Calcium-regulated DNA binding and oligomerization of the neuronal calcium-sensing protein, calsenilin/DREAM/KChIP3.

Calsenilin/DREAM/KChIP3, a member of the recoverin branch of the EF-hand superfamily, interacts with presenilins, serves as a calcium-regulated transcriptional repressor, and interacts with A-type potassium channels. Here we report physicochemical characterization of calcium binding, oligomerization, and DNA binding of human calsenilin/DREAM/KChIP3. Equilibrium Ca(2+) binding measurements indicate that the protein binds 3 Ca(2+) with a dissociation constant of 14 microM and a Hill coefficient of 0.7. Dynamic light scattering and size exclusion chromatography show that the Ca(2+)-bound protein exists as a dimer at protein concentrations lower than 150 microM and forms a tetramer at concentrations above 200 microM. The Ca(2+)-free protein is a tetramer in the concentration range 20-450 microM. Isothermal titration calorimetry and dynamic light scattering indicate that the Ca(2+)-free protein tetramer binds endothermically (DeltaH = +25 kcal/mol) to four molecules of DNA derived from the downstream regulatory element (DRE) of either the prodynorphin or c-fos genes. One DRE molecule binds tightly to the protein with a dissociation constant (K(d)) of 75 nM, and the other three bind more weakly (K(d) = 640 nM). No significant DNA binding was observed for the Ca(2+)-bound protein. The N-terminal protein fragment (residues 1-70) binds nonspecifically to DRE in a Ca(2+)-independent manner, whereas a C-terminal fragment containing the four EF-hands (residues 65-256) binds DRE (K(d) = 200 nM) in a Ca(2+)-regulated and sequence-specific fashion. The C-terminal fragment is a tetramer in the Ca(2+)-free state and dissociates into dimers at saturating Ca(2+) levels.

Structure and calcium-binding properties of Frq1, a novel calcium sensor in the yeast Saccharomyces cerevisiae.

The FRQ1 gene is essential for growth of budding yeast and encodes a 190-residue, N-myristoylated (myr) calcium-binding protein. Frq1 belongs to the recoverin/frequenin branch of the EF-hand superfamily and regulates a yeast phosphatidylinositol 4-kinase isoform. Conformational changes in Frq1 due to N-myristoylation and Ca(2+) binding were assessed by nuclear magnetic resonance (NMR), fluorescence, and equilibrium Ca(2+)-binding measurements. For this purpose, Frq1 and myr-Frq1 were expressed in and purified from Escherichia coli. At saturation, Frq1 bound three Ca(2+) ions at independent sites, which correspond to the second, third, and fourth EF-hand motifs in the protein. Affinity of the second site (K(d) = 10 microM) was much weaker than that of the third and fourth sites (K(d) = 0.4 microM). Myr-Frq1 bound Ca(2+) with a K(d)app of 3 microM and a positive Hill coefficient (n = 1.25), suggesting that the N-myristoyl group confers some degree of cooperativity in Ca(2+) binding, as seen previously in recoverin. Both the NMR and fluorescence spectra of Frq1 exhibited very large Ca(2+)-dependent differences, indicating major conformational changes induced upon Ca(2+) binding. Nearly complete sequence-specific NMR assignments were obtained for the entire carboxy-terminal domain (residues K100-I190). Assignments were made for 20% of the residues in the amino-terminal domain; unassigned residues exhibited very broad NMR signals, most likely due to Frq1 dimerization. NMR chemical shifts and nuclear Overhauser effect (NOE) patterns of Ca(2+)-bound Frq1 were very similar to those of Ca(2+)-bound recoverin, suggesting that the overall structure of Frq1 resembles that of recoverin. A model of the three-dimensional structure of Ca(2+)-bound Frq1 is presented based on the NMR data and homology to recoverin. N-myristoylation of Frq1 had little or no effect on its NMR and fluorescence spectra, suggesting that the myristoyl moiety does not significantly alter Frq1 structure. Correspondingly, the NMR chemical shifts for the myristoyl group in both Ca(2+)-free and Ca(2+)-bound myr-Frq1 were nearly identical to those of free myristate in solution, indicating that the fatty acyl chain is solvent-exposed and not sequestered within the hydrophobic core of the protein, unlike the myristoyl group in Ca(2+)-free recoverin. Subcellular fractionation experiments showed that both the N-myristoyl group and Ca(2+)-binding contribute to the ability of Frq1 to associate with membranes.

Ca(2+)-binding proteins in the retina: structure, function, and the etiology of human visual diseases.

The complex sensation of vision begins with the relatively simple photoisomerization of the visual pigment chromophore 11-cis-retinal to its all-trans configuration. This event initiates a series of biochemical reactions that are collectively referred to as phototransduction, which ultimately lead to a change in the electrochemical signaling of the photoreceptor cell. To operate in a wide range of light intensities, however, the phototransduction pathway must allow for adjustments to background light. These take place through physiological adaptation processes that rely primarily on Ca(2+) ions. While Ca(2+) may modulate some activities directly, it is more often the case that Ca(2+)-binding proteins mediate between transient changes in the concentration of Ca(2+) and the adaptation processes that are associated with phototransduction. Recently, combined genetic, physiological, and biochemical analyses have yielded new insights about the properties and functions of many phototransduction-specific components, including some novel Ca(2+)-binding proteins. Understanding these Ca(2+)-binding proteins will provide a more complete picture of visual transduction, including the mechanisms associated with adaptation, and of related degenerative diseases.

Diversity of conformational states and changes within the EF-hand protein superfamily.

The EF-hand motif, which assumes a helix-loop-helix structure normally responsible for Ca2+ binding, is found in a large number of functionally diverse Ca2+ binding proteins collectively known as the EF-hand protein superfamily. In many superfamily members, Ca2+ binding induces a conformational change in the EF-hand motif, leading to the activation or inactivation of target proteins. In calmodulin and troponin C, this is described as a change from the closed conformational state in the absence of Ca2+ to the open conformational state in its presence. It is now clear from structures of other EF-hand proteins that this "closed-to-open" conformational transition is not the sole model for EF-hand protein structural response to Ca2+. More complex modes of conformational change are observed in EF-hand proteins that interact with a covalently attached acyl group (e.g., recoverin) and in those that dimerize (e.g., S100B, calpain). In fact, EF-hand proteins display a multitude of unique conformational states, together constituting a conformational continuum. Using a quantitative 3D approach termed vector geometry mapping (VGM), we discuss this tertiary structural diversity of EF-hand proteins and its correlation with target recognition.

Three-dimensional structure of guanylyl cyclase activating protein-2, a calcium-sensitive modulator of photoreceptor guanylyl cyclases.

Guanylyl cyclase activating protein-2 (GCAP-2) is a Ca2+-sensitive regulator of phototransduction in retinal photoreceptor cells. GCAP-2 activates retinal guanylyl cyclases at low Ca2+ concentration (<100 nM) and inhibits them at high Ca2+ (>500 nM). The light-induced lowering of the Ca2+ level from approximately 500 nM in the dark to approximately 50 nM following illumination is known to play a key role in visual recovery and adaptation. We report here the three-dimensional structure of unmyristoylated GCAP-2 with three bound Ca2+ ions as determined by nuclear magnetic resonance spectroscopy of recombinant, isotopically labeled protein. GCAP-2 contains four EF-hand motifs arranged in a compact tandem array like that seen previously in recoverin. The root mean square deviation of the main chain atoms in the EF-hand regions is 2.2 A in comparing the Ca2+-bound structures of GCAP-2 and recoverin. EF-1, as in recoverin, does not bind calcium because it contains a disabling Cys-Pro sequence. GCAP-2 differs from recoverin in that the calcium ion binds to EF-4 in addition to EF-2 and EF-3. A prominent exposed patch of hydrophobic residues formed by EF-1 and EF-2 (Leu24, Trp27, Phe31, Phe45, Phe48, Phe49, Tyr81, Val82, Leu85, and Leu89) may serve as a target-binding site for the transmission of calcium signals to guanylyl cyclase.

Core mutations that promote the calcium-induced allosteric transition of bovine recoverin.

Recoverin is a small calcium binding protein involved in regulation of the phototransduction cascade in retinal rod cells. It functions as a calcium sensor by undergoing a cooperative, ligand-dependent conformational change, resulting in the extrusion of the N-terminal myristoyl group from a hydrophobic pocket. To test the role of certain core residues in tuning this allosteric switch, we have made and characterized two mutants: W31K, which replaces Trp31 with Lys; and a double mutant, I52A/Y53A, in which Ile52 and Tyr53 are both replaced by Ala. These mutations decrease the hydrophobicity of the myristoyl binding pocket. They are thus expected to make sequestering of the myristoyl group less favorable and destabilize the Ca2+-free state. As predicted, the myristoylated forms of the mutants exhibit increased affinity for Ca2+, whether monitored by equilibrium binding of 45Ca2+ (Kd = 17.2, 7.9, and 8.1 microM for wild type, W31K, and I52A/Y53A, respectively) or by the change in tryptophan fluorescence associated with the conformational change (Kd = 17.9, 3.6, and 4.4 microM for wild type, W31K, and I52A/Y53A, respectively). The mutants also exhibit decreased cooperativity of binding (Hill coefficient = 1.2 and 1.0 for W31K and I52A/Y53A vs 1. 4 for wild type). Binding of the mutant proteins to rod outer segment membranes occurs at lower Ca2+ concentrations compared to wild-type protein (K1/2 = 5.6, 2.2, and 1.0 microM for wild type, W31K, and I52A/Y53A, respectively). The unmyristoylated forms of the mutants exhibit biphasic Ca2+ binding curves, nearly identical to that observed for wild type. The binding data for the two mutants can be explained by a concerted allosteric model in which the mutations affect only the equilibrium constant L between the two allosteric forms, T (the Ca2+-free form) and R (the Ca2+-bound form), without affecting the intrinsic binding constants for the two Ca2+ sites. Two-dimensional NMR spectra of the Ca2+-free forms of the mutants have been compared to the wild-type spectrum, whose peaks have been assigned to specific residues (1). Many resonances assigned to residues in the C-terminal domain (residues 100-202) in the wild-type spectrum are identical in the mutant spectra, suggesting that the backbone structure of the C-terminal domain is probably unchanged in both mutants. The N-terminal domain, in which both mutations are located, reveals in each case numerous changes of undetermined spatial extent.

Differential isotype labeling strategy for determining the structure of myristoylated recoverin by NMR spectroscopy.

The three-dimensional solution structure of recombinant bovine myristoylated recoverin in the Ca(2+)-free state has been refined using an array of isotope-assisted multidimensional heteronuclear NMR techniques. In some experiments, the myristoyl group covalently attached to the protein N-terminus was labeled with C and the protein was unlabeled or vice versa; in others, both were C-labeled. This differential labeling strategy was essential for structural refinement and can be applied to other acylated proteins. Stereospecific assignments of 41 pairs of beta-methylene protons and 48 methyl groups of valine and leucine were included in the structure refinement. The refined structure was constructed using a total of 3679 experimental NMR restraints, comprising 3242 approximate interproton distance restraints (including 153 between the myristoyl group and the polypeptide), 140 distance restraints for 70 backbone hydrogen bonds, and 297 torsion angle restraints. The atomic rms deviations about the average minimized coordinate positions for the secondary structure region of the N-terminal and C-terminal domains are 0.44 +/- 0.07 and 0.55 +/- 0.18 A for backbone atoms, and the 1.09 +/- 0.07 and 1.10 +/- 0.15 A for all heavy atoms, respectively. The refined structure allows for a detailed analysis of the myristoyl binding pocket. The myristoyl group is in a slightly bent conformation: the average distance between C1 and C14 atoms of the myristoyl group is 14.6 A. Hydrophobic residues Leu28, Trp31, and Tyr32 from a cluster that interacts with the front end of the myristoyl (C1-C8), whereas residues Phe49, Phe56, Tyr86, Val87, and Leu90 interact with the tail end (C9-C14). The relatively deep hydrophobic pocket that binds the myristoyl group (C14:0) could also accommodate other naturally occurring acyl groups such as C12:0, C14:1, C14:2 chains.

Lateral self-assembly of E-cadherin directed by cooperative calcium binding.

We report the Ca2+ binding characteristics of recombinant Ecad12, a construct spanning the first two repeats of epithelial cadherin, and demonstrate the links between Ca2+ binding and dimer formation. Sedimentation equilibrium and dynamic light scattering experiments show that weak dimerization of Ecad12 occurs in the presence of 10 mM Ca2+ (KdP = 0.17 mM), while no appreciable dimer formation was detected in the absence of Ca2+. Ca2+-induced dimerization was also observed in electron microscopy images of Ecad12. We conclude from Ca2+ titration experiments monitored by tryptophan fluorescence and flow dialysis that dimerization does not affect the equilibrium binding constant for Ca2+. However, the value of the Hill coefficient for Ca2+ binding increases from 1.5 to 2.4 as the protein concentration increases, showing that dimer formation largely contributes to the cooperativity in Ca2+ binding. Based on these observations and previous crystallographic studies, we propose that calcium acts more likely as a geometrical aligner ensuring the proper assembly of cadherin molecules, rather than a simple adhesive.

Molecular mechanics of calcium-myristoyl switches.

Many eukaryotic cellular and viral proteins have a covalently attached myristoyl group at the amino terminus. One such protein is recoverin, a calcium sensor in retinal rod cells, which controls the lifetime of photoexcited rhodopsin by inhibiting rhodopsin kinase. Recoverin has a relative molecular mass of 23,000 (M[r] 23K), and contains an amino-terminal myristoyl group (or related acyl group) and four EF hands. The binding of two Ca2+ ions to recoverin leads to its translocation from the cytosol to the disc membrane. In the Ca2+-free state, the myristoyl group is sequestered in a deep hydrophobic box, where it is clamped by multiple residues contributed by three of the EF hands. We have used nuclear magnetic resonance to show that Ca2+ induces the unclamping and extrusion of the myristoyl group, enabling it to interact with a lipid bilayer membrane. The transition is also accompanied by a 45-degree rotation of the amino-terminal domain relative to the carboxy-terminal domain, and many hydrophobic residues are exposed. The conservation of the myristoyl binding site and two swivels in recoverin homologues from yeast to humans indicates that calcium-myristoyl switches are ancient devices for controlling calcium-sensitive processes.

Portrait of a myristoyl switch protein.

Myristoylated proteins transduce a diverse range of cellular signals. Recoverin is a myristoylated, calcium-binding protein in the retina that serves as a calcium sensor in vision. The recent elucidation of the structures of several forms of myristoylated and unmyristoylated recoverin provides insight into how calcium induces the binding of recoverin to membranes.

Nuclear magnetic resonance evidence for Ca(2+)-induced extrusion of the myristoyl group of recoverin.

Recoverin, a recently discovered member of the EF-hand protein superfamily, serves as a Ca2+ sensor in vision. A myristoyl or related N-acyl group covalently attached to the amino terminus of recoverin enables it to translocate to retinal disc membranes when the Ca2+ level is elevated. Two-dimensional 1H-13C shift correlation NMR spectra of recoverin containing a 13C-labeled myristoyl group were obtained to selectively probe the effect of Ca2+ on the environment of the attached myristoyl group. In the Ca(2+)-free state, each pair of methylene protons bonded to carbon atoms 2, 3, 11, and 12 of the myristoyl group gives rise to two peaks. The splittings, caused by nonequivalent methylene proton chemical shifts, indicate that the myristoyl group interacts intimately with the protein in the Ca(2+)-free state. By contrast, only one peak is seen for each pair of methylene protons in the Ca(2+)-bound state, indicating that the myristoyl group is located in an isotropic environment in this form. Furthermore, the 1H-13C shift correlation NMR spectrum of Ca(2+)-bound recoverin is very similar to that of myristic acid in solution. 1H-(13)C shift correlation NMR experiments were also performed with 13C-labeled recoverin to selectively probe the resonances of methyl groups in the hydrophobic core of the protein. The spectrum of Ca(2+)-bound myristoylated recoverin is different from that of Ca(2+)-free myristoylated recoverin but similar to that of Ca(2+)-bound unmyristoylated recoverin. Hence, the myristoyl group interacts little with the hydrophobic core of myristoylated recoverin in the Ca(2+)-bound state. Three-dimensional (13C/F1)-edited (13C/F3)-filtered heteronuclear multiple quantum correlation-nuclear Overhauser effect spectroscopy spectra of recoverin containing a 13C-labeled myristoyl group were obtained to selectively probe protein residues located within 5 A of the myristoyl group. The myristoyl group makes close contact with a number of aromatic residues in Ca(2+)-free recoverin, whereas the myristoyl group makes no observable contacts with the protein in the Ca(2+)-bound state. These NMR data demonstrate that the binding of Ca2+ to recoverin induces the extrusion of its myristoyl group into the solvent, which would enable it to interact with a lipid bilayer or a hydrophobic site of a target protein.

Sequestration of the membrane-targeting myristoyl group of recoverin in the calcium-free state.

Recoverin, a retinal calcium-binding protein of relative molecular mass (M(r)) 23K, participates in the recovery phase of visual excitation and in adaptation to background light. The Ca(2+)-bound form of recoverin prolongs the photoresponse, probably by blocking phosphorylation of photoexcited rhodopsin. Retinal recoverin contains a covalently attached myristoyl group or related acyl group at its amino terminus and two Ca(2+)-binding sites. Ca2+ binding to myristoylated, but not unmyristoylated, recoverin induces its translocation to bilayer membranes, indicating that the myristoyl group is essential to the read-out of calcium signals (calcium-myristoyl switch). Here we present the solution structure of Ca(2+)-free, myristoylated recombinant recoverin obtained by heteronuclear multidimensional NMR spectroscopy. The myristoyl group is sequestered in a deep hydrophobic pocket formed by many aromatic and other hydrophobic residues from five flanking helices.

Amino-terminal myristoylation induces cooperative calcium binding to recoverin.

Recoverin, a new member of the EF-hand protein superfamily, serves as a Ca2+ sensor in vision. A myristoyl or related N-acyl group covalently attached to the amino terminus of recoverin enables it to bind to disc membranes when the Ca2+ level is elevated. Ca(2+)-bound recoverin prolongs the lifetime of photoexcited rhodopsin, most likely by blocking its phosphorylation. We report here Ca2+ binding studies of myristoylated and unmyristoylated recombinant recoverin using flow dialysis, fluorescence, and NMR spectroscopy. Unmyristoylated recoverin exhibits heterogeneous and uncooperative binding of two Ca2+ with dissociation constants of 0.11 and 6.9 microM. In contrast, two Ca2+ bind cooperatively to myristoylated recoverin with a Hill coefficient of 1.75 and an apparent dissociation constant of 17 microM. Thus, the attached myristoyl group lowers the calcium affinity of the protein and induces cooperativity in Ca2+ binding. One-dimensional 1H and two-dimensional 15N-1H shift correlation NMR spectra of myristoylated recoverin measured as a function of Ca2+ concentration show that a concerted conformational change occurs when two Ca2+ are bound. The Ca2+ binding and NMR data can be fit to a concerted allosteric model in which the two Ca2+ binding sites have different affinities in both the T and R states. The T and R conformational states are defined in terms of the Ca(2+)-myristoyl switch; in the T state, the myristoyl group is sequestered inside the protein, whereas in the R state, the myristoyl group is extruded. Ca2+ binds to the R state at least 10,000-fold more tightly than to T. In this model, the dissociation constants of the two sites in the R state of the myristoylated protein are 0.11 and 6.9 microM, as in unmyristoylated recoverin. The ratio of the unliganded form of T to that of R is estimated to be 400 for myristoylated and < 0.05 for unmyristoylated recoverin. Thus, the attached myristoyl group has two related roles: it shifts the T/R ratio of the unliganded protein more than 8000-fold, and serves as a membrane anchor for the fully liganded protein.

Secondary structure of myristoylated recoverin determined by three-dimensional heteronuclear NMR: implications for the calcium-myristoyl switch.

Recoverin, a new member of the EF-hand superfamily, serves as a Ca2+ sensor in vision. A myristoyl or related N-acyl group is covalently attached at its N-terminus and plays an essential role in Ca(2+)-dependent membrane targeting by a novel calcium-myristoyl switch mechanism. The structure of unmyristoylated recoverin containing a single bound Ca2+ has recently been solved by X-ray crystallography [Flaherty, K. M., Zozulya, S., Stryer, L., & McKay, D. B. (1993) Cell 75, 709-716]. We report here multidimensional heteronuclear NMR studies on Ca(2+)-free, myristoylated recoverin (201 residues, 23 kDa). Complete polypeptide backbone 1H, 15N, and 13C resonance assignments and secondary structure are presented. We find 11 helical segments and two pairs of antiparallel beta-sheets, in accord with the four EF-hands seen in the crystal structure. The present NMR study also reveals some distinct structural features of the Ca(2+)-free myristoylated protein. The N-terminal helix of EF-2 is flexible in the myristoylated Ca(2+)-free protein, whereas it has a well-defined structure in the unmyristoylated Ca(2+)-bound form. This difference suggests that the binding of Ca2+ to EF-3 induces EF-2 to adopt a conformation favorable for the binding of a second Ca2+ to recoverin. Furthermore, the N-terminal helix (K5-E16) of myristoylated Ca(2+)-free recoverin is significantly longer than that seen in the unmyristoylated Ca(2+)-bound protein. We propose that this helix is stabilized by the attached myristoyl group and may play a role in sequestering the myristoyl group within the protein in the Ca(2+)-free state.

Resonance Raman study of halorhodopsin photocycle kinetics, chromophore structure, and chloride-pumping mechanism.

Kinetic resonance Raman spectra of the HR520, HR640, and HR578 species in the halorhodopsin photocycle are obtained using time delays ranging from 5 microseconds to 10 ms in 0.3 M NO3-, 0.3 M Cl-, and 3 M Cl-. The Raman intensities are converted to absolute concentrations by using a conservation of molecules constraint. The simplest kinetic scheme that satisfactorily models the data is HR578-->HR520 in equilibrium with HR640-->HR578. The rate constant for the HR640-->HR578 transition increases with Cl- concentration, suggesting that Cl- is taken up between HR640 and HR578. The ratio of the forward to the reverse rate constants connecting HR520 and HR640 increases as the inverse of the Cl- concentration, suggesting that Cl- is released during the HR520-->HR640 step. The configuration about the C13 = C14 bond of the retinal chromophore in HR640 is examined by regenerating the protein with [12,14-2H2]retinal. The C12-2H + C14-2H rocking vibration for HR640 is observed at 943 cm-1, demonstrating that the chromophore is 13-cis. The changes in the resonance Raman spectrum of HR640 in response to 2H2O suspension indicates that the Schiff base linkage to the protein is protonated. None of the HR640 fingerprint vibrations shift significantly in 2H2O, suggesting that the Schiff base adopts a C = N anti configuration; this assignment is supported by the frequency of the C15-2H rocking mode (1002 cm-1). The 13-cis structure for the chromophore in HR640 requires that thermal isomerization back to all-trans occurs in the HR640-->HR578 transition. These structural and kinetic results are incorporated into a two-state C-T model for Cl- pumping.

Time-resolved ultraviolet resonance Raman studies of protein structure: application to bacteriorhodopsin.

Time-resolved ultraviolet resonance Raman spectra of bacteriorhodopsin are used to study protein structural changes on the nanosecond and millisecond time scales. Excitation at 240 nm is used to selectively enhance vibrational scattering from tyrosine so that changes in its hydrogen bonding and protonation state can be examined. Both nanosecond and millisecond UV Raman difference spectra indicate that none of the tyrosine residues change ionization state during the BR----K and BR----M transitions. However, intensity changes are observed at 1172 and 1615 cm-1 in the BR----M UV Raman difference spectra. The 1615-cm-1 feature shifts down 25 cm-1 in tyrosine-d4-labeled BR, consistent with its assignment as a tyrosine vibration. The intensity changes in the BR----M UV Raman difference spectra most likely reflect an increase in resonance enhancement that occurs when one or more tyrosine residues interact more strongly with a hydrogen-bond acceptor in M412. The frequency of the v7a feature (1172 cm-1) in the BR----M UV Raman difference spectra supports this interpretation. The proximity of Tyr-185 and Asp-212 in the retinal binding pocket suggests that deprotonation of the Schiff base in M412 causes Tyr-185 to stabilize ionized Asp-212 by forming a stronger hydrogen bond.

The role of back-reactions and proton uptake during the N----O transition in bacteriorhodopsin's photocycle: a kinetic resonance Raman study.

The kinetics of bacteriorhodopsin's photocycle have been analyzed at pH 5, 6, 7, 8, and 8.6 by using time-resolved resonance Raman spectroscopy. The concentrations of the various intermediates as a function of time were determined by following their resonance Raman intensities using 502-nm (L550, N550, BR568), 458-nm (M412), and 752-nm (O640) excitation. The spectral contributions to the pump + probe data from each intermediate were quantitatively separated by least-squares decomposition. These relative concentrations were then converted to absolute concentrations by using a conservation of molecules constraint. This enabled the unambiguous refinement of a variety of kinetic models to find the simplest one that accurately describes the data. The kinetic data, including the biphasic decay of L550 and M412, are best reproduced by a sequential scheme including back-reactions (BR----L----M----N----O----BR). In addition, the kinetics of the L----M and N----O steps are found to be pH-dependent. Both the forward and reverse rate constants connecting L550 and M412 increase with pH, confirming earlier proposals of catalyzed Schiff base deprotonation at alkaline pH. Below pH 7, the N550----O640 rate constant is independent of pH, but it decreases linearly with pH above 7. This indicates that the protein must pick up a proton during the N550----O640 transition and that this process becomes rate determining above pH 7. There must, therefore, be an intermediate between N550 and O640 which we denote as N+550. A molecular graphics model is presented which incorporates these observations into a mechanism for proton pumping.