X-ray radiolytic labeling reveals the molecular basis of orange carotenoid protein photoprotection and its interactions with fluorescence recovery protein

In cyanobacterial photoprotection, the orange carotenoid protein (OCP) is photoactivated under excess light conditions and binds to the light-harvesting antenna, triggering the dissipation of captured light energy. In low light, the OCP relaxes to the native state, a process that is accelerated in the presence of fluorescence recovery protein (FRP). Despite the importance of the OCP in photoprotection, the precise mechanism of photoactivation by this protein is not well-understood. Using time-resolved X-ray–mediated in situ hydroxyl radical labeling, we probed real-time solvent accessibility (SA) changes at key OCP residues during photoactivation and relaxation. We observed a biphasic photoactivation process in which carotenoid migration preceded domain dissociation. We also observed a multiphasic relaxation process, with collapsed domain association preceding the final conformational rearrangement of the carotenoid. Using steady-state hydroxyl radical labeling, we identified sites of interaction between the FRP and OCP. In combination, the findings in this study provide molecular-level insights into the factors driving structural changes during OCP-mediated photoprotection in cyanobacteria, and furnish a basis for understanding the physiological relevance of the FRP-mediated relaxation process.

Photosynthetic organisms have a protective mechanism known as nonphotochemical quenching (NPQ) 3 to dissipate excess energy and prevent oxidative damage, which would otherwise lead to cell death under high-light conditions (1). Unlike plants and algae, in which the NPQ mechanism involves membrane-embedded protein-pigment interactions within lightharvesting complexes, cyanobacteria use the orange carotenoid protein (OCP) to trigger photoprotection of their extramembrane, water-soluble antenna complexes called the phycobilisomes (PB) (2). Upon absorption of blue-green light, darkadapted OCP (OCP O ) switches to the active, red form (OCP R ), which interacts with the PB and effectively reduces the amount of energy transferred to the photochemical reaction center. The OCP R form is metastable; it relaxes back to OCP O slowly in darkness, but instantaneously in the presence of the fluorescence recovery protein (FRP) (3)(4)(5). The FRP plays an important role in the photoprotective mechanism by detaching the OCP R from PB and accelerating the OCP R to OCP O conversion to restore ambient photosynthesis. Determining the exact structural changes that accompany OCP R formation and its interactions with the FRP is critical for a complete mechanistic understanding of the regulation of NPQ in cyanobacteria. OCP O , which crystallizes as an antiparallel dimer, is composed of an all-helical N-terminal domain (NTD) and a mixed ␣/␤ C-terminal domain (CTD) (6,7). A noncovalently bound carotenoid spans both domains in the OCP O . The NTD and CTD interact through two regions, the major interface, which buries 1722 Å 2 of surface area, and the minor interface, which is the interaction region between the N-terminal ␣ helix (␣A) and the CTD (6,7). The FRP, in contrast, is 13 kDa and entirely ␣-helical. The FRP has been captured in two distinct conformations, with distinct quarternary structures, in a single crystal form (5). The first high-resolution structural information on full-length monomeric OCP R in solution was obtained by the combination of structural methods of X-ray crystallography, X-ray radiolytic labeling, or X-ray footprinting with MS (XFMS), hydrogen/ deuterium exchange with MS, and small angle scattering (8). These results showed that upon photoactivation the chromophore translocates 12 Å into the NTD, and the NTD and CTD completely dissociate from each other. To understand how chromophore movement might be linked to larger conformational changes in the protein, XFMS was further used to pinpoint subtle structural changes in the specific amino acids and structurally conserved water molecules along a potential signal propagation pathway from carotenoid to the protein surface (8). Recently, crystallography has captured light-induced initial structural changes in the dimeric form of the OCP (9). Collectively, these data provided a comprehensive view of global and local intra-protein structural end points at the start and end of photoactivation, but could not delineate the steps along the temporal path during OCP photoactivation and how FRP accelerates the relaxation process.
UV-visible absorption, which is an important analytical tool to characterize OCP O and OCP R , is highly sensitive to the interactions of multiple amino acid side chains with the carotenoid and to the polarity of the carotenoid-binding pocket (7,10). The time course of UV-visible absorption, which occurs in minutes at room temperature, represents the overall kinetics resulting from the change in the carotenoid-protein interaction and/or environment, but cannot delineate specific changes that result/ occur in the protein domain and distant from the carotenoid-binding site (7). Although carotenoid dynamics during photoactivation in the OCP has been studied by ultrafast spectroscopic methods (11)(12)(13), which have revealed details of the electronic structure of short-lived carotenoid excited states and suggested changes in the conformational orientation of the carotenoid, these studies have not provided any direct information on the global structural changes in the protein occurring on a longer time scale during photoactivation or relaxation. Intrinsic fluorescence from tryptophan and other aromatic amino acids also provides an overall picture of the dynamics of photoactivation in seconds but represents an ensemble average of protein conformation resulting from multiple aromatic residues (12,14). Binding assays using size exclusion chromatography together with mutational analysis have indicated that the minor interface and the CTD influence the OCP-FRP interaction and have revealed the variable stoichiometry of the FRP-OCP complex (14,15). These data provide a global view of OCP-FRP interactions and indicate likely surface regions involved in the interactions, but do not pinpoint the specific residue interactions necessary to provide a mechanistic understanding of the recovery of OCP R by FRP for the regulation of NPQ.
XFMS is ideally suited for revealing conformational dynamics at the single residue level. In this method, a high flux density X-ray beam ionizes water to produce hydroxyl radicals in situ and covalently modifies amino acid side chains on the microsecond time scale (16,17). These modifications are analyzed by high-resolution MS to locate the sites and extent of modification. Here, we used time-resolved XFMS to study the OCP photocycle in solution in real time. We captured the progress of OCP R formation from OCP O under constant blue light illumination, and also the reverse process of OCP R relaxation to OCP O under dark conditions by exposing the sample to X-rays at time intervals of seconds to minutes. The residue-specific and time-dependent solvent accessibility (SA) changes provide a mechanistic understanding of the carotenoid-induced signal transfer within the protein under native conditions. Furthermore, we monitored the residue-specific SA changes in the formation of complexes of OCP-FRP and CTD-FRP. We identified a binding site on the CTD domain of the OCP and, in addition, the data indicate that the FRP interacts with the OCP at a second location on the N-terminal ␣ helix. These data support a previous suggestion that the FRP effectively holds together the OCP domains at the major and minor interfaces, and that this forced proximity of the two domains is what leads to the acceleration of OCP R relaxation to OCP O (14). Furthermore, by temporally differentiating SA changes we have identified residue-level events that at least partially constitute the driving force for the global conformational changes that underlie activation and relaxation, and have pinpointed the exact sequence of these structural changes.

Time-evolution of residue-specific conformation changes in OCP photoactivation and relaxation
To follow the activation of OCP O to OCP R , we continuously illuminated OCP O with blue light, and after selected delays, subjected the protein to a 200-s pulse of synchrotron X-rays (17) (Fig. 1A). The microsecond X-ray pulse ionizes both bulk and bound water rapidly and isotropically to generate hydroxyl radicals in situ, which then react with proximal residues to yield stable modification products (18). As photoactivation progresses, XFMS captures sequential snapshots of changing SAs of the various residues in parallel across the protein. Following X-ray exposure, samples were digested and then analyzed by HPLC and MS (LC-MS), which generated Ͼ94% sequence coverage and identified all the side chain modifications reported in our previous studies (8). To follow the relaxation process, preilluminated OCP R in the dark was subjected to microsecond pulses after various selected delay times ( Fig. 2A). For both experiments, the sample conditions were chosen so as to enable comparison with previous steady-state XFMS and complement UV-visible studies at 15°C (Fig. 1B) (8,10). The plot of the fraction of unmodified side chain versus delay time produced residue-specific kinetics traces that show that photoactivation and relaxation followed a different pathway (Figs. 1C and 2A). The time-resolved XFMS strategy thus provided both a qualitative view and quantitative measure of the lifetime of local conformational changes at a high structural resolution and was able to follow these changes at various residues as a function of time during photoactivation or relaxation.

The OCP O photoactivation pathway is concerted
The site-specific kinetic progress curves obtained for the OCP O to OCP R photoactivation process showed a single component or was biphasic in all cases ( Fig. 1C and Fig. S1), and the overall trend of SA changes were consistent with our previous steady-state XFMS experiments (8). Although the time-dependent progress curve consists of a single component, the result can be interpreted by three general categories of conformational transition ( Fig. 1D and Table S1). In the first category, residues Trp-41, Tyr-44, and Met-47, which are located in the carotenoid-binding pocket of the NTD, displayed SA changes on the fastest time scale ( ϭ Ͻ1.5 min). The overall decrease in X-ray footprinting reveals OCP photoprotection  (Table S1) are mapped onto the OCP O structure. Carotenoid is in purple, residues in blue show a SA decrease upon activation, and those in red show an increase. The speed of decrease or increase is indicated by lifetime and the depth of the color. Single letter amino acids are shown in the figures.  (Table S2). Error bars represent the standard deviation from three independent measurements. Below are the same experiments carried out at 4°C, collecting accessibility changes from 0 to 30 min in the OCP R to OCP O transition. B, residues in the two categories are mapped onto the OCP O structure (PDB code MG1). Carotenoid is in purple, residues in blue show a net SA decrease upon activation, and those in red and gray show a net increase and no change, respectively.

X-ray footprinting reveals OCP photoprotection
SA during photoactivation is consistent with the movement of the carotenoid deeper into the binding pocket in the NTD, as observed previously (8), and indicates that rearrangement within the carotenoid tunnel is one of the first structural changes after chromophore activation by light (8,9). In the second category (1.5 min Ͻ Ͻ 3.5 min) were residues at the major interface (Tyr-98, Trp-101, Pro-103, Arg-155, Asn-156, Pro-276, Trp-277, Phe-278, and Met-284) and NTD residues Phe-163, Lys-167, residues 172-185, and Phe-227. Most of these residues are at the major interface, suggesting that domain dissociation is completed after the movement of the carotenoid. In the third or slowest category ( Ͼ 3.5 min) were the NTD ␣A (Ala-8, Arg-9) and its adjacent helical region (13)(14)(15)(16)(17)(18)(19)(20)(21)(22), residues at the minor interface that are in proximity to CTD residues 255-268, 298 -299, 301-303, and the loop joining the NTD (residues 119 -146). Together, these data suggests that the sequence of events begin with carotenoid migration, followed by domain separation at the major interface, and subsequent detachment of the NTD ␣A at the minor interface. We emphasize here that the relative order of events in photoactivation is significant, rather than the specific lifetime of each step, as these will be affected by experimental conditions such as light intensity or ionic strength of the buffer. In particular, the light intensity is of critical importance because of spontaneous reversal of OCP O to OCP R .

OCP relaxation occurs through a compact intermediate
The OCP R to OCP O relaxation follows a single exponential decay with a lifetime of ϳ10 min when measured by the absorbance change at 550 nm at 15°C (Fig. 1A) (12). In contrast, XFMS kinetic progress curves did not show a smooth transition from initial to end state, but instead a complex curve consisting of an initial component representing a change to a more compact (less accessible) state, followed by a slower transition to a more accessible state, and then a final transition to the end state, which could be either more or less accessible than the initial state depending on the specific residue ( Fig. 2A and Fig.  S2). Although the relative amplitude of SA changes within the kinetic progress curves varied, all of them showed three main components ( Fig. 2B and Table S2). The first component entailed a decrease in SA with a lifetime of 0.5-2 min. An increase in SA followed this step with a lifetime of 2-3 min, and then a slow decrease in SA with a lifetime of 5-10 min. In all cases, the net SA changes we observed in the relaxation process corroborate our steady-state data on the relaxation process, which involves re-association of the minor and major interface, movement of the carotenoid to span both domains, and attachment of the NTD ␣A (8). The time constraints to conduct XFMS at the beamline facility and subsequent MS measurements generated a limited number of data points in the timeresolved progress curves in contrast to other contemporary spectroscopic approaches. However, the trend of SA changes was reproducible, and the simulated multiexponential trace provides a clear temporal view of these residue-specific changes. To simplify our interpretation, we categorized sitespecific kinetic progress curves into two, fast and slow, based on the relative amplitude of SA changes of the kinetic components (Table S2). Within the slow category, we identified two subcat-egories representing curves for which there is a net accessibility change from initial to final state and curves for which there is no net accessibility change.
The fast category consisted of progress curves in which the first component showed an initial large decrease in SA within 0.5-2 min, and after that only small reciprocal changes, implying that most of the structural rearrangement for these residues in the relaxation process were completed early on, followed by minor conformational readjustments leading to OCP O . Residues in this category: Ala-8, Arg-9, residues 13-22; Tyr-98, Trp-101, Pro-103, Arg-155, Asn-156, Pro-259, Phe-264, Lys-268, Pro-276, Trp-277, Phe-278, Met-284, Phe-290, Ile-298, Phe-299, Val-301, Ala-302, and Ile-303, are located at the major and minor interfaces. Most of these residues are involved in an inter-domain H-bonding network between conserved water clusters and directly involved in the signal transfer from carotenoid to the surface of the protein (8). The second category is comprised of progress curves that showed a considerable change in SA in the second component, and which occurred on a time scale of 10 min. We further subdivided these curves into two subcategories. The NTD residues, Pro-2, Phe-3, Trp-41, Tyr-44, Met-47, Phe-121, Pro-124, Pro-126, Phe-163, Pro-175, Pro-178, Pro-179, Arg-185, and CTD residues, Pro-225, Phe-227, and Pro-230, were in the first subcategory. These residues showed an overall increase in accessibility and are consistent with structural determinants retrieved from our previous steady-state XFMS studies (8). Residues Trp-41, Tyr-44, and Met-47 are important for carotenoid translocation and stabilization, residues Pro-175 to Arg-185 are on the linker joining the NTD and CTD, and residues Pro-225 to Pro-230 are involved in stabilization of the inter-domain H-bonding network. The decrease in SA in the first phase indicated exclusion of local water from critical regions of OCP R and formation of a compact intermediate state, which is unlike that of the native OCP O . We observed that the reciprocal ratio of net SA change between the first and second kinetic components was ϳ1:2 (Table S2). Therefore, the second component resulted in major rearrangements in the water-side chain H-bonding network in the chromophore-binding pocket to establish the native-like state of OCP O . The presence of the slow decrease in SA in the third component might be due to further conformational rearrangements for the formation of final dark-adapted OCP O . The NTD residues Met-74, Met-83, Pro-94, Phe-109, Trp-110, Tyr-111, Met-117, Met-161, Lys-167, and CTD residues Met-202, Leu-291-292, Pro-297, Pro-309, Lys-310 fell into the second subcategory. For these residues, the SA decreased and then increased in a 1:1 ratio (Table S2), resulting in no net SA change. This observation is consistent with previous studies (8) in which many of these residues showed no net change in SA upon photoactivation, but here, the decrease in SA or compaction in the first component of the kinetic trace might indicate an incorrectly folded intermediate in the relaxation process. The relaxation process is temperature dependent and is slowed significantly at lower temperatures. To confirm the formation of a compact kinetic intermediate we carried out the same experiment at 4°C, collecting accessibility changes from 0 to 30 min in the OCP R to OCP O transition ( Fig. 2A, lower panel, and Fig.  S3). The progressive decrease in SA of all residues further sup-

X-ray footprinting reveals OCP photoprotection
ports the formation of a compact intermediate state in the OCP R to OCP O relaxation process.

XFMS reveals contacts between the OCP and FRP by comparative SA analysis between the OCP and OCP/CTD-FRP complexes
To determine the interaction between of the FRP and OCP, we conducted steady-state XFMS analysis on the OCP and its complex with FRP. We subjected isolated CTD, OCP O , OCP R , and their respective complexes with FRP to a series of increasing X-ray exposures in the microsecond time range as previously described (Fig. 3A) (17). High-resolution quantitative bottom-up LC-MS provided a residue-specific dose response for each site of modification and covered nearly all regions of the protein domains. The pseudo-first order fit of the dose response gives the hydroxyl radical reactivity rate constant (k s Ϫ1 ), which is a measure of both intrinsic reactivity and SA of the modified side chain (Fig. S5-S10). In the comparative structural analysis, the ratio (R) of rate constants between samples determines the relative SA changes, which are independent of any sequence context (8,17). FRP binding may also induce protein conformational changes associated with an increase (R Ͼ 1) or decrease (R Ͻ 1) in SA of certain residues. In our study, we considered a more than 3-fold change in accessibility as significant and indicative of binding and/or conformational change (Fig. 3, B-D, and Table S3).
To obtain a comprehensive view of the interaction of the OCP with the FRP we compared the reactivity rate of an equimolar molar concentration of OCP O and OCP R to that of the OCP-FRP complexes under dark-(OCP O -FRP) and lightilluminated (OCP R -FRP) conditions (Fig. 3, B and C). Notably, we observed strong protection (RϽ0.2) of OCP residues Pro-276, Trp-277, Phe-278, and Met-284 for both the dark-and Red indicates an increase in SA and blue indicates a decrease in SA of OCP residues when comparing the two states. C, relative SA difference arising from the FRP binding on OCP R . D, relative SA differences arising from the FRP binding on CTD. The error bars represent the standard deviation from three independent measurements.

X-ray footprinting reveals OCP photoprotection
light-illuminated complexes. Also, binding of FRP to the CTD alone resulted in marked protection of Met-284 (Fig. 3D). We summarize that this high degree of protection is due to the presence of a stable and bound FRP to both OCP O and OCP R during the microsecond X-ray irradiation time course. Although the protection of these residues indicates binding of FRP at the major interface, the SA of residues Arg-155-Asn-156 increases Ͼ3-fold in the dark complex, but not in the lightilluminated complex. The reactivity of these residues in OCP O and OCP R is consistent with previous XFMS studies on the photoactivation of OCP showing that Arg-155, in particular, becomes more solvent accessible during the OCP O to OCP R transition (Figs. S4 and S5) (8). Addition of FRP has no effect on the SA of Arg-155-Asn-156 in the light-illuminated complex when the domains are already separated, but does increase flexibility of Arg-155-Asn-156 in the dark-adapted complex, which is consistent with a model of FRP binding to the CTD half of the major interface, thereby displacing the NTD and increasing Arg-155 solvent exposure. The residues in the region 13-22, in contrast, increases in SA Ͼ3-fold in both dark and light complexes. Recently, it was suggested through mutation and spectroscopic studies that a region of FRP binds to OCP at the minor interface (14). In this model, an FRP N-terminal segment of approximately 10 residues displaces the NTD ␣A of OCP. Our observation of an increase in accessibility of residues 13-22, which is adjacent to NTD ␣A of OCP, suggests that FRP binding occurs on an extended surface area of both FRP and OCP that might cause a displacement of the helical region adjacent to OCP residues 13-22. We also observe that, although protection of FRP during binding to OCP covers a large area, the Ͼ2-fold protection consistently resides only on peptide(71-83), which forms the head region of FRP (Fig.  4, A and B, and Table S4), consistent with recent studies showing substitution of Phe-76 in FRP severely affects FRP-OCP interaction (5,15). In the light-illuminated complex, the near 2-fold protection is extended to the N-terminal peptide, whereas in the dark and CTD-FRP complex it shows either no overall change (Fig. 4A) or a slight increase, respectively (Fig. 4C). The rest of the FRP segments show similar protection in all complexes. The high degree of protection at the head region is likely due to its binding to the CTD, whereas the variable protection pattern in the N-terminal Red indicates an increase in SA and blue indicates a decrease in SA of FRP residues when comparing the two states. Error bars represent the standard deviation from three to six independent measurements. D, rate of reactivity of FRP and OCP peptides, which are involved in binding interactions. The reactivity rate is obtained from the dose-response plots at various molar concentrations of FRP with respect to OCP R .

X-ray footprinting reveals OCP photoprotection
segment of FRP might be caused by the dynamic interaction of the FRP with OCP.
The rate of hydroxyl radical reactivity of the OCP residues implicated in binding to FRP did not change when the FRP concentration was increased from 1-to 5-fold in excess, suggesting that the OCP-binding site was saturated with FRP ( Fig.  4D and Fig. S11). In contrast, the hydroxyl radical reactivity of the OCP-binding sites on the FRP gradually reversed to the original values as molar excess of FRP increased, indicating the presence of unbound FRP (Fig. 4D). These results are corroborated by studies on the stoichiometry of FRP-OCP complexes using size exclusion chromatography and mutational analysis, which found that OCP likely recruits dimeric FRP, but that further rearrangement is accompanied by FRP monomerization, possibly because OCP domain reassociation disrupts the FRP dimer, and that carotenoid motion occurs after FRP monomerization (15).

Physiological relevance of hysteresis in the photoactivation and relaxation processes
Time-resolved XFMS showed that the OCP photoactivation and relaxation processes occur through different pathways. Under constant illumination the activation process is characterized by single phase kinetic profiles, although with subtle residue-specific rate differences in which the fastest SA changes are confined to prominent carotenoid-binding residues in the carotenoid-binding pocket, followed by SA changes at the inter-domain interface. Because XFMS is highly sensitive to any change in a protein-water network (18), our data also show that the reorganization of the carotenoid environment and carotenoid translocation directly influence the accessibility of local bound and conserved waters, which are present adjacent to the carotenoid cavity and domain interface (8). The destabilization of the carotenoid environment associated with several aromatic residues in the hydrophobic carotenoid-binding pockets was shown as the initial photoactivation event by dynamics crystallography and single crystal microspectrometry (9), and in addition, recent kinetics studies have shown that during photoactivation, carotenoid movement precedes domain rearrangement by over two decades of time (19). However, it has been previously shown that water molecules can also play an important role in hydrophobic and ionic interactions to define the shape of protein cavities (20,21). Therefore, although XFMS could not directly probe changes on those buried aromatic residues, it identified additional residues, directly or indirectly connected to the H-bonding network involving conserved water at the carotenoid-binding pocket undergoing conformational changes during the initial event of the photoactivation process (8) and chromophore migration (22). In contrast to the dynamic crystallography study, we observed the slowest conformational transition at the NTD ␣A. The discrepancy could be due to either the influence of continual back conversion to OCP O in solutionortheinherentstructuraldifferencebetweenthemonomeric OCP in solution versus the strongly associated antiparallel dimeric form in the crystal. Overall our data indicated the carotenoid migration first destabilizes the water-protein net-work at the major interface, which then drives the overall structural reorganization and domain dissociation, thus propagating the signal from the active site to the protein surface.
Previous reports showed that the initial rate of photoactivation is temperature independent and the overall process is ionic strength dependent (12,23). SA measurements by XFMS studies on Arthrospira OCP, an ortholog of Synechocystis OCP, showed that an increase in salt concentration affects local protein-water interactions, which are critical for maintaining the native OCP O state (Fig. S4). The residues at the major and minor interfaces are tightly packed through an extensive H-bonding network between side chains and bound water (8). As shown earlier, this dense hydration layer connects both buried water clusters at the carotenoid cavity and the surfacebound waters at the major and minor interfaces (8). It is possible that water, both as part of bound water clusters and as part of the hydration layer, can mitigate local conformational fluctuations around the carotenoid cavity resulting from temperature changes by a mechanism similar to thermophilic proteins (20,24,25). This can make the initial stage of the photoactivation process strictly driven by the intensity of the light and thus providing a very specific and directional driving force for the signal transfer from the carotenoid to the protein surface. Such temperature independence might provide a way for the organism to induce photoprotection by the intensity of light as the single controlling factor.
The relaxation process showed generally multiphasic kinetic profiles in which the first step involved the formation of a compact intermediate and exclusion of water. Although previous reports (12) using time-resolved fluorescence static quenching measurements indicated the formation of a distinct orange intermediate before the final structural rearrangement, our study pinpointed residue-specific information characteristic of the intermediate state. The domain association and re-structuring of conserved water-protein interactions at the major and minor interface appear faster than the final rearrangements at the carotenoid pocket residues and the minor structural rearrangements throughout the entire protein in the process of returning to the native OCP O conformation. At the same time, because the OCP O and OCP R are always at equilibrium, the fast domain association in the reverse process can contribute to the slower rate of the forward process that we observed under constant illumination. It is likely that the overall rate will depend on the temperature, as indicated in previous UV-visible progress curves (26). The entire relaxation process is indeed highly temperature dependent, suggesting that the compact intermediate might represent one of the local conformational energy minima along the pathway to formation of the final OCP O state. In this view, the relaxation process is a kinetically controlled process in which a decrease in temperature can trap the conversion to a local minimum or a misfolded state as observed by time-resolved XFMS at 4°C ( Fig. 2A and Fig. S3). It has previously been shown that the activation entropy for the OCP-FRP relaxation process is significantly lower than for the unaided relaxation of OCP R to OCP O (14). This is consistent with our interpretation of the data, which reveal the presence of an unfavorable intermediate in the absence of FRP; FRP binding may serve to elim-

Mode of FRP-OCP interactions
XFMS confirms that FRP binds to CTD at the major interface as well as interacts at the minor interface at equimolar concentration. We performed docking simulations with XFMS-based constraints to generate a model of the OCP-FRP complex. The crystal structure of the FRP shows that the protein can exist in two conformations with distinct oligomeric states, a dimer with a folded helix at the N-terminal half, and a tetramer with a long extended helix (5). The tetramer is not prominently observed in solution and another recent crystal structure consisted of the dimer only (27). Docking simulations used the FRP's head region (residues 54 -109) from the FRP dimer, and the FRP tetramer crystal structure with the full-length OCP without carotenoid or the isolated C-terminal domain. The most acceptable solution indicated by a narrow funnel in a room mean square deviation versus interface energy plot (Fig. S12) was obtained from a complex of the FRP head domain originating from the dimer with the OCP-CTD. Our model indicates that the conserved head region remained intact and interacted with the CTD. The head region of FRP binds at the major interface, protecting residues Pro-276, Trp-277, Phe-278, and Met-284. Specifically, a hydrophobic binding interface exists with Leu-81 and Leu-86 of FRP surrounding Met-284 of OCP-CTD that show marked protection upon FRP binding (Fig. 5A).
When we use the FRP head domain-CTD complex as a template and fit the full chain of the FRP dimer, the ␣1 helix clashes with the CTD (Fig. 5B). However, a rotation of the ␣1 helix by about 50 degrees (Fig. 5C) would alleviate the clash and we used such a template as a starting point for a docking simulation. The resulting complex is energetically favorable, indicating that such an interaction would be possible. This type of rotation of ␣1 is also very likely to occur for FRP in solution because it is a prerequisite for the formation of the tetramer form with a long extended helix (a fusion of ␣1 and ␣2) observed in the crystal structure.
The FRP ␣1 helix in this complex is located where the NTD ␣A helix of OCP is located in OCP O , consistent with previous results that suggest that FRP can bind in that region (14). However, in our complex a different region of FRP (residues 36 -40 versus 12-21 in Ref. 14) is in the location where NTD ␣A is in OCP O . Such an extensive mode of binding requires conformational changes in the helical regions of the N-terminal half of the FRP. The slight increase of SA in the N-terminal helix of FRP in the presence of CTD might be due to the conformation change from a folded N-terminal half helical segment to a more elongated one. In contrast, a slight decrease in the SA of the FRP ␣1 helix in the presence of OCP R indicates a relatively stable interaction of the elongated ␣1 helix with the CTD in OCP R . The overall change and a high degree of fluctuation in the SA of the ␣1 helix in the presence of OCP O suggests a dynamic nature of the interaction, which is plausible because of the presence of the NTD ␣A in OCP O . The structure of FRP is known to be flexible, evident by the two very different forms observed in the same crystal, so an influence of the OCP-CTD on the structure of FRP that is connected to its function is quite likely (5,28).
It has also been previously suggested that the FRP acts as a scaffold for OCP domain reassociation (14). Our model indicates that the extended form of FRP binding may bring the two domains of the OCP close to each other, reducing their motion, so that they can efficiently sample different structural configurations and provide an environment conducive to carotenoid movement. Scaffolding by FRP might help order water clusters along the major interface; interdomain interactions between conserved residues and bound water have been shown to be

X-ray footprinting reveals OCP photoprotection
important for proper domain association (8). Formation of the correct H-bonding network, in turn, might facilitate the formation of H-bonds between the carbonyl group of ␤-ring of the carotenoid and Tyr-201 and Trp-288. Noticeably, many residues in the cavity surrounding the carotenoid showed a slightly higher SA in the presence of FRP, in particular, residues Trp-41, Tyr-44, Met-47, Met-161, Phe-163, in the OCP-FRP complex compared with OCP O (Fig. 3, B and C). This indicates that FRP binding may also induce some flexibility in the carotenoid environment and the large cavity at the major interface and prevent any drastic change in a conformation such as compaction in the absence of FRP. An overall model of the OCP photocycle, including the compact intermediate, is shown in Fig. 6.
Recently, new paralogous OCP families have been discovered in cyanobacteria (27). For example, the cyanobacterium Tolypothrix sp. PCC 7601 encodes two distinct forms of the OCP. OCP1, the canonical OCP, is functionally equivalent to the Synechocystis OCP. In contrast, OCP2, a new family, exhibits both faster photoactivation and back-conversion, is less temperature sensitive, and FRP does not influence its dark reversion. Given that FRP may act as a scaffold to accelerate back-conversion of OCP1, it is interesting to compare the differences in primary structure between OCP1 and OCP2 (Fig. S13). One of the key sequence differences between OCP2 and OCP1 is the substitution of isoleucine for methionine in position 284. In the OCP1 photoactivated form, the residue Met-284 becomes highly accessible, and during relaxation in the presence of FRP, it forms a rapid compact or misfolded intermediate. In the orange form, residue Met-284 resides very close to the conserved water cluster and the carotenoid at the major interface. Methionine residues are unique in the protein core as they can have hydrophobic interactions while also engaging in polar contacts (6, 29 -31). The presence of isoleucine might eliminate the requirement for FRP to bind and assist the formation of the correct polar contacts at the major interface involving water clusters. Two other critical differences, the substitution of Tyr-240 to Phe-240, and the substitution of Lys/Arg for Glu-229, might have a direct influence on the non-FRP relaxation of OCP2. Phe-240 in OCP1 is located close to the ␤keto group of the carotenoid in the CTD and a tyrosine at the same position further stabilizes the carotenoid by H-bond formation. Lys/ Arg-229 in OCP1 is projected toward Arg-9 and participates in H-bonding to stabilize the ␣A-CTD interaction at the minor interface. Arg-9 is involved in an H-bonding network with water clusters at the minor interface, which in turn connects several functionally important residues such as Arg-155 (NTD) and Phe-227 (CTD). The substitution of Arg-229 by Glu-229 might further strengthen the domain interaction at the minor interface further by electrostatic stabilization. Overall in the OCP2, substitution of several conserved residues might have led to a more rapid sampling of the domain interface to form the correct geometry of the dark-adapted conformation in the absence of FRP. Recent studies (3) have pinpointed OCP residues Phe-299 and Asp-220 as playing an important role in the acceleration of the relaxation process in the presence of FRP. Although this study did not explicitly show a change in accessibility of these residues upon FRP binding, likely because the residues are in proximity to water molecules in both the bound and unbound FRP forms, it is interesting to note that they are in

X-ray footprinting reveals OCP photoprotection
the region in which the N-terminal extension of the FRP binds in the scaffolding model (Fig. 5).

Conclusion
In this study, X-ray radiolytic labeling revealed new mechanistic details of the key events in OCP photoprotection, providing both a spatial and temporal view of site-specific conformational changes in the OCP and its interaction with the FRP. By temporally differentiating residue-specific SA changes we have determined the driving force for photoactivation and relaxation under native conditions by pinpointing the sequence of structural events during these processes. We determined that FRP provides an extended binding region, which effectively holds together the OCP domains at the major and minor interfaces, and forces proximity of the two domains to lead to the acceleration of OCP R relaxation to OCP O . Together, these data allowed the construction of a molecular model of the FRP-binding process and its role in the regulation of the OCP. Recent studies have suggested that the OCP-CTD is accessible when OCP is bound to the PB (32), and therefore future studies will include investigating the role of FRP in detaching OCP from the OCP-PB complex. Studying residue-specific kinetics of water accessibility changes in OCP R in the presence of the FRP, and as a function of OCP binding to PB will generate a more complete understanding of the driving force underlying the NPQ mechanism in cyanobacteria. Understanding the switching on and off of thermal dissipation, and the action of the "photoswitch," which controls this mechanism has long-ranging implications in synthetic biology, such as design of optogenetic switches (33) or the finer control of photoprotection and consequent enhancement of photosynthetic efficiency (34).

Purification of OCP, FRP, and CTD
OCP and FRP samples were purified using procedures described previously (5,8). CTD preparations incorporating canthaxanthin (CAN) were obtained by expressing the gene for Synechocystis sp. PCC 6803 in BL21(DE3) competent Escherichia coli cells with His 6 tag added. BL21(DE3) cells were transformed simultaneously with the pCDFDuet-1 vector and the pAC-CANTHipi plasmid (Addgene plasmid number 53301) 36 containing the genes enabling CAN biosynthesis in E. coli. The CTD CAN holoprotein isolated by nickel-nitrilotriacetic acid affinity chromatography (HisTrap Affinity column, GE Healthcare) yielded a mixture of apo-and holoproteins. The CTD-CAN holoprotein was further isolated by hydrophobic interaction chromatography (HiTrap HIC phenyl column, GE Healthcare). The purity of the final OCP fractions was confirmed by SDS-PAGE. The protein samples were exchanged into 20 mM potassium phosphate (pH 7.4), 100 mM NaCl by size exclusion chromatography (SEC) on a Superdex-75 10/300 GL column prior to XFMS experiments.

Time-dependent XFMS measurements to follow OCP photoactivation and relaxation
The OCP ECN (expressed in Synechocystis PCC 6803) samples were exchanged into 20 mM potassium phosphate (pH 7.4), 100 mM NaCl by SEC on a Superdex-75 10/300 GL column prior to XFMS experiments. Based on UV-visible spectrometry measurements of OCP photoactivation and deactivation, the time window to follow the conformation transition by XFMS is set to a range of "zero" to 30 min. To follow the photoactivation pathway, 2.5 ml of OCP O inside the sample loading syringe of microfluidic flow system was illuminated with a blue LED array (470 nm Luxeon Rebel, Philip Lumileds) at room temperature for 30 min. At various intervals (delay time) within this duration of the LED illumination, 200-l sample was passed through a 200-m inner diameter Polymicro capillary tube and 200 ϫ 500 m 2 beam spot to receive a 200 s of X-ray exposure at beamline 5.3.1 at the Advanced Light Source (17). The zero time interval sample was collected at the beginning under dark conditions. All samples were immediately quenched with methionine amide to stop the secondary oxidations and stored at Ϫ80°C for LC-MS analysis. To follow the time course of the reverse pathway, which is the decay of OCP R to OCP O , 2.5 ml of OCP ECN sample was first illuminated inside the sample loading syringe of a microfluidic flow system for 15 min at room temperature to prepare OCP R . The LED was then turned off, and at various intervals, 200 l of sample was exposed to X-ray as described above. The zero time interval of the reverse transition was collected at the beginning in the presence of LED illumination. In both forward and backward transitions, we were able to collect the first data point only at ϳ40 s because of the inherent delays associated with a manual operation of beamline components and sample handling system.

Steady-state XFMS measurements to study interaction of FRP with OCP and CTD
The steady-state XFMS were carried out at the Advanced Light Source beamline 5.3.1 using a standard microfluidic set-up (8,17). All protein samples (OCP O , OCP R , OCP O -FRP, OCO R -FRP, CTD, FRP, and CTD-FRP) at 5 M concentration in 20 mM potassium phosphate (pH 7.4) buffer containing 100 mM NaCl were subject to X-ray exposure from 0 to 800 s, and immediately quenched with methionine amide to stop the secondary oxidations and stored at Ϫ80°C for LC-MS analysis. The sample handling and exposure of OCP O -FRP and OCO R -FRP complex were carried out under dark-and light-illuminated conditions, respectively. The protein complexes OCP-FRP and CTD-FRP were prepared at 5 M concentration by mixing an equimolar ratio of individual protein components. Additionally, the OCP R -FRP complex at various amounts of molar excess of FRP (1-, 2-, and 5-fold) was exposed to monitor the effect of a molar excess of FRP on SA changes at the binding sites. The SA changes are proportional to the reactivity rates, which rates have been corrected for the scavenging effect of the excess amount of FRP determined by the standard Alexa 488 assay (35).

Mass spectrometry and data analysis
The pH of the exposed samples were increased to pH 8 by ammonium bicarbonate to a final concentration of 50 mM. The disulfide linkages in the protein samples were reduced by treatment with DTT to a final concentration of 5 mM and incubating at 55°C for 30 min. The Cys residues were alkylated by treating X-ray footprinting reveals OCP photoprotection the samples with iodoacetamide to a final concentration of 15 mM at room temperature and in the dark for 30 min. The Cysalkylated proteins were then desalted and buffer exchanged with 50 mM ammonium biocarbonate (pH 8) using Zeba TM desalting spin columns (Thermo Fisher Scientific). Samples were digested overnight with trypsin (1:20, w/w, enzyme:protein) and endoproteinase GluC (1:10, w/w, enzyme:protein) at pH 8 and 37°C. Proteolyzed samples were analyzed in an Agilent 6550 iFunnel Q-TOF mass spectrometer (Agilent Technologies, Santa Clara, CA) coupled to an Agilent 1290 LC system (Agilent) using Sigma Ascentis Peptides ES-C18 reverse phase column (2.1 ϫ 100 mm, 2.7 m particle size; Sigma). Approximately 10 pmol of samples were loaded onto the column via an Infinity Autosampler (Agilent) with buffer A (2% acetonitrile, 0.1% formic acid) flowing at 0.400 ml/min. The peptides were separated and eluted into the mass spectrometer via a gradient with initial conditions of 5% buffer B (98% acetonitrile, 0.1% formic acid) increasing to 70% B over 15 min. Subsequently, B was increased to 90% over 1 min and held for 3 min at a flow rate of 0.6 ml/min followed by a ramp back down to 5% over 1 min where it was held for minutes to re-equilibrate the column to the original condition. Peptides were introduced to the mass spectrometer from the LC using a Jet Stream source (Agilent) operating in positive-ion mode (3500 V). The data were acquired with MassHunter B.05.00 operating in Auto MS/MS mode, whereby the three most intense ions (charge states 2-5) within 300 to 1400 m/z mass range above a threshold of 1000 counts were selected for MS/MS analysis. MS/MS spectra were collected with the quadrupole set to "narrow" resolution and collision energy to optimize fragmentation. MS/MS spectra were scanned from m/z 100 to 1700 and collected until 40,000 total counts were collected or for a maximum accumulation time of 333 ms. Parent ions were excluded for 0.1 min following MS/MS acquisition. MS/MS data of native and modified peptide fragments were interpreted by Mascot MS/MS Ions Search as well as verified manually. The abundance of native and modified peptides at any irradiation time point area were measured (peak area) from their respective extracted ion chromatogram using Agilent Mass Hunter version 2.0.
The peak area from the extracted ion chromatograms of a specific peptide fragment with a particular m/z and associated ϩ16, ϩ32, or ϩ48 Da side chain modifications was used to quantify the amount of modification yielded at 200 s of exposure time for time-resolved studies. The increase in delay time progressively reduces or increases the fraction of the unmodified product. The plot of fraction unmodified versus delay time provides a site-specific kinetic trace. The multiple experimental repeats of the kinetic traces are fitted globally to an exponential decay function with Origin version 6.1 (OriginLabs) to determine the lifetime () of conformational change. The reported errors of the rate data were determined by the Origin program using 95% confidence limits of the fitting results.
For FRP binding studies the peak area from the extracted ion chromatograms of a specific peptide fragment with a particular m/z and associated ϩ16, ϩ32, or ϩ48 Da side chain modifications was used to quantify the amount of modification at a given irradiation time. Increasing irradiation progressively reduces the fraction of unmodified products and provides a site-specific dose-response plot (17,22). The hydroxyl radical reactivity rate (k), which depends on both intrinsic reactivity and SA, was obtained by fitting the dose response to a single exponential decay (based on a pseudo-first order reaction scheme using Origin 7.5 (OriginLabs) (Fig. S5-S11). The ratio (R) of the measured reactivity of the side chains residues between free protein and bound complex (R ϭ k bound /k free ) gave information on SA changes independent of the intrinsic reactivity (Tables  S3 and S4).

Docking simulations
XFMS-based restraints were used to generate the starting positions of the docking partners using PatchDock (36). These starting positions of the docking partners were then uploaded to RosettaDock (37). 1000 independent simulations were performed and the 10 best-scoring structures were returned at the end of the simulation. One of those solutions was then used for a refinement run with RosettaDock (10,000 structures) with standard deviation perturbation parameters of 4 Å for translation and 6 degrees for rotation. The energetic "funnel" with low energy structures clustered around a single position with minimal room mean square deviation was obtained (Fig. S12). The model of full-length FRP with a rotated ␣1 helix was generated based on the full-length FRP dimer with a manually rotated helix that was subsequently relaxed with RosettaRelax and then used for docking. Simulations with full-length OCP and FRP as well as combinations without restraints did not indicate successful runs. Structures were visualized in PyMol 2.0 (Schrodinger) and Chimera (38).