A Highly Conserved Cysteine of Neuronal Calcium-sensing Proteins Controls Cooperative Binding of Ca2+ to Recoverin*

Background: Recoverin contains a cysteine (Cys-39) that is highly conserved in neuronal calcium-sensing (NCS) proteins. Results: The C39A mutation shifts the conformational equilibrium from the R to T state, inducing cooperative calcium binding. Conclusion: Cys-39 controls the conformational equilibrium of calcium-free recoverin. Significance: This mutation assigns a previously unknown function to the conserved cysteine in recoverin and possibly all NCS proteins. Recoverin, a 23-kDa Ca2+-binding protein of the neuronal calcium sensing (NCS) family, inhibits rhodopsin kinase, a Ser/Thr kinase responsible for termination of photoactivated rhodopsin in rod photoreceptor cells. Recoverin has two functional EF hands and a myristoylated N terminus. The myristoyl chain imparts cooperativity to the Ca2+-binding sites through an allosteric mechanism involving a conformational equilibrium between R and T states of the protein. Ca2+ binds preferentially to the R state; the myristoyl chain binds preferentially to the T state. In the absence of myristoylation, the R state predominates, and consequently, binding of Ca2+ to the non-myristoylated protein is not cooperative. We show here that a mutation, C39A, of a highly conserved Cys residue among NCS proteins, increases the apparent cooperativity for binding of Ca2+ to non-myristoylated recoverin. The binding data can be explained by an effect on the T/R equilibrium to favor the T state without affecting the intrinsic binding constants for the two Ca2+ sites.

an amphipathic helix also recognized by rhodopsin (7), and thus prevents phosphorylation of activated rhodopsin. When the Ca 2ϩ concentration is low, RK is released by recoverin and is then free to phosphorylate rhodopsin in a reaction that helps terminate the photoactivated state (8).
Recoverin is a member of the neuronal calcium sensor (NCS) protein family (9). Like all members of the family, it is a small globular protein with four EF hand motifs and a post-translationally added fatty acyl group (myristoyl or related acyl chain) on the N-terminal glycine residue (Fig. 1A) (10). Two of the EF hands are non-functional in recoverin (Fig. 1B): EF1 as a result of the highly conserved CPXG sequence (residues 39 -42) and EF4 as a result of a salt bridge between Lys-161 and Glu-171 (11). Consequently, recoverin binds only two Ca 2ϩ ions, one each at EF2 and EF3 (12). The intrinsic affinities of the two sites for Ca 2ϩ are different, but the binding is cooperative because Ca 2ϩ and the myristoyl chain bind preferentially to two different conformations of the protein (R and T, respectively). The T state is a more compact structure than the R state and sequesters the myristoyl chain within a hydrophobic cavity of recoverin. The binding of two Ca 2ϩ ions to recoverin results in the extrusion of the myristoyl chain from the protein interior into the aqueous milieu as recoverin changes from the T to the R state. This conformational transition is known as the "myristoyl switch" (13,14). In the absence of the acyl chain, non-myristoylated recoverin exists predominantly in the R state, and binding is distinctly non-cooperative, with K D values reflecting the intrinsic affinities of the two sites for Ca 2ϩ ions. Although recoverin clearly binds two Ca 2ϩ ions (12), it should be noted that the protein has never been crystallized with Ca 2ϩ bound to both sites. In all structures of the Ca 2ϩ -bound form, only the high affinity EF3 site is occupied (11,15,16).
We show here that mutation of the highly conserved Cys-39 (to alanine) within the CPXG motif of EF1 has a dramatic effect on the binding of Ca 2ϩ to non-myristoylated recoverin. The mutation increases the apparent cooperativity for binding Ca 2ϩ by a mechanism that is most parsimoniously interpreted as an effect on the T/R equilibrium to favor the T state without significantly affecting the intrinsic affinities of EF2 or EF3 for Ca 2ϩ . We also show from the x-ray crystal structure of wildtype (WT) recoverin under non-reducing conditions that Cys-39 can form a stable sulfenic acid. Cys-39 is the only cysteine residue in recoverin and has been suggested to function as a redox sensor in rod photoreceptor cells through formation of derivatives with higher oxidation states of the sulfur (i.e. sulfenic, sulfinic, or sulfonic acids) (17,18).
Preparation of Rv-Recoverin was expressed in T7 Express Escherichia coli (New England Biolabs). Cultures were grown with orbital shaking (220 rpm) to A 660 Ϸ 0.6 -0.8 at 37°C before inducing with 0.25 mM IPTG. Following incubation for an additional 3 h at 37°C, cells were harvested by centrifugation, resuspended in Buffer A, and frozen at Ϫ80°C. Recoverin was purified (19,20) from thawed cell suspensions by sonicating (in the presence of 1.6 mM PMSF) to break cells open, centrifuging to clarify the supernatant fraction, and applying the clarified lysate to a low substituted phenyl-Sepharose matrix (Amersham Biosciences) that had been equilibrated with Buffer A. Bound recoverin was eluted with a linear gradient of Buffer A to Buffer B, and fractions were analyzed by SDS-PAGE. Those fractions containing recoverin were combined and concentrated with a YM-10 Centricon to 1-2 ml and then dialyzed against Buffer C at 4°C to remove EGTA. Purified protein was quantified spectrally by ⑀ 280 (WT, 24,075 M Ϫ1 cm Ϫ1 ; C39A, 24,000 M Ϫ1 cm Ϫ1 ) (17,21) and frozen at Ϫ80°C until use.
Preparation of mRv-mRv was prepared from T7 Express E. coli cells coexpressing yeast N-myristoyltransferase1 in pBB131 (20) with recoverin. Cells were grown to A 660 Ϸ 0.15 at 37°C with orbital shaking at 220 rpm at which point sodium myristate was added to a final concentration of 0.1 mM (19). Following a 30-min incubation period, the temperature was decreased to 27°C, and expression was induced by addition of 0.5 mM IPTG. Cells were harvested after 2 h, resuspended in Buffer C, and frozen at Ϫ80°C until use in purification of recoverin. Bulk recoverin was purified using the low substituted phenyl-Sepharose matrix as described above, followed by chromatography on a Q-Sepharose FF matrix (GE Healthcare) using a KCl step gradient in a 20 mM Tris buffer (pH 8.0) to separate mRv from Rv; Rv elutes at Ϸ80 mM KCl, whereas mRv elutes between 100 and 120 mM KCl (20). Fractions containing mRv were then pooled, concentrated, dialyzed against Buffer C, and frozen at Ϫ80°C until use. The purification of mRv was monitored by reverse-phase HPLC using a 0 -80% acetonitrile gradient (in 0.1% trifluoroacetic acid) with a C18 analytical column (Advanced Chromatography Technologies, Aberdale, Scotland) on an Agilent 1100 series HPLC (Agilent Technologies, Berkshire, UK). All mRv preparations used in this study were Ͼ95% myristoylated.
Myristoylation of recoverin was also followed by electrospray ionization mass spectrometry (ESI-MS). The LC-MS system consisted of an Agilent 1200 series HPLC connected to an Agilent series 6520 ESI Q-TOF. Protein samples (10 M) dissolved in a 5% acetonitrile and 0.1% formic acid buffer were separated on a C18 Poroshell 300SB column (1 mm ϫ 75 mm ϫ 5 m) at 0.5 ml min Ϫ1 using a linear gradient of 5-70% acetonitrile in 0.1% formic acid. MS data were collected up to 3000 m/z, and raw spectra were deconvoluted using the maximum entropy algorithm of Agilent Masshunter version B.03.01 software. External mass calibration was performed using a mixture of purine (121 m/z) and HP-0921 (922 m/z) immediately prior to measuring protein samples. B, sequence alignment of the four EF hand loop regions of recoverin. The consensus sequence for EF hand loops is shown with the Ca 2ϩ -coordinating residues denoted as X, Y, and Z (underline indicates the negative position), n represents any non-polar residue, and a dash is any residue. The conserved Ca 2ϩ -binding residues are marked in red boldface. The gray box in EF1 highlights the conserved CPXG motif.
Preparation of Oxidized WT Rv (oxWT Rv)-Cell pellets containing WT Rv were resuspended in Buffer D and sonicated to release the expressed protein. The lysate was clarified by centrifugation and then passed over a Q-Sepharose FF matrix. The column was washed with 5 column volumes of Buffer D, and bound protein eluted with a linear gradient of Buffers D and E in 15 column volumes. Column fractions were analyzed by SDS-PAGE. Fractions containing Rv were concentrated with a YM-10 Centricon and stored at 4°C in the presence of 10 mM CaCl 2 for 1 month before crystallographic trials.
Preparation of the Regulator of G Protein Signaling (RGS) Domain of RK-RGS is a truncated form of RK in which the catalytic domain is replaced with a small linker peptide (i.e. GSGS) joining residues 1-181 to 512-557 of the kinase (these two regions form the RGS homology domain (22)). RGS is soluble, is expressed well in E. coli, and can be purified by means of a C-terminal His 6 tag. This tag was also used to immobilize RGS for binding assays.
RGS was expressed from the pET28a vector in T7 Express E. coli cells. Cells were grown to A 660 Ϸ 0.6 -0.8 at 37°C with orbital shaking at 220 rpm before induction with 0.25 mM IPTG. Induced cultures were grown an additional 2-3 h before harvesting cells by centrifugation. Cell pellets were resuspended in Buffer F and frozen at Ϫ80°C until purification. RGS was purified from thawed cell suspensions by sonication to break cells open, centrifuging to clarify the supernatant fraction, and applying the clarified lysate to a Ni-NTA matrix (GE Healthcare) that had been equilibrated with Buffer F. Bound RGS was eluted with a linear imidazole gradient using Buffers F and G. Purified protein was quantified spectrally using ⑀ 280 (33,640 M Ϫ1 cm Ϫ1 ) (21) and stored for up to 1 week at 4°C. One liter of culture typically yielded 2-4 mg of purified RGS.
Ca 2ϩ -binding Assays-Ca 2ϩ -binding assays were performed at 25°C according to published protocols (14, 23) using a Hitachi F-2500 fluorescence spectrometer (Tokyo, Japan) to follow changes intrinsic tryptophan fluorescence in recoverin. Ca 2ϩ standards were prepared by serial dilutions from a 0.1 M CaCl 2 standard (Ricca Chemical Co.). Titration data were fit initially to one of two models: a model for cooperative interaction of the two Ca 2ϩ -binding sites and a non-cooperative model for two independent binding sites. In the cooperative model, data are fit to the Hill equation, is the fraction of sites bound to Ca 2ϩ , [Ca 2ϩ ] is the free Ca 2ϩ concentration, K d is the apparent dissociation constant, and n is the Hill coefficient. In the non-cooperative model, the K D values (K 1 , K 2 ) are fit as independent parameters, where K 1 is the first dissociation constant, and K 2 is the second dissociation constant. Data were also analyzed using Equation 3 for the concerted allosteric model shown in Scheme 1, where T refers to the conformation of the Ca 2ϩ -free protein (subscripts indicate the number a Ca 2ϩ bound at each site), R the conformation of the Ca 2ϩ -bound protein, L is the T/R ratio, K 1 and K 2 are the same as in Equation 2, c is the ratio of Ca 2ϩ affinity for the T and R states (0.003; assumed to be the same for both binding sites), ␤ is equal to c ϩ L Ϫ1 , and ␥ is equal to c 2 ϩ L Ϫ1 (14,23). Titration data were fit to Equations 1-3 using KaleidaGraph 4.1 from Synergy Software (Reading, PA).
RGS-binding Assay-The formation of a complex between RGS and Rv was followed using a Ni-NTA matrix in a type of pulldown assay using spun columns. Typically, 10 nmol of purified RGS was first immobilized on the column. After removing unbound protein with Buffer H, recoverin was added in 1.5-fold molar excess over RGS. Following an incubation period of 30 min on ice, the column was washed again with Buffer H, and the recoverin-RGS complex eluted with 250 mM imidazole in Buffer H. Fractions were analyzed by SDS-PAGE.
The K D for the recoverin-RGS interaction was measured using a MicroCal VP-ITC microcalorimeter (GE Healthcare) as described previously (6). WT or C39A Rv was dialyzed against two changes of Buffer I at 4°C and then loaded into the sample syringe at a concentration of 0.35-0.66 mM, to be used for titration of the RGS sample (30 -60 M).
Rv Crystallization-Initial crystallization trials used sitting drop vapor diffusion at 20°C with Hampton sparse matrix crystallization screens (Hampton Research, Aliso Viejo, CA). Protein was mixed 1:1 with crystallization mother liquor from a reservoir solution of 50 l, and drops were set up using a Phoenix robot (Art Robbins Instruments, Sunnyvale, CA). Once initial conditions were identified, subsequent exploration of crystallization space used the hanging drop method. Final crystallization conditions were: C39A, 2.0 M ammonium citrate (pH 7.0); WT, 1.8 M ammonium citrate (pH 7.0); oxWT, 1.8 M ammonium citrate (pH 7.0); and C39D, 2.4 M sodium malonate (pH 7.0). All samples contained 1-5 mM CaCl 2 . oxWT Rv was stored for 2-4 weeks at 4°C in the presence of 10 mM CaCl 2 before use in crystallization trials.
X-ray Data Collection and Analysis-Crystals were soaked in reservoir solution containing 10% glycerol as cryoprotectant before flash freezing in liquid nitrogen. Diffraction data were SCHEME 1

Cys-39 Influences the Conformation of Calcium-free Recoverin
collected at 100 K with beam line 8.2.1 at the Advance Light Source (Lawrence Berkeley National Laboratory, Berkeley, CA) using ADSC Q315R CCD detectors (Area Detector Systems Corporation, San Diego, CA). Data sets were integrated using MOSFLM (24) and scaled using SCALA (25) from the CCP4 software suite v6.3 (26,27). All structures were solved by the molecular replacement method using PHASER (28) with WT Rv (PDB ID code 1OMR (15)) as a search model. Rigid body refinement followed by positional and B-factor refinement was carried out using phenix.refine (29) from the PHENIX software suite v1.8 (30). Manual model building was done using COOT v0.7 (31).
Continuous electron density was observed for residues: 7-197, C39A; 7-198, C39D; 7-198, WT; and 7-194, oxWT. In the oxWT structure, additional 3 F o Ϫ F c density was observed at the side chain of residue Cys-39. This density was interpreted to be an oxygen atom of a sulfenic acid side chain and was included from a built-in monomer library of COOT. Water molecules were included in the final refinement after satisfying the criteria of 3 cutoff F o Ϫ F c and 1 cutoff 2 F o Ϫ F c . Several iterative cycles of refinement were carried out before final submission of data. The data collection and final refinement statistics are given in Table 1. Data sets for C39A Rv (PDB ID code 4M2O), C39D Rv (PDB ID code 4M2P), WT Rv (PDB ID code 4MLW), and oxWT Rv (PDB ID code 4M2Q) have been submitted to the Protein Data Bank.

RESULTS
In the following presentation, we use mRv to indicate the myristoylated form of recoverin, Rv to indicate the non-myristoylated form, and recoverin when we do not want to specify whether or not the protein is acylated.
C39A Mutant-As part of an ongoing effort to study the structure and function of recoverin, we explored a variety of approaches to fluorescently label the protein. Although the single cysteine residue Cys-39 has been labeled with a maleimide derivative of Alexa Fluor 647 (32), the location of this residue in the binding pocket for RK (6) made it non-ideal for a fluorophore label in our studies. To evaluate whether the unique cysteine could be moved to another location in the protein, we first prepared the single mutant C39A to determine the impact of removing this highly conserved cysteine. Coexpression of recoverin and yeast N-myristoyl transferase 1 in E. coli followed by purification on a phenyl-Sepharose column consistently yielded a greater fraction of the C39A mutant in the myristoylated form (78.5 Ϯ 9.7%) than was observed for WT recoverin (53.8 Ϯ 5.5%), as judged by analysis with reverse-phase HPLC (data not shown). We decided to characterize the C39A mutant further because these data suggested that the mutation might affect the conformation of recoverin. In the following experiments, mRv was purified from contaminating Rv using chromatography on Q-Sepharose and then analyzed for purity (generally Ͼ95% mRv) by HPLC and ESI-MS (data not shown).
Binding of Ca 2ϩ to the C39A Mutant-The binding of Ca 2ϩ to recoverin was monitored using a method that measures intrinsic tryptophan fluorescence as has been described extensively by Ames and co-workers (14,23,33). In agreement with the earlier studies, binding of Ca 2ϩ to WT mRv was cooperative, and the titration data were fit well by the Hill equation (Equation 1) using a K d of 25.9 Ϯ 0.9 M and Hill coefficient n ϭ 1.46 Ϯ 0.07 ( Fig. 2A). Also in agreement with the earlier studies, binding of Ca 2ϩ to WT Rv was non-cooperative, and the titration data were fit well with an equation for two non-identical sites (Equation 2) using dissociation constants of 0.21 Ϯ 0.03 M (K 1 ) and 2.23 Ϯ 0.33 M (K 2 ) for the high and low affinity sites corresponding to EF3 and EF2, respectively ( Fig. 2A). In stark contrast, C39A Rv displayed a much steeper Ca 2ϩ binding curve compared with WT Rv, and the data were fit well by the Hill equation for cooperative binding with K d of 1.33 Ϯ 0.06 M and Hill coefficient of n ϭ 1.23 Ϯ 0.06 (Fig. 2B), suggesting that the effect of mutation was on the cooperative interaction of the two binding sites. Certainly, an n of 1.2 is not large, and the data could also be fit reasonably well with the model for independent   DECEMBER 13, 2013 • VOLUME 288 • NUMBER 50 sites (Equation 2) using K D values for the two sites that were fortuitously identical (K 1 ϭ K 2 ϭ 2.4 M) as a result of the mutation. In this case, the mutation would be interpreted to affect the intrinsic affinity of EF3 directly. Although we cannot unequivocally rule out a direct effect of mutation on EF3, we favor the model in which the mutation affects the T/R equilibrium for reasons outlined under "Discussion." To explore further the possible effect of the mutation on the cooperative interaction of the two Ca 2ϩ -binding sites, we analyzed the Ca 2ϩ titration data for WT and C39A recoverin using the concerted allosteric model developed by Ames et al. (Scheme 1 and Equation 3) (14). According to this model, the two Ca 2ϩ -binding sites, EF2 and EF3, bind Ca 2ϩ with different intrinsic affinities. Furthermore, the Ca 2ϩ -free protein exists in an equilibrium mixture of R and T states, where the myristoyl chain binds preferentially to the T state and Ca 2ϩ binds preferentially to the R state. Binding of Ca 2ϩ to WT mRv is cooperative because the T state is favored in mRv, and much of the energy from binding Ca 2ϩ to the high affinity EF3 site is utilized in converting the protein into the Ca 2ϩ -binding R state. The R state is favored in WT Rv, and the two sites bind Ca 2ϩ independently. In this model, K 1 and L are not independent parameters. Therefore, we fixed K 1 and K 2 to be values close to those observed for the WT Rv titration and fit the data to determine L. As shown in Table 2, the equilibrium ratio of T to R states for the non-myristoylated form of the protein is dramatically shifted from L ϭ 0.31 Ϯ 0.14 in WT Rv to 11.1 Ϯ 0.8 in C39A Rv, providing a clear rationale for why this mutant exhibits coop-erative interaction of the Ca 2ϩ binding sites in the non-myristoylated protein.

Cys-39 Influences the Conformation of Calcium-free Recoverin
Interestingly, the effect of the C39A mutation is lost in the myristoylated protein. As shown in Fig. 2B, the cooperative binding of Ca 2ϩ to C39A mRv was almost indistinguishable from that of WT mRv, with a K d of 19.5 Ϯ 0.4 M and Hill coefficient of n ϭ 1.52 Ϯ 0.04. A fit of the data to the concerted allosteric model indicates that the equilibrium constant L changes from 490 Ϯ 30 in WT mRv to 280 Ϯ 10 in the mutant ( Table 2).
RGS-binding Assay-Recoverin binds the N-terminal helix of RK (4 -6). In this work, we used a truncated form of RK, designated RGS, composed of the N-terminal 185 amino acids fused to residues 510 -558 from the C terminus using a short GSGS linker (see "Experimental Procedures"). C39A Rv binds RGS with a 1:1 stoichiometric ratio, similar to WT Rv (Fig. 3). The K D of WT and C39A Rv for RGS, as measured by isothermal titration calorimetry (ITC), is 1.0 Ϯ 0.4 M and 0.91 Ϯ 0.15 M, respectively (Fig. 3, B and C). Hence, the C39A mutant is unimpaired in its ability to bind RK.
P40A Mutant-The Pro-40 residue of the highly conserved CPXG motif was also of interest because of its covariance with Cys-39 in the NCS family. For this reason, we generated the P40A mutant and tested for the effect of mutation on Ca 2ϩ and RGS binding to the protein. As is shown in Fig. 4A, the P40A mutation did not restore cooperative binding of Ca 2ϩ to Rv, as was the case for C39A. The titration curves were less steep than for C39A and, in fact, less steep than for WT Rv, displaying clearly non-cooperative binding of Ca 2ϩ , with dissociation constants of less affinity (K 1 ϭ 0.34 Ϯ 0.06 M and K 2 ϭ 10.3 Ϯ 2.0 M) than those of WT Rv. P40A was able to bind RGS as judge by the pulldown assay of Fig. 4C.
X-ray Crystal Structure of C39A Rv-C39A Rv crystals were grown in 2.0 M ammonium citrate (pH 7.0) containing 2 mM CaCl 2 and diffracted to 1.5 Å resolution. As is shown in Fig. 5A, the overall structure of the C39A mutant is virtually identical with that for WT Rv (WT Rv crystals were grown in 1.8 M ammonium citrate (pH 7.0) containing 2 mM CaCl 2 and diffracted to 1.45 Å resolution). In particular, the structure of Ca 2ϩ -bound EF3 in C39A Rv differed from EF3 in WT Rv by an r.m.s.d. of only 0.042 Å (Fig. 5B), showing very little perturbation to the Ca 2ϩ -bound structure resulting from the mutation.
X-ray Crystal Structure of WT Rv under Oxidizing Conditions (oxWT Rv) and the C39D Rv Mutant-oxWT Rv crystals were grown in 1.8 M ammonium citrate (pH 7.0) containing 2 mM CaCl 2 and diffracted to 1.9 Å resolution. Interestingly, there is extra density for Cys-39 consistent with an additional oxygen  atom attached to the sulfur (Fig. 6), which we have modeled as a sulfenic acid. We also prepared the C39D Rv mutant, which was originally created by Permyakov et al. (18) to mimic a higher oxidation state (sulfinic acid) of the Cys-39 sulfur atom. As is shown in Fig. 4B, C39D Rv bound Ca 2ϩ with lower affinity at both sites, shifting the titration curve to the right with little change in the overall slope. Interestingly, C39D Rv was incapable of binding RGS in the pulldown assay shown in Fig. 4C. Because of the disruption to interaction with RGS, it was of interest to determine the crystal structure of the Ca 2ϩ -bound C39D Rv mutant. C39D Rv crystals were grown in 2.4 M sodium malonate (pH 7.0) containing 2 mM CaCl 2 and diffracted to 1.45 Å resolution. The global C39D Rv structure displayed no obvious differences from that of WT Rv, suggesting that the disruption to RGS binding resulted from a direct effect of the mutation in the RGS-binding pocket. Electron density for residue 39 is compared for WT, oxWT, C39A, and C39D Rv in Fig. 6, A-D.

DISCUSSION
The main effect of the C39A mutation on the non-myristoylated form of recoverin is an increase in the observed K D for the high affinity EF3 Ca 2ϩ -binding site with little or no change to the low affinity EF2 site (Fig. 2). This result is superficially similar to the effect of the ⌬191-202 truncation in the mutant Rv  reported by Koch and co-workers (16) but has been interpreted very differently here. The ⌬191-202 truncation affects directly the intrinsic affinity of EF3 for Ca 2ϩ , whereas we have interpreted the C39A mutation as directly affecting the T/R equilibrium without significant perturbation of the intrinsic binding constants for either EF2 or EF3.
The ⌬191-202 truncation removes a C-terminal helix that is in direct contact with the two helical segments of EF3 in WT recoverin. The two Ca 2ϩ ions bind to Rv 2-190 non-cooperatively; the K D for EF3 shifts from 0.21 M in WT to 1.3 M in the mutant, whereas the K D for EF2 is unaltered (6.2 M) (16). Analysis of the Rv 2-190 titration data using the Hill equation yields a Hill coefficient, n, of 0.86 (our analysis of their data), which is consistent with a model for two independent binding sites. Importantly, the effect of the ⌬191-202 truncation is preserved in the myristoylated protein; binding of Ca 2ϩ to mRv  is cooperative, as expected, but the affinity (K d ϭ 30 M) is less than for WT mRv (K d ϭ 14 M) as is the Hill coefficient (WT mRv, n ϭ 1.5; mRv  , n ϭ 1.3). These results are consistent with a direct effect of the mutation on the intrinsic affinity of EF3 for Ca 2ϩ .
In contrast, we have concluded that the apparent decreased Ca 2ϩ affinity in the C39A mutant results from an indirect effect of the mutation on the T/R equilibrium with essentially no change in the intrinsic K D of the two binding sites for the following reasons: (i) Cys-39 is located in the EF1 loop, close to the myristoyl-binding pocket and close to the site of other mutations known to affect the T/R equilibrium (e.g. W31K; see  below), with no obvious connection to EF3 or the C-terminal residues in contact with EF3 in either the Ca 2ϩ -free (34) or Ca 2ϩ -bound (15,35) proteins. (ii) The structure of Ca 2ϩ -bound EF3 is the same in the C39A mutant as it is in WT Rv (r.m.s.d. ϭ 0.04 Å for residues 110 -121; Fig. 5B), suggesting that the mutation affects the Ca 2ϩ -free protein. (iii) The Ca 2ϩ -binding isotherm for C39A Rv is well fit with the Hill equation using a single K d and n Ͼ 1 (n ϭ 1.2), indicating that binding of the two Ca 2ϩ ions is cooperative (Fig. 2). (iv) The isotherms for binding of Ca 2ϩ to WT mRv and C39A mRv are indistinguishable within experimental error (Fig. 2), indicating that myristoylation eliminates the effect of the C39A mutation.
Whereas this last result is clearly different from that found for mRv  , as discussed above, and inconsistent with a direct effect of the mutation on the binding site in EF3, we must note that a direct effect of mutation on the T/R equilibrium (to favor the T state) would also be expected to express itself in the myristoyated protein. The fact that an effect is not observed in the myristoylated mutant may indicate that the C39A mutation has an adverse effect on sequestration of the myristoyl chain in the T state as well as an effect promoting the T state in the unmyristoylated protein. Although there is an ad hoc element to this suggestion, we note that the site of mutation in the N-terminal domain is close to the myristoyl-binding pocket.
In aggregate, these data are most easily interpreted to indicate that the C39A mutation restores cooperative binding of Ca 2ϩ to Rv because it shifts the T/R equilibrium in favor of T without significantly affecting the intrinsic Ca 2ϩ affinity of EF2 and EF3. Baldwin and Ames (23) have reported two different mutants (W31K and I52A/Y53A) in which the primary effect of the mutations is on the T/R equilibrium. However, in their case the mutations shift the equilibrium in favor of the R state, increasing affinity of the myristoylated protein for Ca 2ϩ with little effect on the non-myristoylated forms. To our knowledge, C39A is the first recoverin mutant for which the binding of Ca 2ϩ to the non-myristoylated protein is cooperative. As such, the mutant should be useful for elucidating the atomic details stabilizing each conformation and thus controlling the position of the T/R equilibrium, especially given the fact that the mutation does not affect downstream interaction of the Ca 2ϩ -bound protein with the RGS domain of RK (Fig. 3).
Finally, the crystal structure of WT Rv under non-reducing conditions shows that the sulfur atom of Cys-39 can oxidize to sulfenic acid (Fig. 6B). These data are of interest because Cys-39 is the only cysteine residue in recoverin and is strictly conserved among all members of the NCS family. Permyakov et al. showed that recoverin lost the ability to react with the thiol-specific Ellman's reagent (5,5Ј-dithiobis-(2-nitrobenzoic acid)) upon incubation under non-reducing conditions, suggesting that the sulfur of Cys-39 underwent oxidation (17). For this reason, noting also that the retina is one of the most vascularized tissues in the body, they further proposed that Cys-39 may function as a redox sensor in rod photoreceptor cells. Trying to mimic higher oxidation states of the sulfur atom, they made the C39D mutant and demonstrated that the Ca 2ϩ -dependent interaction of the mutant mRv with rod outer segment membranes was significantly disrupted by the mutation (18). These data agree well with our results showing that C39D Rv does not bind RGS in a pulldown assay for interaction of the two proteins (Fig. 4C). Further work will be required to understand fully the possible role Cys-39 may play as a redox sensor in photoreceptor cells,  but it is clear at this point that the sulfur atom can sense environmental redox conditions.