL5 variant (Extended Data Fig. 5c). All GAP concentrations required for ATP dissociation can be rationalized OEP21 variants were properly folded and showed comparable stability by partial unspecific binding to the detergent micelle, as probed by (Extended Data Fig. 5d,e). Next, we probed the binding affinity of these one-dimensional (1D) NMR (Extended Data Fig. 7b). As evident from variants by fluorescence polarization (FP) experiments with MANT-ATP. our ITC experiments and literature reports32, Mg2+ has a strong affinity As shown in Figure 3e, the KD values for ATP determined with this set for ATP (50 µM) but interacts only weakly with GAP or phosphate (KD, of proteins all lie within 50% of the value obtained with the wild-type ~9 mM each) (Extended Data Fig. 7c). These data suggest that phos- (WT) protein. This implies that the removal or shifting (with the R66A phorylated metabolites utilize the same positively charged binding N164R variant) of individual positive charges can be compensated by surface of OEP21 and that dissociation of the high-affinity binder ATP other positively charged side chains in close proximity. Furthermore, can be achieved at cellular solute concentrations. deletion of L5 does not affect the high-affinity binding site inside the pore. These data suggest a promiscuous binding mechanism that is Size-selective metabolite translocation through OEP21 largely dependent on the bulk electrostatic properties of the OEP21 To gain further insights into the metabolite translocation pathway pore. To further validate this conclusion, we performed ATP-binding across OEP21, we conducted MD simulations, where GAP was initially assays at increasing NaCl concentrations (Fig. 3f, left) and observed placed at the entrance of the pore in the IMS. To facilitate translocation an almost 120-fold drop in binding affinity at 2 M salt. Mg2+ is associ- within a 3-µs simulation time, we applied a 180 mV membrane potential ated with ATP in the cell and leads to a reduction in its negative charge with a positive pole at the cytosolic side. The obtained trajectory shows density. Consequently, binding assays in the presence of increasing that GAP is hopping along patches of positive charges on one side of MgCl2 concentrations indicated an almost 50-fold decrease in affinity the barrel and eventually sticks to the exterior binding site involving at 50 mM MgCl2 (Fig. 3f, right). L5 (Fig. 4a and Supplementary Video 1) that provides two additional Next, we used NMR to characterize the interaction between OEP21 positively charged residues, Lys76 and Arg79 (Fig. 4b). Finally, GAP and GAP, the primary product of photosynthesis. Owing to the weaker is released by an outward movement of L5. The interaction with L5 affinity than for ATP, the addition of GAP induced line broadening of slows down translocation, which is supported by a simulated translo- the NMR signals in the β-barrel structure (Extended Data Figs. 6a–c) cation frequency of GAP that is about ten times higher with OEP21ΔL5 in a concentration-dependent manner. The amino acids that experi- (Extended Data Fig. 8). In line with the MD data, the amide hydrogen ence the most pronounced line broadening effects are located at the exchange rates in L5 are reduced in the complex with GAP and, more side of the β-barrel that is involved in dimerization (Extended Data prominently, with ATP (Extended Data Fig. 9a–d), confirming that Fig. 3a,b) and which has the highest density of positively charged resi- L5 interacts with negatively charged metabolites. MD simulations of dues (red spheres in Fig. 3g). To exclude nonspecific binding effects of OEP21 in the apo state show that L5 can transiently cover the pore in a GAP with the LDAO micelle as a reason for the observed line broaden- µs time scale, which was confirmed by the observation of NMR chemi- ing, we recorded 2D-NMR spectra with the unrelated bacterial β-barrel cal shift perturbations between OEP21-WT and OEP21ΔL5 only in close membrane protein OmpX and did not observe any effect, even at a GAP proximity to L5 (Extended Data Fig. 9e–h). concentration of 5 mM (Extended Data Fig. 7a). Next, we performed Finally, we experimentally investigated the channel functionality unrestrained MD simulations with GAP to explore its possible binding by metabolite translocation assays using OEP21 proteoliposomes. poses with OEP21 (Supplementary Methods). As expected, GAP binds For this purpose, metabolite-filled liposomes were subjected to to positively charged patches inside the pore but also interacts with size-exclusion chromatography (SEC), and the loss in metabolite con- cytosolic L5 (Extended Data Fig. 6d, bottom), a finding that is corrobo- tent was quantified by comparing liposomes with and without increas- rated by an increase in the NMR signal intensity of residues in L5 upon ing amounts of OEP21 (Fig. 4c). In the deadtime of the assay, which was the addition of GAP (Extended Data Fig. 6b, bottom). To quantify this ~4 min, GAP and ATP both passed through the channel (Extended Data finding, we next used NMR to derive affinities of GAP with the OEP21 Fig. 10a), even if both molecules were added at the same time. GAP β-barrel, L5 and OEP21ΔL5 (Fig. 3h). These data show that L5 is involved translocation assays in the presence of 5 mM ATP inside and outside the in GAP binding and causes approximately a fourfold increase in bind- liposomes show that GAP transport is slightly enhanced by ATP (Fig. 4d ing affinity. Furthermore, the addition of Mg2+ decreased the binding and Extended Data Fig. 10b), suggesting an activating, rather than an affinity to a similar extent. Residues in L5 interacted with GAP with lower inhibitory, role for ATP. As proposed by MD simulations, deletion of affinity (~3 mM). As suggested by the lower binding affinity (Fig. 3c), the L5 slightly increased the translocation activity of OEP21, leading to a addition of phosphate did only cause NMR chemical shift perturbations lower metabolite content than with the WT protein (Fig. 4d). The addi- but did not alter the intensity of residues in L5 (Extended Data Fig. 6e,f). tion of MgCl2 or elevated concentrations of NaCl to the transport assay Next, we aimed to explore whether metabolites can bind to OEP21 (Fig. 4d and Extended Data Fig. 10c) resulted in reduced GAP transloca- in a competitive manner. We performed FP experiments in which tion, in accordance with a reduced binding affinity (Fig. 3f,h). A similar GAP, GAP and Mg2+, or phosphate was added to a preformed complex behavior was observed for the translocation of ATP, where deletion between OEP21 (WT or ΔL5) and MANT-ATP (Fig. 3i). The addition of of L5 increased and the addition of Mg2+ decreased the translocation Nature Structural & Molecular Biology | Volume 30 | June 2023 | 761–769 766 Article https://doi.org/10.1038/s41594-023-00984-y b Loop 5 a 6 5 Cytosol R79 K76 4 3 E149 R84 GAP 2 R66 R121 1 IMS Molecular weight d e f c 100 100 No translocation 100 0 Control 1,688 Da