In-Situ Flexibility of Both Redox-States of the Chloroplast Regulatory Protein CP12

In the chloroplast, Calvin-Benson-Bassham enzymes are active in the reducing environment imposed in the light via the electrons from the photosystems. In the dark these enzymes are inhibited, and this regulation is mainly mediated via oxidation of key regulatory cysteine residues. CP12 is a small protein that plays a role in this regulation with four cysteine residues that undergo a redox transition. Using amide-proton exchange with solvent measured by nuclear magnetic resonance (NMR) and mass-spectrometry, we confirmed that reduced CP12 is intrinsically disordered. Using real-time NMR, we showed that the oxidation of the two disulfide bridges are simultaneous. In oxidized CP12, the C23-C31 pair is in a region that undergoes a conformational exchange in the NMR-intermediate timescale. The C66-C75 pair is in the C-terminus that folds into a stable helical turn. We confirmed that these structural states exist in a physiologically relevant environment that is, in cell extract from Chlamydomonas reinhardtii. Consistent with these structural equilibria, the reduction is slower for the C66-C75 pair compared to the C23-C31 pair. Finally, the redox mid-potentials for the two cysteine pairs differ and are similar to those found for phosphoribulokinase and glyceraldehyde 3-phosphate dehydrogenase, that we relate to the regulatory role of CP12.


Introduction
Redox regulation based on disulfide-dithiol exchanges constitutes a rapid and reversible post translational modification (PTM) that affects protein conformation. PTM in intrinsically disordered regions (IDRs) can confer flexibility or in contrast order, thereby allowing quick response to a stimulus for a rapid regulation process in cell. The proteins with disorder triggered by different stimuli (redox, pH, temperature…) are categorized as conditionally disordered proteins (CDPs) [1]. PTMs on CDPs can contribute to the diversification and functionality of proteomes, by regulating different properties of proteins and this is termed "proteoform concept" [2]. Among these CDPs, some are particularly sensitive to redox changes, and these have been termed redox-dependent CDPs [3]. In these CDPs, key cysteine residues can be either involved in disulfide bridge or not, depending on the redox conditions. Among the twenty natural amino acid residues, cysteine was believed to be the more order promoting residue, however many examples proteins containing IDRs, or intrinsically disordered proteins (IDP) have been described in the literature that contain redox-sensitive cysteine  [30,31]. The complex Trx network is derived from Cejudo et al [15]. The sulphur atoms of the two pairs of cysteine residues are in yellow. Trx and Fdx stand for thioredoxins and ferredoxin respectively.
The objective of this study was to monitor the structural and redox properties of both cysteine residues pairs depending on the physical-chemical conditions. In order to understand better the redox-dependent conformational sampling of CP12 in vitro, the amide proton exchange rates were measured by mass-spectrometry and Nuclear Magnetic Resonance (NMR), as well as the temperature and pH variation of the NMR observables. The conformational states of both cysteine residues pair regions in CP12ox and CP12red were also observed into more physiological conditions, in C. reinhardtii cell extracts. We thus confirmed the distinct dynamical nature of both cysteine residues pair regions in-situ with all the possible interacting partners, PTM mediators, or solubilizing molecules present in the cell extract. Since, as mentioned above, it is the CP12 redox transition that governs the GAPDH and PRK regulation, we monitored in-real time the redox transition of CP12 in isolation and within C. reinhardtii cell extract, and determined the oxidation-reduction midpoint potentials of its two pairs of cysteine residues.

Amide-water proton exchange kinetic measurements by NMR
The kinetics of amide-water proton exchange was measured on CP12red (in the presence of 20 mM DTTred) and CP12ox in 50mM NaPi pH 6.5, 50mM NaCl, 1mM EDTA, 5% D2O, with traces of sodium trimethylsilylpropanesulfonate (DSS) by NMR using the CLEANEX-PM [39] pulse sequence on a 600MHz Advance III spectrometer equipped with a cryoprobe, pH 7.0 and pH 6.0 and at 283K, with the following spin-lock delays (τm in ms): 1.6, 3.2, 6.5, 9.7, 20, 39 and 81 and as a pseudo-3D experiments. The signal integral for each residue (Vτm) was normalized against that in a fast-HSQC spectrum (V0) acquired using the same parameters, that is a 15 N acquisition time of 42ms, a 1 H acquisition time of 243 ms, and 64 scans. The spectra were transformed in nmrPipe [40], and the crosspeak integrals measured using NMRFAM-SPARKY [41]. The increase of resonance integral as a function of the spin-lock delay (τm) was fitted to the following equation using a Levenberg-Marquardt nonlinear regression function in Octave [42]: where R1B,app is the apparent water relaxation rate and is set to 0.6s -1 , and k (the water-amide exchange rate) and R1A,app (the amide relaxation rate) are optimized. The fits are shown in Sup. Fig. 1. The exchange rate (kex) is set to k/0.85 to take into account the water saturation effect in the CLEANEX-PM experiment [39,43]. The uncertainties on kex values are obtained by repeating the optimization upon randomly varying the experimental integrals within 10% error (which is an overestimation of the experimental error). The intrinsic rates of exchange (kint) were predicted using Sphere (https://protocol.fccc.edu/research/labs/roder/sphere/) [44].

Gibbs-free energy derived from protection factors
The protection factor for CP12ox (P) is calculated with: The Delta Gibbs free energy for the backbone structure of oxidized CP12 is calculated from P with the following equation: with R the gas constant (1.987 10 -3 kcal.K −1 .mol −1 ) and T the temperature (293 K).

Amide-water proton exchange kinetic measurements by mass-spectrometry
Mass-spectrometry was also used to monitor the rate of exchange of proton amide with deuterated solvent. CP12ox and CP12red in 15 mM Tris pH 6.6, 50 mM NaCl were diluted 10x in either 10 mM potassium phosphate in 100% H2O, pH 7.0 for control "undeuterated" experiments, or 10 mM potassium phosphate in 99.96% D2O, pD 7.0 for the "deuterated" experiments. pH values of D2O solutions were adjusted to the corresponding pD values using the equation (pD = pHread + 0.40). Final concentration of CP12ox and CP12red was 17 μM and 20 μM, respectively. The samples were left from 10 s to 15 min for the exchange to occur. At different times, the exchange reaction was quenched by diluting twice in pre-chilled 100 mM potassium phosphate, pH 2.66 containing 2 mM Tris(2carboxyethyl)phosphine (TCEP), and immediately injected on a nanoACQUITY UPLC™ system with HDX technology (Waters Corporation, Milford, MA, USA). The proteins were online digested at 20°C in an immobilized pepsin column (2.1 × 30 mm, Applied Biosystems, CA, USA) for 5 min in 0.1% formic acid/ H2O at a flow rate of 100 µL/min. Peptides were subsequently trapped and desalted online using an ACQUITY UPLC® BEH C18 VanGuard™ Pre-column (1.7 μm, Waters) at 0•C, then eluted into an ACQUITY UPLC® BEH C18 column (1.7 μm, 1 mm × 100 mm, Waters) held at 0 •C, and separated with a linear acetonitrile gradient containing 0.1% formic acid. Mass spectra were acquired on a SYNAPT-G1 mass spectrometer (Waters) with an electrospray ionization source and lock-mass corrected by a Glu-fibrinogen peptide solution, in MSe mode, over the m/z range of 50-2,000. Undeuterated peptides were identified using ProteinLynx Global Server software 3.1 (Waters). Deuterium uptake data for each peptic peptide from the "deuterated" experiments were automatically calculated using DynamX 2.0 software (Waters) and the results were manually checked.

Temperature dependence of the NMR chemical shift and signal intensity
1 H-15 N fast-HSQC of CP12ox (400 μM) in 50mM NaPi pH 6.5, 50mM NaCl and with 20 mM DTTox, 5% D2O were recorded at varying temperature (277K, 284K, 291K, 298K, 305K, 313K, 319K), with a proton acquisition time of 243.3 ms and a 15 N acquisition time of 42 ms on a 600MHz and a 900MHz Advance III spectrometer equipped with cryo-probes. When mentioned, one equivalent of copper (CuSO4) was added to the sample to catalyze oxidation, followed by the addition of 1 mM EDTA and dialysis. The samples recorded at varying pH were recorded in 20 mM Tris, and pH was increased by the addition of NaOH. The spectra were processed with nmrPipe [40] with Sine Bell window function. The referencing of the spectra were done using the frequency of water (carrier frequency) at these temperature [40], and controlled via the chemical shift of DSS, with uncertainty bellow 0.05ppm. The chemical shift of all CP12 amide protons, as well as the signal intensity were determined in NMRFAM-SPARKY [41], and exported in Octave [42]. The chemical shift dependence with temperature was fitted with the following equation: where c is a constant, T the temperature in K.

Thermodynamic of the redox transition
The redox mid-potentials for the transition for each pair of cysteine residues were probed by NMR as proposed in [45]. 50 μM of CP12 in 50mM NaPi pH 6.5, 50mM NaCl, pH 7 were prepared with addition of 20 mM DTT with a varying ratio of DTTox and DTTred in a glovebox under anaerobic atmosphere. The samples were left overnight at ambient temperature to reach the equilibrium. The ratios of DTTox and DTTred were controlled by NMR using the proton signal intensity at 2.85 and 3.05 ppm (DTTox) and at 2.6 ppm (DTTred). 1 H-15 N fast-HSQC of each sample were recorded. The electropotential for each sample was calculated from the ratio of the concentrations of DTTox over DTTred using the Nernst equation [45]: where EDTT, pH7 is the redox mid-potential of DTT (-332 mV [45,46]), R is the gas constant (1.987 10 -3 kcal.K −1 .mol −1 ) and T is the temperature (283 K), n is the number of electrons (2) [41]. The signal intensity of the reduced resonance as a function of the electropotential was fitted in Octave using the following sigmoidal function [46]: The signal intensity of the oxidized resonance as a function of the electropotential was fitted using the following sigmoidal function [46]: The same experiment was repeated at pH 8 in 30mM Tris, 50mM NaCl. The redox potential of DTT is expected to change as a function of pH using the following equation [45]: (8) and was modified consequently in equations 5, 6 and 7.

Real-time monitoring of the reduced-to-oxidized transition
20 mM DTTred was added to 1 mM CP12 (50mM NaPi pH 6.5, 50mM NaCl, pH 7, 5% D2O), the sample was left overnight to ensure complete reduction of its cysteine residues. The absence of oxidized cysteine was monitored by the acquisition of a 1 H-15 N HSQC. The reducing agent was then removed by a PD10 column, followed by dialysis (using 10 kDa cut-off vivaspin concentrators) with a buffer to which air was bubbled for one hour to oxidize the buffer. A series of 1 H-15 N fast-HSQC was then recorded at 293 K. All spectra were transformed in nmrPipe [40], and the signal intensity at frequency of reduced C23, C31, C66, C75 and oxidized C66 and C75 were obtained using the autoFit.tcl script in nmrPipe [40]. The oxidized C23 and C31 resonances are broadened beyond detection [31]. The intensities were then transferred into Octave [42], and the decrease in intensity of the reduced resonances as a function of time were fitted with the following equation using Levenberg-Marquardt nonlinear regression function: where c is a constant, k the rate constant of the oxidation of the cysteine residues pair. The increases in intensity of the oxidized resonance as a function of time were fitted with the following equation: where c is a constant, k the rate constant of the oxidation of the cysteine residues pair.

2.7.
Real-time monitoring of the reduction 20 mM DTTred was added a 400 μM CP12ox (50mM NaPi pH 6.5, 50mM NaCl, 1mM EDTA, 5% D2O). A series of 1 H-15 N fast-HSQC was then recorded at 293 K. The spectra were analyzed, and the data were fitted as above. The decrease of oxidized resonance intensity as a function of time was fitted with the following equation: where c is a constant, k is the rate constant of reductions of disulfide bridges.
The increases in intensity of the reduced resonance as a function of time were fitted with the following equation: 0 ⁄ = c − e -k ox→red t (12) where c is a constant, k is the rate constant of reductions of disulfide bridges.

Cell extract preparation
C. reinhardtii CC124 cells were cultured at ambient temperature (22 °C), with a light/dark cycle, 110 rpm rotation for five days until they reached an absorbance at 680 nm of 1.24 (18.10 6 cells.mL -1 ). 50 mL of culture were centrifuged at 2000 g for 15 min at 4°C and resuspended in 1 mL of buffer 30mM Tris pH 7.0, 20mM DTTred, protease inhibitor Complete EDTA free (Roche) following the recommended concentration. The sample was again centrifuged 2000 g for 15 min at 4°C, and resuspended in 300 μL of the same buffer and D2O was added to the sample. The sample was sonicated twice one min on ice. 15 μL of CP12red was added to reach a final concentration of 50 μM. The total protein concentration in the sample was measured using Bradford assay [47]. 1 H-15 N fast-HSQC of the sample was recorded at 283 K, with proton acquisition time of 243 ms, and nitrogen acquisition time of 42 ms, on a 600 MHz Advance III NMR spectrometer equipped with a cryo-probe. The data were transformed in nmrPipe [40] and signal intensity for each resonance was acquired in NMRFAM-Sparky [41].
For the oxidized sample, the same procedure was applied with the following modification: C. reinhardtii cells were left 24 hrs in the dark before collection. 20 mM DTTox was added to the sample instead of DTTred, as well as 0.1 mM 3-(3,4-dichlorophenyl)-1,1-dimethylurea (DCMU) (Durion). 15 μL of CP12ox was added to reach a final concentration of 50 μM.

Diffusion coefficient determination
In the presence of cell extract, the translational diffusion coefficients of CP12 were recorded using DOSY-NMR selectively on the 15 N labelled protein using the heteronuclear stimulated echo experiment (XSTE) from Ferrage et al [48]. Ten experiments were recorded in a pseudo-2D fashion using bipolar square gradients of 1.4 ms (δ) of strength varying from 2 to 98% of the maximum gradient strength (G, G100% = 0.5146 T.m -1 ) with a fixed diffusion delay of 200 ms (Δ).
In the absence of cellular extract, the same experiment was used to probe for the translational diffusion delay of isolated CP12, and the result was compared with that obtained using standard bipolar stimulated echo experiment (STE) diffusion experiment using the same parameter. The translational diffusion coefficients were identical within 4% uncertainty.
The date were processed in nmrPipe [40] and fitted to the Stejskal-Tanner equation in Octave [42]: where δ, Δ and G are the gradient duration, diffusion delay and gradient strengths defined above. γH is the proton gyromagnetic ratio (2.67 10 8 rad.s -1 .T -1 ). The hydrodynamic radius associated with the diffusion coefficient was determined using the Stokes-Einstein equation: where kB is the Boltzmann constant, T the temperature in Kelvin and η the viscosity.

Determination of delta-Gibbs free energy of binding
The delta-Gibbs free energy for binding of CP12ox to its partner is determined from the following equation: where R is the gas constant (1.987 10 -3 kcal.K −1 .mol −1 ) and T is the temperature (283 K). The dissociation constant for the CP12-PRK complex is 1.3 μM, for the CP12-GAPDH complex is 0.4 nM and for the PRK-(CP12-GAPDH) complex is 60 nM [23].

Structural transition of the region encompassing the C66-C75 disulfide bridge upon oxidation of the isolated protein.
We first investigated the dynamics of CP12red by measuring the amide proton exchange rate with H2O by NMR on CP12red to probe for the protected and exposed backbone amide protons. The measured kex rates were related to very fast amide-water exchange (on the order of 6.10 -2 to 3 s -1 ) and were only slightly slower than the predicted ones using SPHERE (Figure 2a, Figure S1 and S2), in a quasi-uniform manner. These data confirm that in its reduced state, CP12 is highly disordered [30] with amide proton exposed to the solvent, and the rate-limiting step of the exchange is the intrinsic proton exchange rate [49].
Upon oxidation, only a small subset of protons was protected as compared to the reduced state; they are all in the region that surrounds the C66-C75 disulfide bridge. Residues E63 to D68 and E74 to Y78 for which the exchange rates were beyond our detection limit, and A69 and A72 for which exchange rates showed proton exchange rates significantly lower than in the fully disordered CP12red. Assuming that the rate limiting step for this exchange is the unfolding of the stable helical turn, these exchange rates can be related to protection factors ( Figure 2b). These protection factors are in good agreement with our previously described C-terminal structure [31], with the most protected amide being within the two small helical structures connected by the disulfide bridge ( Figure 2c).
Interestingly, the variation of proton chemical shift according to the temperature for the same set of CP12ox residues (E63, F65, K67, D68, A69, A72, C75 and R76) had slope more positive than -4.6.10 -3 ppm.K -1 (Figure 2d, Figure S4), which is also indicative of hydrogen bond [50]. Together, these data confirm that the C-terminal region surrounding the C66-C75 disulfide bridge folds in a helical turn stabilized by H-bonds upon oxidation [31].  [44] (black) and solvent-amide proton exchange rates measured using CLEANEX-PM experiment [39,43] on CP12red (red). The similarity between both values indicated that CP12red is intrinsically disordered. The signal intensity as a function of the spin-lock delay in the CLEANEX-PM experiments is shown in Figure S1. (b) Gibbs-free energy of CP12ox backbone amides derived from the measured solvent-amine proton exchange using CLEANEX-PM on oxidized versus reduced CP12 (refer to Material and Method for details), assuming that the exchange falls into the EX2 regime that is highly probable at this low pH. The measured rates of exchange for CP12ox amide at pH 7 and pH 6 are shown Figure S2. (d) Value of the slope of the temperature dependence of the proton chemical shift for CP12ox residues. The line at -4.6 ppm.K -1 indicates the threshold below which the data indicate the absence of hydrogen bond [50]. The temperature dependence curves of the chemical shift of the amine proton for all CP12ox residues are shown Figure S4. (c) Structure of the stable C-terminal helical turn with the amide protons protected from the exchange with water highlighted in red, and those exposed to exchange in green. (e) On the same structure, the amide proton for which the temperature dependence of the chemical shift has a positive slope highlighted in red.

Structural transition of the region encompassing the C23-C31 disulfide bridge upon oxidation of the isolated protein.
Contrary to the residues of the stable C-terminal region, most residues of the N-terminal region presented a kinetic of HN → H2O exchange similar to the disordered reduced state (Figure 2b, Figure  S1 and S2). Mass-spectrometry measurement of the proton-deuterium exchange on CP12ox also confirmed that these N-terminal amide protons exchange very quickly with the deuterated solvent ( Figure S3). The variation of chemical shift according to the temperature for these resonances had all more negative gradients than -4.6.10 -3 ppm.K -1 (Figure 2d), indicative of more labile protons [50] and this confirms the amide proton-exchange measurements. These data indicated that the N-terminal region surrounding the C23-C31 disulfide bridge remains highly unstable.
The temperature dependence of the proton chemical shift is expected to be linear for a proton in a stable conformational state [51]. For a high number of amide protons in CP12ox the temperature dependence of the NMR frequency of amide proton of residues deviated from linearity, in particular those of neighboring the dynamic N-terminal region; for example the resonances assigned to residues A5, V45 as well as a high field shifted glycine resonance that is putatively assigned to the single glycine of the N-terminal region G26 ( Figure S4). This was also the case for several unassigned resonances, that we have ascribed to the region L8-A43 [31]. Of note, a few residues of the C-terminal turn also presented a non-linear temperature dependence of their amide proton chemical shift (L62, E63, A69, A72, R76, Figure S4).
The curvature in the temperature dependence of the chemical shift indicates a chemical exchange between two forms or more for which the relative population varies with temperature [51]. In line with these observations, two resonances were observed for the side chain indole amide of the single tryptophan residue at position 35 ( Figure 3). One of these two resonances overlaid with that of CP12red Trp-Hε (Figure 3a). The 1 H chemical shift of this resonance had a linear dependence with a negative slope of -3.3 10 -3 ppm.K -1 (Figure 3c), which is close to the threshold for disordered amide [50]. To ensure that this resonance, assigned to unfolded Trp-Hε, did not rise from a fraction of reduced proteins, Cu 2+ was added to the sample to fully oxidize the protein and then removed before NMR acquisition. The spectrum remained identical (Figure 3b), indicating that the unfolded Trp-Hε resonance belongs to CP12ox. The chemical shift temperature dependence of the second resonance had a positive slope (+8 10 -3 ppm.K -1 ), indicative of folded conformation. Interestingly, the difference in frequency between these two resonances assigned to the same proton decreased at high temperature, and this relates to an increase in intensity (and decrease in linewidth) for the folded Trp-Hε resonance (Figure 3c, Figure S5). The difference in frequency also decreased at higher pH (Figure 3b), and the resulted resonances at pH 8 and 9 fell in a linear chemical shift pattern defined by the two above mentioned unfolded and folded Trp-Hε resonances [52,53]. This behavior is strongly indicative of a chemical exchange in the intermediate timescale (μs-100 ms) between a disordered and a folded conformation, that becomes faster at higher temperature and at higher pH [54].  15 N spectra of CP12ox recorded at varying temperatures (277 K, 284 K, 291 K, 298 K, 305 K, 313 K, 319 K) and at pH 6.5 and plotted with the same contour levels. To ease the reading, the contour levels of each resonance are colored differentially. In (a) and (b), above the series, the spectrum of CP12red at 284 K is shown (magenta). (b) 1 H-15 N spectra of CP12 recorded at 284K at varying pH (pH 6.5 as in A, pH 8 and pH 9). An additional spectrum is shown at pH 6.5 where the protein has been pre-treated with Cu 2+ to ensure full oxidation. The paramagnetic ion was removed by dialysis before NMR acquisition. (c) Temperature dependence of the proton chemical shift for each of these CP12ox Trp-ε resonances using the same color coding ( Figure S4). (d) Temperature dependence of the resonance intensity for each of these CP12ox Trp-ε resonances using the same color coding (refer to Figure S5).

Structural properties of CP12red in the presence of C.reinhardtii cell extract.
CP12 is located in a highly dense organelle, the chloroplast, and we aimed at grasping the effect of the presence of the macromolecules of its physiological environment on the structural states of CP12 in its reduced (light) and oxidized (dark) states. To mimic the chloroplast environment, we thus added crude cell extract at high concentration (15 to 19 mg.mL -1 of proteins) to purified 15 N-CP12 such that the observed CP12 proteins represent 2 to 3% of the total mass of protein. On the reduced and disordered protein, there was no significant chemical shift displacement, indicating that the protein remains in a disordered state (Figure 4b), with the exception of the last two resonances E79 and D80. The resonances of the N-terminal residues up to K17 including the His-tag were broadened beyond detection compared to the isolated protein (Figure 4d). Other resonances were broadened in the presence of cell extract compared to the isolated protein: E40 and L41, K48 and K49, as well as few resonances of the C-terminal region (D70 to D72, C75 and the last two residues E79, D80). On the contrary, the region C31-A34 did not present any chemical shift variation, or specific line-broadening in the presence of cell extract. The translational diffusion coefficient at 4°C of reduced 15 N-CP12 in the presence of cell extract was less than 2 10 -11 m 2 .s -1 that is significantly lower compared to that of the isolated reduced protein (9.3 ± 0.15 10 -11 m 2 .s -1 , Figure S6). The 1 H-15 N HSQC spectrum of oxidized and partially folded CP12 in the presence of cell extract also remarkably resembles that of the isolated protein ( Figure 5b). As for CP12red, the His-tag and the N-terminal residues (up to D14) resonances were also broadened beyond detection in the presence of cell extract (Figure 5d). On the contrary, the C-terminal region (from K53 onwards) presented relatively narrow linewidths and identical chemical shift as compared to the isolated protein. The resonances for the residues A15 to S42 were broadened beyond detection both in the isolated protein and in the presence of cell extract. On the isolated protein, resonance linewidths were large up to residues S42 but in the presence of cell extract resonance linewidths were large for all residues until D50. The translational diffusion coefficient at 4°C of oxidized 15 N-CP12 in the presence of cell extract was 4 ± 0.5 10 -11 m 2 .s -1 that is significantly lower compared to that of the isolated oxidized protein (9.8 ± 0.2 10 -11 m 2 .s -1 , Figure S6).

Thermodynamical properties of the redox transition of both disulfide bridges in isolated CP12.
CP12 is a CDP, and its function is related to the significant structural transition upon reduction or oxidation of the two cysteine residues pair. From the above results, we showed that the two disulfide bridges are located into two structurally distinguishable regions separated by a flexible linker. We monitored the thermodynamical properties of these two disulfide bridges using a titration with the redox mediator DTTox/DTTred [45,55] on isolated CP12. Surprisingly, the use of the wellknown redox mediator GSSG/GSH was not possible because the redox transition was not reversible with this mediator [45]. The standard redox mid-potentials at pH 7.0 were -284 mV for the C23-C31 cysteine pair (Figure 6a) and -291 mV for the C66-C75 pair (Figure 6b). The variation of the standard redox mid-potential of the N-terminal pair followed the expected pH-dependence for a disulfide bridge (-59.1 mV per pH unit [45], Figure 6a). On the contrary, the redox mid-potential measured at pH 8 for the C-terminal disulfide bridge was more negative than expected (Figure 6c), indicating a possible pH-induced stability for the C-terminal folded oxidized state.  Figure S7. (e) Real-time monitoring of the reduction. The color coding is the same as above. The fits using the equations I I 0 ⁄ =e -k ox→red t +c and I I 0 ⁄ =c-e -k ox→red t are shown, where c is a constant, k is the rate constant of the transition from oxidized to reduced cysteine residues. The reduction rate for all CP12 residues is shown in Figure S8.

Rate of oxidation and reduction of both disulfide bridges in isolated CP12, probed by real-time NMR.
Using NMR spectroscopy, we followed the oxidation or reduction in real-time of each disulfide bridge of isolated CP12 upon changing the redox potential of the solution by buffer exchange into air-aerobic buffer or upon addition of 20 mM DTTred. The rates of the oxidation for both disulfide bridges were identical, indicating their synchronisation (Figure 6d, Figure S7). On the contrary, the reduction of the N-terminal bridge was much faster than that of the C-terminal bridge, indicating asynchrony in their reduction (Figure 6e, Figure S8). These rates measured on the isolated protein are not representative of those in the cell, where redox transitions are catalyzed by the complex Trx network but reflect the structural stability of both regions encompassing the two respective disulfide bridges.

Reversible redox transition of CP12 in the presence of C. reinhardtii cell-extract probed by real-time NMR.
Upon addition of crude cell extract from cells exposed to 24 hours of dark, the 15 N-CP12ox protein was highly unstable, and its 1 H-15 N spectrum was quickly converted and overlaid that of CP12red within 20-40 min, even when 20 mM DTTox was added. This indicates that the electron transfer of the photosynthesis was active in these crude cell extract and can activate the Trx network resulting in the reduction of DTT and CP12 when cells were transferred from dark to the spectrometer. We therefore added 0.1 mM DCMU that inhibits photosystems II and electron transfer reactions, and this prevented CP12 reduction.
We monitored in real time the redox transition triggered by strong reducing or oxidizing agents (in the presence of PSII inhibitor not to interfere with the Trx network). When dark extracts were exposed to strong reducing conditions (20mM Na2S2O2), the 1 H-15 N spectrum of the protein was quickly converted and overlaid that of CP12red (Figure 4c). Reciprocally, when light extracts were exposed to oxidizing conditions (20mM H2O2) the 1 H-15 N spectrum of the protein was quickly converted within 20 min and overlaid that of CP12ox (Figure 5c). Altogether, these results show that both oxidized-to-reduced and reduced-to-oxidized transitions of CP12 are very fast in C. reinhardtii cell extract, contrary to isolated CP12, very likely thanks to the presence of TRXs in the extracts.

Discussion
CP12 is present in the chloroplast of photosynthetic organisms from cyanobacteria, green algae and plants, red algae as well as heterokonts [27,28,32]. Various functions have been assigned to this protein, in particular the dark-down regulation of two CBB enzymes: GAPDH and PRK, as well as their fast re-activation upon dark-to-light transition [22]. Besides, genetic and proteomic studies on higher plants and heterokonts report results that suggest that CP12 has other functions, for example related to stress regulation [32,[56][57][58][59]. The enigmatic and multiple functions of CP12 is not an unusual property for a protein belonging to the CDP family [1]. In this study, we report on the structural transition of CP12 at the molecular level occurring upon change in redox potential, which is the basis for its proposed regulatory functions.

Isolated CP12red is intrinsically disordered
In the light, the photosystems I and II impose a strong reducing potential that is mediated to the chloroplastic proteins via the Fdx-dependent Trx [60], and this network is probably the best studied within the chloroplast [61]. Under reducing conditions, we recorded the amide protons exchange rate with solvent by NMR and mass spectrometry for the purified CP12 and these data confirmed that the unfolded reduced state of CP12 is highly flexible (Figure 2a), as we have described previously by SAXS and NMR spin-relaxation experiments [30,31]. The absence of pre-formed motif is also confirmed in this study by the homogeneous NMR resonance intensity for all CP12red residues ( Figure  4d). Indeed, the existence of a small proportion of pre-formed motifs would give rise to selective broadening of the NMR resonances [30,62]. Together, these data confirm that CP12red is intrinsically disordered (Figure 7a). The midpoint redox potentials for the CP12 C-terminal disulfide bridge (blue filled circle) and Nterminal disulfide bridge (green filled circle) measured in this study are compared to the midpoint potential for A. thaliana CP12 isoforms (blue and green squares), as well as for A2B2-GAPDH, A4-GAPDH and PRK activity as determined in [63][64][65]   In the dark, oxidizing conditions prevail in the chloroplast [61], principally because the source of electron (water splitting) does not fuel the Fdx-dependent Trx system. Oxidation of chloroplast proteins is then catalyzed by both the Fdx-dependent and the NADPH-dependent Trx systems. In the case of CP12, the four cysteine residues form two disulfide bridges. Our results showed that the C-terminal disulfide bridge (C66-C75) is located in a region in which amide proton are highly protected against solvent exchange (Figure 2b). This confirms that this region encompassing residues L62-Y78 is highly stable, as we have previously shown by NMR spin-relaxation measurements [31] (Figure  7b). The N-terminal disulfide bridge (C23-C31) is in a region that remains highly dynamic, as shown here by the absence of protection against amide proton exchange by NMR and MS (Figure 2b, Figure  S3). As shown previously for C. reinhardtii and A. thaliana CP12 proteins, the residues of this long Nterminal region encompassing residues L8-A44 gives rise to broad NMR resonances [29,31]. We used the resonances from the unique tryptophan side chain to probe for the origin of this line broadening. Two resonances are observed (Figure 3). One resonance that is the most down field overlays with that of the same proton in CP12red. Its variation with temperature follows what is expected for a residue in a disordered region [50,51]. Also, this resonance is not observed at pH 9, and this is also indicative of fast amide proton exchange with water, and therefore absence of structure [67]. The second resonance is always observed even at high pH, and varies with temperature with a positive slope, that is indicative of hydrogen bond ( Figure 3) [50,51]. Together, these data confirm our previous SAXS analysis that suggested that the N-terminal region can be either disordered or folded [31]. When oxidizing conditions prevail in the chloroplast, the pH in the stroma decreases to pH 7 [68]. Such a drop in pH induced a displacement of the resonances arising from Trp-Hε within a linear chemical shift pattern defined by the above mentioned two resonances with a larger difference in intensity (Figure 3b). This displacement is accompanied by a decrease in linewidth and an increase in signal intensity. This indicates that the speed of interconversion between the two states is modified at high pH, and becomes slower in NMR intermediate exchange timescale (μs-ms) [52,53,69], as we have previously suggested based on NMR spin relaxation experiments [31].
Altogether, these data suggest that in its oxidized state, CP12 is in an atypical conformational ensemble, in which the C-terminal region (L62-Y78) is highly stable, whereas the N-terminal region (L8-A44) co-exists in a folded and unfolded states in exchange in the ms timescale (Figure 7b). In other word, the C-terminal region of CP12 undergoes a complete disorder-to-order transition upon increasing the redox potential, as described for several redox-dependent CDP [3]. On the contrary, the N-terminal region is not fully ordered in the oxidized state. For this region, the folded state is only partially stabilized under oxidizing condition. Our previous SAXS analysis suggested relative population for the disordered and ordered conformation of 40%:60%, and our current temperature and pH dependent analysis suggest that this equilibrium is strongly dependent on the physicochemical conditions (Figure 7b). These two regions (folded C-terminal helical turn and highly dynamic N-terminal region) are separated by a flexible linker of ten residues (D50-D60) that also delimits two regions of interaction with different partners.

The distinct regions of CP12ox differ in structural dynamic and in affinity for their interacting partners.
CP12ox associates and inhibits PRK and GAPDH in a ternary complex [23]. The structure of this complex has been solved recently for proteins from A. thaliana [26] and from the cyanobacterium Thermosynechococcus elongatus [25]. In these structures, the C-terminal region of CP12ox folds into a helical turn that is identical to the stable structure that we have computed from NMR data [31]. The very high affinity for the C-terminal region of CP12 for GAPDH (sub nM) relates to a low delta Gibbsfree energy of -12.7 kcal.mol -1 , a value that is below the mean value for complexes between two ordered proteins from ~200 studies [70] (Figure 7c). On the contrary, the N-terminal region has a much lower affinity for PRK (-1.3 μM) [23], that relates to a higher delta Gibbs-free energy of -8 kcal.mol -1 . This value of delta Gibbs free energy is identical to the mean value for complexes between an ordered and a disordered protein [70]. The CP12ox binding region with PRK is located in the unstable and highly dynamic N-terminal region [24]. In line with the above described structural dynamic for this region in the isolated protein, in the crystal structure of the PRK-CP12-GAPDH from A. thaliana [26] and in the cryo-EM structure from T. elongatus [25], the least resolved regions are the interaction surface between CP12 and PRK. Altogether, the interface between Nter-CP12ox and PRK seems to be highly dynamic contrary to the Cter-CP12ox interface with GAPDH, and these can be related to the relative structural stability of these two regions.
The two structurally and functionally distinguishable regions of CP12 -a stable C-terminal helical turn and an unstable N-terminal region -fused via a flexible linker offer a complex and tightly controlled mean to regulate the two binding partners. Interestingly, photosynthetic organisms possess CP12 homologs where only the C-terminal region (C66xxxP70xxxC75) is fused to other enzymes such as GAPDH-B isoforms [71] or adenylate kinase, ADK3 [72,73]. These proteins also bind to GAPDH with a high affinity. Besides, other organisms possess CP12 homologs with the conserved N-terminal motif A34WxxVEEL41 that is embedded in the dynamical PRK-binding region defined above [27,32], and the function of these proteins remains to be elucidated.

Thermodynamically independent and reversible redox transition of both disulfide bridges in CP12.
As mentioned above, the principal known function for CP12 is the redox regulation of CBB enzymes in the chloroplast. Here, we report on redox titrations by NMR that demonstrated that the four conserved cysteine residues of CP12 from C. reinhardtii could form two intermolecular disulfide bridges with different redox mid-potentials (Figure 7d). The redox mid-potential for the C23-C31 cysteine residues pair (E pH 8 C.r N-ter 0 = -343 mV) is more negative than those measured for the Nterminal disulfide bridge of A. thaliana CP12 isoforms (E pH 7.9 A.t N-ter 0 = −326mV in the two isoforms CP12-1 and CP12-2 and E pH 7.9 A.t N-ter 0 = −332mV in the isoform CP12-3 [65]), analyzed from redox titrations using 5,5′-dithiobis(2-nitrobenzoic acid) to detect thiol groups. The redox mid-potential for the C. reinhardtii C-terminal disulfide (E pH 8 C.r C-ter 0 −356 mV) is an intermediate with that found for the isoforms CP12-1 & CP12-2 of A. thaliana (E pH 7.9 A.t C-ter 0 = -350mV) and that found for the isoform CP12-3 (E pH 7.9 A.t C-ter 0 = -373mV) [65]. These more negative values in the alga CP12 redox mid-points are in contradiction with the previous observation that the redox potential of Trxs in the alga are less negative than higher plant counterpart [74]. The different technique might also be responsible for the difference. Above all, and similarly with what has been observed for A. thaliana, our results indicate that the N terminal disulfide in C. reinhardtii requires less reducing conditions than the C-terminal disulfide, i.e. in thermodynamic terms, the N-terminal disulfide is easier to reduce than the C terminal disulfide.
During light to dark transition, partially oxidizing conditions would cause the formation of the C-terminal disulfide of CP12 (E pH 8 0 = -358 mV) and then favour the formation of the binary complex A4-GAPDH-CP12 (Figure 7c & d). At these electropotential values, the A4-GAPDH tetramer from A. thaliana (E pH 7.9 A4-GAPDH 0 = -335 mV) is still reduced, but the A2B2-GAPDH tetramer -that contains a CP12 homolog extension is also getting oxidized (E pH 7.9 A2B2-GAPDH 0 = -359 mV) [64]. The redox mid-potential of CP12 C-terminal bridge is consistent with the hypothesis proposed in [17] that CP12 is a redox mediator for GAPDH regulation. By approaching darkness, further oxidation of Trxs would lead to the formation of N-terminal disulfide formation in CP12 (E pH 8 N-ter 0 = -343mV) allowing the PRK-CP12-GAPDH formation, and finally the regulatory disulfide in PRK is able to form ( E pH 7.9 PRK 0 = -333 mV) [63]. These results provide key insights to understanding the biological assembly and the regulation of this ternary complex.

Asynchrony of the reduction of the two disulfide bridges, synchrony of their formation
From a kinetic point of view, the rates of oxidation of the cysteine residues at the N and C terminal extremities were similar, which validates the above discussion on equilibrium. On the contrary, the rates of reduction were different ( Figure 7E). The synchrony of oxidation can be explained by the fully disordered nature of the reduced form. Similarly, the asynchrony of reduction can be explained by the dual nature of CP12ox with a very dynamic N terminal end, being prone to fast reduction, and a stable C terminal helical turn less prone to reduction. On the isolated CP12, these processes, oxidation and reduction, were slow. In contrast, in-situ -that is in CP12 physiological environment with active cellular redox-mediators -the redox transitions were much faster [17]. It would be interesting to compare the rate of thiol reduction in the cell (kinetic of CP12 thiol to disulfide transition) with the rate of reduction of the electropotential in the chloroplast upon dark to light transition [15], as this could provide a further fine regulation mean for CBB enzyme activity.

Effect of cell extract on Nter-CP12 and Cter-CP12
Within the chloroplast, redox modulation is clearly open to modification by a number of other parameters such as stromal pH, Mg 2+ and metabolite concentrations, and above all, crowding. The chloroplast contains numerous phases including starch granules, thylakoid membranes [75], pyrenoid (the liquid non-membranous compartment containing the ribulose-1,5-bisphosphate carboxylase oxygenase) [76,77] as well as natural deep eutectic molecules with an extremely high local concentration [78][79][80]. The physico-chemical conditions prevailing in the chloroplast stroma therefore have to be taken into account when trying to unravel the processes of protein folding. CP12 is known to exist in chloroplast of in A. thaliana both in light and dark conditions [81], we thus aimed at characterizing the effect of the chloroplast environment on both CP12red and CP12ox. In order to approach the physico-chemical condition that CP12 encounters in the chloroplast, we added C. reinhardtii cell extract to purified 15 N-CP12. The cytoplasm of C. reinhardtii is very limited in volume compared to other eukaryotic cells [82], such that chloroplast molecules contribute to a major fraction of the cell extract. Besides, the interaction network of CP12 is expected to be present in cell extract as it has been previously shown that sonication does not break the networks of interactions between partners [83].
We observed the structural state of CP12 in a large excess of C. reinhardtii cell extract (2 to 3% of protein are 15 N-CP12 in a total protein concentration of 15 to 19 mg.mL -1 ). We observed that CP12red in-situ remains disordered, with a few residues being affected. Some residues present specific line broadening of their NMR resonances (for example the N-terminal residues, E40-L41, K48-K49), and this could be ascribed to specific binding of molecule with CP12red. In the literature, no binding partners for CP12red have been identified, and this could be because their affinity are probably weak as this is often the case for disordered proteins [70]. These affinities might be increased in-situ, giving rise to the observed line-broadening. Indeed, molecular crowding is known to increase the equilibrium constants of association events [84,85]. For example, the association equilibrium constant for the dimerization of a 40-kDa monomer is 8 to 40-fold higher if the protein is in E. coli cytoplasm, where the concentration of proteins is extremely high (200 to 320 mg/mL), compared to the same phenomenon in an aqueous solution [86].
In the presence of oxidized cell extract, only the stable C-terminal helical turn is observed in the NMR spectrum ( Figure 5). This could be because the region concerned with line broadening ascribed to the millisecond-timescale equilibrium on the isolated protein (L9-A43) is expanding in the presence of cell extract (N-terminal to D50). CP12ox thus behaves like many proteins that were shown to be more thermodynamically stable in the presence of crowding agents because of excluded-volume effects and higher chances of intra-molecular interactions [87]. Another origin for line broadening could be the association within a macro-molecular complex as discussed above for CP12red.
The translational diffusions of both CP12ox and CP12red are also dampened by the presence of C. reinhardtii cell extract ( Figure S6) and are reduced by a factor of two for CP12ox and by more than 5 for CP12red. The macromolecular concentration in our experiment was 15 to 19 mg/mL of protein, and an increase in viscosity is thus expected. Ordered proteins are expected to be more affected by the increase in viscosity than disordered protein [88,89], but this is not the case for the partially folded CP12ox that is less affected than the disordered CP12red. A succession of association-dissociation events with several chloroplast partners can also explain this very slow diffusion. The presence of broadened resonances in both spectra as discussed above indicates the presence of such associationdissociation events. Surprisingly, diffusion of CP12red is more affected than CP12ox, and this may suggest a higher number of weak (specific or not) interaction events for the disordered reduced state.