Structural Characterization of Ferrous Ion Binding to Retinal Guanylate Cyclase Activator Protein-5 from Zebrafish Photoreceptors

Abstract

Sensory guanylate cyclases (zGCs) in zebrafish photoreceptors are regulated by a family of guanylate cyclase activator proteins (called GCAP1-7). GCAP5 contains two non-conserved cysteine residues (Cys15 and Cys17) that could in principle bind to biologically active transition state metal ions (Zn²⁺ and Fe²⁺). Here, we present nuclear magnetic resonance (NMR) and isothermal titration calorimetry (ITC) binding analysis that demonstrate the binding of one Fe²⁺ ion to two GCAP5 molecules (in a 1:2 complex) with a dissociation constant in the nanomolar range. At least one other Fe²⁺ binds to GCAP5 with micromolar affinity that likely represents electrostatic Fe²⁺ binding to the EF-hand loops. The GCAP5 double mutant (C15A/C17A) lacks nanomolar binding to Fe²⁺, suggesting that Fe²⁺ at this site is ligated directly by thiolate groups of Cys15 and Cys17. Size exclusion chromatography analysis indicates that GCAP5 forms a dimer in both the Fe²⁺-free and Fe²⁺-bound states. NMR structural analysis and molecular docking studies suggest that a single Fe²⁺ ion is chelated by thiol side chains from Cys15 and Cys17 in the GCAP5 dimer, forming a [Fe(SCys)₄] complex like that observed previously in two-iron superoxide reductases. Fe²⁺ binding to GCAP5 decreases its ability to activate photoreceptor human GC-E by decreasing GC-activity more than 10-fold. Our results indicate a strong Fe²⁺-induced inhibition of GC by GCAP5 and suggest that GCAP5 may serve as a redox sensor in visual phototransduction.

Graphical abstract

Introduction

Retinal guanylate cyclase activator proteins (GCAP1¹ and GCAP2²) were first identified as EF-hand calcium sensor proteins in mammalian photoreceptor rod and cone cells³. Both GCAP1 and GCAP2 control Ca²⁺-sensitive activation of retinal guanylate cyclases (RetGCs^{4, 5}) that is crucial for promoting the recovery phase of vertebrate visual phototransduction^{6, 7}. Additional GCAP homologs were discovered recently in zebrafish (GCAP3-5 and 7⁸). In situ hybridization showed that expression of zebrafish specific GCs and GCAPs starts between 3 and 4 days post fertilization coinciding with the onset of visual function^{9, 10}. Initial in vitro studies further demonstrated that GCAPs activate mammalian GCs in a step-by-step Ca²⁺-relay mode fashion similar to the GC-GCAP system in mammals^11-13. Interestingly, these earlier studies showed that GCAP5 is a much weaker GC activator compared to the other GCAPs under standard assay conditions¹³. Furthermore, GCAP5 has the most divergent amino acid sequence compared to mammalian and other zebrafish GCAP forms (Fig. 1). In particular, GCAP5 contains two non-conserved cysteine residues (Cys15 and Cys17) not found in the other GCAPs that in principle could be used to ligate biological transition metals such as Fe²⁺ and Zn²⁺. Ferrous ion has been shown to serve as a redox sensor in a variety of cell types ¹⁴, and it is tempting to speculate that redox sensing by GCAP5 might control phototransduction in zebrafish photoreceptors. Indeed, the accumulation of iron levels in the retina have been correlated to age-related macular degeneration in humans, suggesting the involvement of redox-sensitive processes in the pathogenesis of the disease¹⁵.

Figure 1. Amino acid sequence alignment of GCAP proteins.

EF-hand motifs are shaded in color (EF1 green, EF2 red, EF3 cyan and EF4 yellow). Non-conserved cysteine residues (Cys15 and Cys17) in GCAP5 are highlighted in bold and red. Swiss Protein Database accession numbers are Q90WX4 (zebrafish GCAP1), Q90WX3 (zebrafish GCAP2), Q8UUX9 (zebrafish GCAP3), Q6ZM98 (zebrafish GCAP4), and Q5MAC8 (zebrafish GCAP5).

In the current study, we characterized both the structure and binding properties of Fe²⁺ binding to GCAP5 in zebrafish. Our ITC binding analysis demonstrates that one Fe²⁺ binds to two molecules of GCAP5 with nanomolar affinity, consistent with one Fe²⁺ bound per GCAP5 dimer. At least one additional Fe²⁺ binds to GCAP5 in the micromolar range that likely binds to the second EF-hand. NMR structural studies reveal that both Cys15 and Cys17 are essential for nanomolar Fe²⁺ binding to GCAP5, which form a [Fe(SCys)₄] complex. Our functional studies indicate that Ca²⁺-free/Fe²⁺-bound and Ca²⁺-bound/Fe²⁺-bound GCAP5 both strongly inhibit mammalian GC activity compared to that of Fe²⁺-free GCAP5. We suggest that Fe²⁺ levels in photoreceptors (and hence metabolic activity) may be controlled by light activation, and therefore propose that Fe²⁺ binding to GCAP5 may serve as a light-dependent redox sensor in zebrafish photoreceptors.

Materials and Methods

Cloning, expression and purification of GCAP5 forms

Recombinant myristoylated GCAP5 (hereafter designated as GCAP5) was used throughout this study and bacterial expression of myristoylated GCAP5 was accomplished by co-expressing the GCAP5 D3N mutant¹⁶ and yeast N-myristoyl CoA transferase (NMT) in E. coli strain, BL21(DE3) as described previously¹⁶ for GCAP1¹⁷. Acylation of the GCAP5 D3N mutant was originally confirmed in living cells by a click-chemistry approach¹⁶. Cloning of the C15A/C17A double mutant was similar as described for other point mutants of GCAPs^{11, 13}. Accordingly, the double mutant was prepared employing primers 5′-CTCAGCGCCgcCAAAgcCCACCAGTGG-3′ (forward) and 5′-CTCGGTGGCGGACATGCTGG-3′ (reverse). Single point mutations were introduced to GCAP5 using the same reverse primer and 5′-CTCAGCGCCgcCAAATGCCACCAGTGG-3′ (forward) for the C15A mutant and 5′-CTCAGCGCCTGCAAAgcCCACCAGTGG-3′ (forward) for the C17A mutant. Purification of GCAP5 was achieved using previously described methods with the following modifications^{18, 19}. Briefly, myristoylated GCAP5 was expressed in the soluble fraction, in contrast to unmyristoylated GCAP5 that was expressed in inclusion bodies. The cell lysate supernatant with the addition of 0.35 M NH₄SO₄ (final concentration) was loaded onto a Butyl sepharose column (HiPrep Butyl FF 16/10, GE Healthcare) equilibrated with loading buffer (20 mM Tris (pH7.5), 0.35M NH₄SO₄, and 1mM Dithiothreitol (DTT)), and the column was washed until the column wash contained no detectable protein (A₂₈₀ < 0.01). GCAP5 protein (both wild type, GCAP5^WT or C15A/C17A double mutant, GCAP5^C15A/C17A) was then eluted from the Butyl sepharose column with a low salt buffer (20 mM Tris (pH7.5), 2 mM ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA) and 1 mM DTT). The eluted protein fraction from the Butyl-sepharose column was then diluted 5-fold and directly applied to a Q sepharose column (HiTrap 5ml Q HP, GE Healthcare) equilibrated in low salt buffer, and GCAP5 was eluted using a standard salt gradient^{18, 19}. The eluted protein fraction from the Q-sepharose column was concentrated to a volume of 2 mL and then applied to a size exclusion chromatography column (HiLoad 26/600 Superdex, GE Healthcare) equilibrated in low salt buffer. The final GCAP5 protein was eluted into low salt buffer. Typically, 10 mg of final purified GCAP5 protein was obtained from 1 liter of cell culture. The final protein sample was more than 95% pure as judged by SDS-PAGE. Protein myristoylation was verified using reverse phase HPLC as described²⁰.

Analytical size exclusion chromatography (SEC)

Measurement of the molar mass of GCAP5 in the presence of Mg²⁺ (5 mM) and/or Fe²⁺ (1 mM) was performed using analytical SEC (Superdex 200 HR 10/30 column, GE Healthcare). A sample volume of 100 μL of GCAP5 (200 μM protein concentration) was applied to the column equilibrated with 30 mM MES (pH 6.6), 5 mM citrate, and 100 mM NaCl. The SEC measurements were made at 4 °C with a flow rate at 0.5 ml/min.

NMR spectroscopy

GCAP5 samples for NMR experiments consisted of ¹⁵N-labeled myristoylated and Ca²⁺-free GCAP5 (0.35 mM) dissolved in 30 mM MES (pH 6.6) buffer containing 1 mM DTT-d₁₀ and 90%:10% H₂O:D₂O. To observe Fe²⁺ ion binding to GCAP5 by NMR, a series of samples of ¹⁵N-labeled GCAP5^WT (and GCAP5^C15A/C17A) was prepared with the addition of 0, 0.25, 0.50, 1.0, 2.0, and 5.0 equivalents of FeSO₄ into the samples prepared anaerobically in a glove bag under constant purging with high purity argon gas. All NMR experiments were performed at 30°C on a Bruker 800 MHz Avance III spectrometer equipped with a triple resonance cryogenic TCI probe and pulsed field gradients. 2D ¹⁵N-¹H HSQC with 2048 (¹H) × 256 (¹⁵N) data points were acquired on ¹⁵N-labeled GCAP1^WT or GCAP5^C15A/C17A samples. Spectra were processed using NMRPipe software package²¹ and analyzed using SPARKY²².

Isothermal titration calorimetry (ITC)

Fe²⁺ binding to GCAP5^WT (and GCAP5^C15A/C17A) with ferrous gluconate was carried out on a VP-ITC calorimeter (Micro-Cal) at 25°C as described previously²³. Metal free GCAP5 protein was prepared by first adding 5 mM EGTA into the protein sample (to remove contaminating Ca²⁺) and residual EGTA was removed by performing repeated buffer exchange with Amicon spin concentrator. The ITC titration buffer contained 30 mM MES (pH 6.6), 100 mM NaCl, 2 mM ferrous gluconate, and 1 mM beta-mercaptoethanol. The concentration of GCAP5 in the titration was 50 μM and Fe²⁺ concentration in the titrant was 2 mM. The sample was titrated with 40 injections of 5 μl aliquots.

Molecular docking calculation

The web-based docking program HADDOCK²⁴ was used to generate a structural model of Fe²⁺-bound to GCAP5. First, a homology modeled structure of monomeric GCAP5 was generated based on the crystal structure of chicken GCAP1 (PDB ID: 2R2I; 63% sequence identity) using Swiss-Modeler software²⁵. This homology model of monomeric GCAP5 was then docked to a single Fe²⁺ using HADDOCK. For the Fe²⁺ docking, an unambiguous distance restraint between the bound Fe²⁺ ion and sulfur atoms of cysteine residues (Cys15 and Cys17) was set to 2.35 ±0.05 Å. This distance represents the average Fe-S bond length extracted from crystal structures of rubredoxin (PDB ID: 1FHM) and rubrerythrin (PDB ID: 1LKO). The docking calculation was initiated with a rigid body energy minimization that generated 1000 structures. The best 200 structures were subjected to a semi-flexible simulated annealing step. In the final step, the 200 structures obtained from the previous simulated annealing step were refined in explicit waters. The coordinate file with the lowest HADDOCK score was selected as the final model of Fe²⁺-bound GCAP5. The structure of the GCAP5 dimer with one Fe²⁺ bound was built using HADDOCK, in which the Fe²⁺-bound GCAP5 monomer (calculated above) was docked onto a second GCAP5 molecule. The same HADDOCK protocol above was used to dock the dimer. The distance between Fe²⁺ and each sulfur atom of the four cysteine residues in the dimer was set to 2.35 ± 0.05 Å. The distance between the four sulfur atoms in the dimer was set to 3.8 ± 0.1 Å, which matched the corresponding distances observed in the crystal structure of rubredoxin. At the end of the HADDOCK dimer calculation, 195 structures formed a single cluster out of 200 water-refined structures. The coordinate file with the lowest HADDOCK score was chosen for the final structural model displayed in this study.

Guanylate cyclase assays

For testing the regulatory properties of zGCAP5 and its mutant, we reconstituted purified GCAP5 forms with cell membranes containing heterologously expressed human GC-E in HEK flip 293 cells essentially as described previously^{26, 27}, but using a cell line that stably expressed GC-E. Cells were cultivated and harvested by centrifugation (300×g, 5 min), the supernatant was discarded and the pellet was resuspended in100 μL of 50 mM HEPES-KOH pH 7.4, 50 mM KCl, 20 mM NaCl, 1 mM DTT, mammalian protease inhibitor cocktail (1:500). Activities were measured according to a detailed protocol that was published before^{13, 26} with the following modifications: GCAP5 and bovine GCAP1 (control) were added at a final concentration of 5 - 10 μM and Ca²⁺-dependent GC activities were obtained by adjusting the free [Ca²⁺ ] using a K₂H₂EGTA/CaH₂EGTA buffer system as described previously¹³. The free Mg²⁺-concentration was 1 mM. For investigating the activity profile of Fe²⁺ bound GCAP5 we first resuspended lyophilized purified GCAP5 forms in 30 mM Mes buffer pH 6.6, 1 mM DTT. GCAP5 was saturated with Fe²⁺ by incubation with 100 μM Fe²⁺, and unbound Fe²⁺ was removed by dialysis against 30 mM Mops pH 7.2, 60 mM KCl, 4 mM NaCl, 3.5 mM MgCl₂, 1 mM DTT, which kept Fe²⁺ at a minimum concentration of 100 nM (FeSO₄). Iron-loaded GCAP5 forms were reconstituted with recombinant GC-E and the activity was tested in the absence and presence of free Ca²⁺ in Mops-buffer (30mM) at pH 7.2, 60 mM KCl, 4 mM NaCl, 3.5 mM MgCl₂, 1 mM GTP, 0.3 mM ATP, 0.16 mM Zaprinast and 5 mM DTT. To avoid interference of the Ca²⁺-chelating EGTA buffer with Fe²⁺, we lowered the EGTA concentration to 50 μM in the assay mixture.

Results

Fe²⁺ binding to GCAP5 monitored by ITC

The binding of Fe²⁺ to GCAP5 was initially detected visually when highly concentrated and purified GCAP5 protein appeared to have a faint brownish or yellow color. Two non-conserved cysteine residues in GCAP5 (Cys15 and Cys17, see Fig. 1) were suggested to participate in Fe²⁺ binding, because the double mutation C15A/C17A in GCAP5 (called GCAP5^C15A/C17A) abolished the faint color. The UV-visible absorbance spectrum of iron-loaded GCAP5 exhibited two peaks at 325 nm and 420 nm indicative of the presence of an iron-sulfur cluster²⁸. GCAP5 may also bind to Fe³⁺, and the observed faint yellow color could also be due to light absorption by Fe³⁺; however, Fe³⁺ binding to GCAP5 was not analyzed in this study, because Fe³⁺ under physiological conditions is not soluble enough for binding titrations by ITC or NMR.

Isothermal titration calorimetry (ITC) was used to quantitatively monitor Fe²⁺ binding to GCAP5 (Fig. 2). Titration of Fe²⁺ ions into wildtype GCAP5 (called GCAP5^WT) resulted in a binding isotherm that is multiphasic and could be fit by a two-site model with 2 or more Fe²⁺ ions (Figs. 2A-B). The binding isotherm in Fig. 2B exhibited two distinct phases: The first phase represents high affinity binding of Fe²⁺ (K_d < 100 nM) with a stoichiometry of 1 Fe²⁺ bound to two GCAP5 molecules (stoichiometric coefficient, n = 0.4 ±0.1). The enthalpy change (∆H) for the high affinity site could not be accurately measured due to confounding enthalpy changes caused by dissociation of Fe²⁺ complexes with gluconate and mercaptoethanol as well as redox dependent heat of dilution. The second phase in the binding isotherm represents exothermic binding of Fe²⁺ (∆H = -2.8 ±0.1 kcal/mol) with a stoichiometry of at least 2 Fe²⁺ bound to each GCAP5 molecule (n = 2 ±0.5) and apparent dissociation constant in the micromolar range (K_d = 3 ±1 μM).

Figure 2. GCAP5 binds to Fe ⁺ as measured by ITC.

Change of heat resulting from incremental addition of Fe²⁺ to GCAP5^WT (A) or GCAP5^C15A/C17A (C) during ITC titration at 25°C. Binding isotherms of Fe²⁺ binding to GCAP5^WT (B) or GCAP5^C15A/C17A (D) were derived from the integrated heat at each injection. The binding isotherm for GCAP5^WT (B) was fit to a two-site model (see Methods). The fit suggests a high affinity site with a dissociation constant (K_d) that is less than 100 nM and stoichiometry (n) equal to 0.4 ±0.1 equivalents of Fe²⁺ bound per GCAP5 polypeptide. The lower affinity site has an apparent K_d of 3 ±1 μM with a stoichiometric coefficient of 2 ±0.5 and ∆H = -2.8 ±0.1 kcal/mol. The binding isotherm for GCAP5^C15A/C17A (D) was fit to a onesite model with dissociation constant of 3 ±1 μ M and ∆H = -0.7 ±0.1 kcal/mol.

Titration of Fe²⁺ ions into the C15A/C17A double mutant (GCAP5^C15A/C17A) resulted in a monophasic exothermic binding isotherm that could be fit by a one-site model (Figs. 2C-D). The binding isotherm in Fig. 2D exhibited exothermic binding of Fe²⁺ (∆H = -0.7 ±0.1 kcal/mol) with a dissociation constant in the micromolar range (K_d = 3 ±1 μM), but the amplitude of the exothermic heat signal was approximately six-fold lower. The double mutant (GCAP5^C15A/C17A) eliminated the nanomolar Fe²⁺ binding site observed for GCAP5^WT, suggesting that Cys15 and Cys17 are both essential for nanomolar Fe²⁺ binding to GCAP5. The micromolar Fe²⁺ binding observed for GCAP5^C15A/C17A is possibly due to non-specific electrostatic interaction of Fe²⁺ with negatively charged EF-hand loops or negatively charged residues on the protein surface.

NMR titration of Fe²⁺ binding to GCAP5

To probe the structural effects of Fe²⁺ binding to GCAP5, ¹⁵N-¹H HSQC NMR spectra were recorded for ¹⁵N-labeled GCAP5 in the presence and absence of saturating Fe²⁺ concentration (Fig. 3). The HSQC spectrum of GCAP5 in the absence of Fe²⁺ (red peaks in Fig. 3) exhibits 191 well resolved peaks with uniform intensity, indicating that GCAP5 is stably folded under NMR conditions. Each peak in the HSQC spectrum represents a particular amide group with a unique chemical shift environment that provides a residue specific fingerprint of protein conformation. Thus, Fe²⁺-dependent spectral changes in the HSQC can be analyzed to infer particular amino acids at or near the Fe²⁺ binding sites. The HSQC spectrum of GCAP5 in the presence of saturating levels of Fe²⁺ (black peaks in Fig. 3) revealed more than a dozen peaks whose spectral frequency and/or intensity were altered by the addition of Fe²⁺. Fe²⁺-binding caused some peaks to become broadened beyond detection (see peaks at 7.6, 7.8, 8.05, 9.1, 9.3, and 10.6 ppm), whereas Fe²⁺ also caused a few new peaks to appear (6.4, 6.8 and 9.7 ppm). The large number of Fe²⁺-induced spectral changes suggests that GCAP5 has multiple Fe²⁺ binding sites, consistent with the Fe²⁺ binding heterogeneity observed in the ITC isotherm (Fig. 2B).

Figure 3. NMR spectroscopy of GCAP5 vs Fe²⁺.

¹H-¹⁵N HSQC spectra of ¹⁵N-labeled GCAP5 in the Fe²⁺-free (red) and Fe²⁺-saturated (black) states. The experimental conditions are defined under Materials and Methods and were the same as described previously for GCAP1¹⁷.

To discriminate the different Fe²⁺ binding sites, HSQC spectra of GCAP5^WT (Fig. 4A) and GCAP5^C15A/C17A (Fig. 4B) were recorded at many different Fe²⁺ concentrations. The NMR titration of GCAP5^WT (Fig 4A) revealed two types of binding sites: one type had two separate resolved peaks that represent Fe²⁺-free and Fe²⁺-bound states (see magenta solid circle and arrows in Fig. 4A), which is indicative of slow exchange kinetics on the NMR chemical shift timescale. The term “slow exchange” in this context refers to Fe²⁺ binding event in which the exchange rate of binding is smaller (slower) than the frequency difference between the NMR peaks (representing the free and bound states). In other words, the exchange rate is slower than the chemical shift timescale and the two states are resolved. By contrast, many other peaks in the NMR titration exhibited fast exchange kinetics in which a single peak titrates from the Fe²⁺-free to Fe²⁺-bound state (see magenta dashed circles in Fig. 4). In this case, the exchange rate is larger (faster) than the frequency difference between the free and bound peaks and the two states are not resolved. Instead, the two states are represented by a single averaged peak whose chemical shift titrates progressively from the Fe²⁺-free to Fe²⁺-bound state (see dashed circles in Fig. 4).

Figure 4. NMR titration of Fe²⁺ binding to GCAP5.

Overlay of ¹H-¹⁵N HSQC spectra of ¹⁵N-labeled GCAP5^WT (A) or GCAP5^C15A/C17A (B) titrated with 0 (red), 0.50 (green), 1.0 (blue), and 3 (black) equivalents of Fe²⁺ ions. Peaks exhibiting slow exchange kinetics during the titration are marked by magenta arrows and solid circle in panel A, and are not present in the titration of GCAP5^C15A/C17A (B). The slow exchange peaks at 7.6/115 and 7.7/116 ppm in the Fe²⁺-free GCAP5^WT spectrum are tentatively assigned to Cys15 and Cys17 because these peaks are abolished in the spectrum of Fe²⁺-free GCAP5^C15A/C17A. The slow exchange peaks at 7.7/115 and 6.8/117 ppm in the Fe²⁺-bound GCAP5^WT spectrum are also assigned to Cys15 and Cys17 because these peaks are abolished in the spectrum of Fe²⁺-bound GCAP5^C15A/C17A. Representative peaks exhibiting fast exchange kinetics are marked by the magenta dashed circles in panels A and B. (C) NMR intensity of the slow exchange peaks from Fe²⁺-free state (7.6/115 and 7.7/116 ppm, red) and Fe²⁺-bound state (7.7/115 and 6.8/117 ppm, black) are plotted as a function of Fe²⁺ concentration. (D) Expanded view of downfield region of HSQC spectrum of GCAP5^C15A/C17A. The downfield peak at 10.60/111.8 ppm, assigned to Gly68 in EF2 in the Fe²⁺-free state (red), is shifted to 10.65/113.8 ppm in the Fe²⁺-bound state (black), suggesting that Fe²⁺ may bind to EF2.

High affinity Fe²⁺ binding site

The Fe²⁺ binding site with slow exchange kinetics (and hence high affinity) is represented by sets of two NMR peaks in Fig. 4A that have different chemical shifts in the Fe²⁺-free versus Fe²⁺-bound states (see solid magenta circle and arrows in Fig. 4A). For example, the NMR peak in the Fe²⁺-free state at 7.6/115 ppm is clearly resolved from its Fe²⁺-bound resonance at 7.7/115 ppm (see solid magenta circle in Fig. 4A). During the NMR titration, the peak at 7.6/115 ppm (Fe²⁺-free state) decreased in intensity as the Fe²⁺ concentration was increased, while the corresponding peak at 7.7/115 ppm (Fe²⁺-bound state) increased in intensity as Fe²⁺ concentration was increased (Fig. 4C). A similar situation occurs for the NMR peak at 7.7/116 ppm in Fe²⁺-free state that titrates into the peak at 6.8/117 ppm in the Fe²⁺-bound state (see magenta arrows in Fig. 4A). The NMR intensity of each slow exchange peak (magenta circle and arrows in Fig. 4A) is plotted as a function of Fe²⁺ concentration (Fig. 4C). The concentration profiles in Fig. 4C indicate that the Fe²⁺-bound peaks (7.7/115 ppm and 6.8/117 ppm) both saturate when the Fe²⁺:GCAP5 ratio is 0.5, consistent with a binding stoichiometry in which one Fe²⁺ binds to two GCAP5 molecules. A similar 1:2 stoichiometry was observed for the high affinity Fe²⁺-binding site detected by ITC (Fig. 2A-B).

The slow exchange peaks in the NMR titration of GCAP5^WT in Fig. 4A (7.6/115 ppm and 7.7/116 ppm in Fe²⁺-free state; vs 7.7/115 ppm and 6.8/117 ppm in Fe²⁺-bound state) are not observed in the NMR titration of GCAP5^C15A/C17A (Fig. 4B). The loss of these slow exchange peaks caused by the C15A/C17A mutation suggests that these peaks (7.6/115 ppm, 7.7/115 ppm, 7.7/116 ppm, and 6.8/117 ppm) are likely assigned to backbone amide resonances of Cys15 and Cys17 in the Fe²⁺-free and Fe²⁺-bound states, respectively. Thus, high affinity Fe²⁺ binding requires both Cys15 and Cys17, consistent with ITC observations above that indicate a lack of nanomolar Fe²⁺-binding to GCAP5^C15A/C17A (Fig. 2D).

Low affinity Fe²⁺ binding sites

A number of peaks in the NMR titration exhibited fast exchange kinetics, consistent with relatively low affinity Fe²⁺ binding (see dashed circles in Figs. 4A-B). The NMR peaks that titrate with fast exchange kinetics are present in NMR spectra of both GCAP5^WT and GCAP5^C15A/C17A, and therefore represent bound Fe²⁺ that must be ligated by non-cysteine residues. These fast exchanging NMR peaks likely represent the multiple Fe²⁺ binding sites detected by ITC with micromolar affinity (Fig. 2). The multiple low affinity Fe²⁺ binding sites probed by NMR and ITC are likely due to non-specific electrostatic Fe²⁺ binding with negatively charged EF-hand loops. To test whether Fe²⁺ might be bound electrostatically to the EF-hand loops, we monitored Fe²⁺ binding to GCAP5 in the presence of saturating Ca²⁺ levels using ITC (Fig. S1). If Fe²⁺ binds electrostatically to EF-hand loops, then the presence of saturating Ca²⁺ should weaken or prevent Fe²⁺ binding to the EF-hand loops. Ca²⁺ binds to two EF-hands of GCAP5 with nanomolar affinity (apparent K_D of 0.37 and 0.61 nM¹⁶) and to one EF-hand with an apparent K_D of 2.91 μM¹⁶ compared to micromolar affinity for Fe²⁺. The ITC Fe²⁺ binding isotherm in the presence of Ca²⁺ showed an overall lower ∆H (∆H = -0.17 kcal/mol) for the exothermic Fe²⁺ binding phase (Fig. S1). Thus, the Ca²⁺-induced lowering of ∆H is consistent with having Ca²⁺ bound to at least two EF-hands in GCAP5 that may impede Fe²⁺ binding to the EF-hands. Indeed, the most downfield NMR peak at 10.60 ppm (assigned to Gly68 in the second EF-hand, EF2) in the Fe²⁺-free/Ca²⁺-free/Mg²⁺-free state undergoes a substantial Fe²⁺-induced spectral shift upon adding a saturating amount of Fe²⁺ (Fig. 4D), consistent with possible Fe²⁺ binding at EF2. In the presence of physiological Mg²⁺ levels (1 mM), Mg²⁺ interferes with Ca²⁺-binding to GCAP5¹⁶ and is known to bind to EF2 in GCAP1^17,29. To test whether Fe²⁺ can bind to EF2 in place of Mg²⁺, an HSQC spectrum of Mg²⁺-bound GCAP5 was recorded in the presence and absence of saturating Fe²⁺ (Fig. S2). In the absence of Fe²⁺, the Mg²⁺-bound GCAP5 exhibited a downfield peak at 10.65 ppm assigned to Gly68 that represents Mg²⁺ binding to EF2 like that observed for GCAP1¹⁷. This downfield peak becomes significantly broadened upon adding a saturating concentration of Fe²⁺, consistent with Fe²⁺ binding to EF2 in place of Mg²⁺ when the Fe²⁺ concentration exceeds that of Mg²⁺.

GCAP5 is a dimer

The molecular size of GCAP5 in solution was determined by size exclusion chromatography (SEC) in Fig. 5. The elution time of GCAP5, calibrated using protein standards (see black peaks in Fig. 5), corresponded to a molar mass of 40 kDa, indicating that GCAP5 is a dimer in solution. The addition of Fe²⁺ and/or Mg²⁺ had no effect on the GCAP5 elution time. Therefore, GCAP5 exists as a dimer in both the Fe²⁺-free and Fe²⁺-bound states. The addition of Ca²⁺ to GCAP5 caused a shorter elution time that corresponded to a molar mass of ∼60 kDa. The Ca²⁺-induced oligomerization of GCAP5 was also detected by NMR, and many NMR peaks of Ca²⁺-bound GCAP5 were broadened beyond detection (Fig. S3). The Ca²⁺-induced broadening of NMR peaks was also observed for unmyristoylated GCAP5, which argues against a possible Ca²⁺-myristoyl switch for GCAP5¹⁶.

Figure 5. GCAP5 is a dimer.

(A) SEC chromatograms of GCAP5^WT in the presence of Mg²⁺ (green) and Fe²⁺ (blue), and GCAP5^C15A/C17A in the presence of Fe²⁺ (red). SEC chromatograms of protein standards (12.4, 29, and 66 kDa) are shown in black. (B) Molar mass calibration curve. The molar mass of GCAP5 was calculated to be 42 kDa (based on the protein standard masses), consistent with GCAP5 forming a dimer in both Fe²⁺-free and Fe²⁺-bound states.

Structural model of a GCAP5 dimer in a [Fe(SCys)₄] complex

The Fe²⁺ binding analysis by NMR and ITC above demonstrates that one Fe²⁺ binds with nanomolar affinity to two molecules of GCAP5. This high affinity Fe²⁺ binding to GCAP5 was abolished by the C15A/C17A mutation, suggesting that side-chain sulfur atoms may ligate the bound Fe²⁺. A structural model of GCAP5 bound to Fe²⁺ was generated in a stepwise fashion. First, a three-dimensional homology model of GCAP5 was generated based on the known crystal structure of chicken GCAP1³⁰. The homology modeled structure of GCAP5 was then docked (using HADDOCK, see Methods) to a single Fe²⁺ ion that was ligated directly to the thiolate atoms of Cys15 and Cys17 (Fig. 6A). In the structure of monomeric GCAP5 bound to Fe²⁺, the indole side-chain of Trp20 is located in close proximity to the bound Fe²⁺, which may explain the Fe²⁺-induced NMR spectral shift of the Trp20 indole resonance at 9.7/127 ppm (Fig. 3). A bound Mg²⁺ at EF2 was included based on the downfield NMR resonance assigned to Gly68 of EF2 in the HSQC spectrum of Mg²⁺-bound GCAP5 (Fig. S2).

Figure 6. Structural model of GCAP5 bound to Fe²⁺.

(A) Homology model structure of Ca²⁺-free/Mg²⁺-bound GCAP5 with bound Fe²⁺ (green) ligated by Cys15 and Cys17 (yellow). Mg²⁺ bound to EF2 (magenta) is supported by the downfield NMR resonance assigned to Gly68 in EF2 of Mg²⁺-bound GCAP5 (Fig. S2), which is similar to that observed for Mg²⁺-bound GCAP1¹⁷. Covalently attached myristic acid is depicted by a space-filling model (brown). (B) Docked structure of GCAP5 dimer in a [Fe(SCys)₄] complex. (C) Schematic model of GCAP5 dimer showing Fe²⁺ bound to residues Cys15 and Cys17. (D) Intermolecular residue contacts at the GCAP5 dimer interface. The H18 imidazole group interacts with side-chain methyl groups of V76, and K84 side-chain amine group is hydrogen bonded to the side-chain carbonyl group of Q88. (E) Close-up view of Fe²⁺ binding site in GCAP5 (left) compared to that of rubredoxin (1FMH, right). Carbon, sulfur, and Fe²⁺ atoms are colored black, yellow, and green.

To generate a dimeric structure of GCAP5, a second homology modeled structure of GCAP5 was docked onto the Fe²⁺-bound GCAP5 structure in Fig. 6A. The modeled structure of the GCAP5 dimer (Fig. 6B) contains a single bound Fe²⁺ that is ligated by the four cysteinyl thiolates of Cys15 and Cys17 from the two docked GCAP5 molecules (Fig. 6B). In essence, a single bound Fe²⁺ bridges two GCAP5 molecules in a [Fe(SCys)₄] complex (Fig. 6C).

Fe²⁺-induced inhibition of guanylate cyclase

Fe²⁺-binding to GCAP5 was unexpected and had not been observed for any other GCAP form before. Therefore, we asked, whether Fe²⁺-bound GCAP5 differs from the Fe²⁺-free form in its regulatory properties. Since we had to adjust the assay conditions for the addition of Fe²⁺, we performed control incubations with bovine GCAP1 (well characterized under different assay conditions). GCAP5 is the weakest activator among all zebrafish GCAP forms showing only very low Ca²⁺-dependent activation (x-fold is 1.5 in the case of nonmyristoylated GCAP5)¹³. In the absence of Fe²⁺, GCAP5 activated recombinant human GC-E roughly by a factor of 20 at high and low Ca²⁺-concentration when compared to the GC-activity in the absence of either GCAP5 or bovine GCAP1 (Figure 7). However, we did not see higher activity in the absence of Ca²⁺ as this is typically seen with other GCAP forms and in our control incubation using bovine GCAP1 (Figure 7B). Further, Fe²⁺-bound GCAP5 strongly decreased GC-E activity independent of the absence or presence of Ca²⁺. The GCAP5 double mutant C15A/C17A activated GC-E to a similar extent as the wildtype did. However, it showed a Ca²⁺-dependent activation yielding an x-fold activation of 1.4 (Figure 7A). In the presence of Fe²⁺, we observed a slight decrease in activity, but the Ca²⁺-dependency was preserved (Figure 7A). In contrast, the presence of Fe²⁺ decreased the activation of GC-E by Ca²⁺-free GCAP1 reaching the same level as with Ca²⁺-bound GCAP1 (Figure 7B). Fe²⁺ did not influence GC-E activity in the absence of any GCAP (Figure 7C).

Figure 7. GC-E enzyme activity regulated by GCAPs in the presence of Fe²⁺.

(A) Activation of recombinant human GC-E by wildtype (WT) GCAP5 or the C15A/C17A double mutant. GCAP concentration was 10 μM, black bars are incubations in the presence of 1.5 μM free [Ca²⁺ ], gray bars are in the presence of Ca²⁺-chelating EGTA (< 10 nM). Fe²⁺ was present at 100 nM. (B) Control incubations of human recombinant GC-E with 5 μM bovine GCAP1 in the absence and presence of 100 nM Fe²⁺. Lyophilized GCAP1 was either resuspended in Na-phosphate buffer (NaPB) or in Mes buffer like GCAP5. (C) Control incubation of GC-E without GCAPs in the presence and absence of Ca²⁺ and/or Fe²⁺. Basal activity showed no influence of either Ca²⁺ or Fe²⁺. Black and gray bars in (B) and (C) are as indicated in (A).

Discussion

We present ITC (Fig. 2), NMR (Figs. 3-4) and cyclase enzymatic assays (Fig. 7) that monitor and structurally characterize Fe²⁺ binding to GCAP5. One Fe²⁺ binds to two molecules of GCAP5 with nanomolar affinity, and additional Fe²⁺ ions bind to GCAP5 with micromolar affinity. The low affinity Fe²⁺ binding to EF2 (Fig. 4D) is likely not physiological, because micromolar Fe²⁺ levels are outside the physiological range. The highest affinity Fe²⁺ binding site is not observed in the GCAP5^C15A/C17A mutant (Figs. 2B and 4B), consistent with the bound Fe²⁺ at this site being ligated by side-chain sulfur atoms in Cys15 and Cys17. A structural model of a GCAP5 dimer forming a [Fe(SCys)₄] complex (Figs. 6B) shows a single bound Fe²⁺ ligated by the four sulfur atoms from Cys15 and Cys17 in a symmetric dimer (Fig. 6E). A similar [Fe(SCys)₄] structural motif was reported previously in the crystal structures of rubredoxin³¹ (Fig. 6E, right panel) and a two-iron superoxide reductase^{32, 33}. Our SEC data on GCAP5 (Fig. 5) indicate that both Fe²⁺-free and Fe²⁺-bound GCAP5 exist as a dimer in solution. A structural model of the Fe²⁺-bound GCAP5 dimer (Fig. 6B) suggests intermolecular protein hydrophobic contacts (between His18 and Val76) as well as an intermolecular hydrogen bond (between Lys84 and Gln88) that may stabilize GCAP5 as a pre-formed protein dimer in the Fe²⁺-free state (Fig. 6D). The intermolecular contact involving Val76 was also implicated in the dimerization of GCAP1, because mutating this Val residue to Glu abolishes dimerization of GCAP1^{17, 19, 34}.

The bound Fe²⁺ in GCAP5 is coordinated by four cysteinyl thiolate atoms (Fig. 6E) that is quite similar to the structure of a bound Zn²⁺ in a Cys4 zinc finger motif³⁵. This structural similarity suggests that Zn²⁺ might bind to GCAP5 in place of Fe²⁺ at the high affinity site. Zn²⁺ is transported into retinal photoreceptor cells and has been suggested to play a role in phototransduction³⁶. Future studies are needed to probe whether Zn²⁺ can replace Fe²⁺ and bind with nanomolar affinity to GCAP5. If so, then it will be interesting to test whether Zn²⁺ binding to GCAP5 (like Fe²⁺ binding) can also regulate zGCs during visual phototransduction.

Functional implications of Fe²⁺-binding to GCAP5

Our results suggest that Fe²⁺ binding to GCAP5 may stabilize and perhaps modulate the quaternary structure of the GCAP5 dimer that could play a role in regulating zebrafish GCs. However, our in vitro functional analysis in this study employed a recombinant mammalian GC, since active recombinant photoreceptor specific GCs from zebrafish are not currently available. Although zebrafish GCAP5 can regulate native zebrafish GCs in retinae membranes⁹, we currently do not have access to zebrafish GC samples in live animals. The addition of Fe²⁺ caused complex effects on the regulation of mammalian GC and showed a strong inhibitory effect mediated by both GCAP5 and GCAP1 (Figs. 7A and 7B). The Fe²⁺-induced GC inhibition meditated by GCAP1 is surprising, and might indicate possible Fe²⁺ binding to the EF-hands in GCAP1 that may mimick Ca²⁺ binding. The Fe²⁺-induced inhibition of GC was somewhat weakened by the GCAP5 double mutant (C15A/C17A), consistent with functional Fe²⁺ binding by Cys15 and Cys17 (Fig. 6). The GCAP5 double mutant also showed a more typical Ca²⁺-dependent activation of the target GC compared to the Ca²⁺-independent GC activation by GCAP5^WT (Fig. 7A). This suggests that Fe²⁺ binding to GCAP5 may stabilize the inhibitory conformational state of GCAP5 even in the absence of Ca²⁺, and therefore Fe²⁺-bound GCAP5 may serve to constitutively disable GC activation. The lack of Ca²⁺-sensitive GC activation by GCAP5^WT might also be explained by differences in our test incubations when compared to previous results^{9, 13}. In the current study, we used myristoylated GCAP5 instead of nonmyristoylated protein that was used previously. The presence of the myristoyl group in GCAP5 appears to prevent the typical Ca²⁺-sensitive activation of a GC, whereas the double mutation (C15A/C17A) restored the Ca²⁺-dependent GC activation. These observations suggest a complex interdependence of the Fe²⁺ binding site (at Cys15 and Cys17), the myristoyl group, and the Ca²⁺-free/Mg²⁺-bound activator state of GCAP5. The source of GC in our test system was not critical for the outcome of the assay. We compared heterologously expressed human GC-E in HEK 293 cells with mammalian GC from native bovine rod outer segment membranes and did not observe a principal difference in the results (data not shown).

Does Fe²⁺-binding to GCAP5 have any physiological meaning? At this point, we can only speculate on possible physiological implications, but several observations suggest that GCAP5 might not be the main activator of GCs in the zebrafish retina. Indeed, all other zebrafish GCAPs are at least 3-fold more active than GCAP5¹³. Furthermore, the expression and activation profiles of the other GCAPs promote a dynamic Ca²⁺-sensitive activation of zebrafish GCs in the absence of GCAP5, suggesting a lack of GCAP5 participation perhaps due to compensatory actions^9-13. Finally, the downregulation of GCAP5 by Fe²⁺ further diminishes the already weak impact of GCAP5 on GC activation. Instead, we suggest that GCAP5 might bind to a different and as yet unknown target protein. Preliminary studies suggest that GCAP5 is expressed in the inner segment (K.W.K and A.S., unpublished results). Therefore, the binding of Fe²⁺ to GCAP5 might serve a regulatory role for a redox-sensitive process in the inner segment. Future experiments are needed to look for a GCAP5 binding partner protein expressed in the inner segment.

The Fe²⁺-induced inhibition of GC mediated by GCAP5 (Fig. 7A) may protect photoreceptor cells from apoptosis and possibly prevent retinal degeneration. Mutations in GCs and/or GCAPs that cause constitutive GC activation are genetically linked to retinal degeneration^{29, 37-41}. Also, the retinal degeneration 3 (RD3) protein (that binds to GC in place of GCAPs and prevents GC activation) is important for turning off basal cyclase activity during the trafficking of GC to the outer segment⁴². The basal cyclase activity of GC that occurs in the absence of RD3 is thought to trigger apoptosis of photoreceptor cells that could lead to retinal degeneration⁴³. Therefore, inhibition of basal cyclase activity by RD3 is believed to be important for preventing retinal degeneration. We suggest that Fe²⁺-induced inhibition of GC by GCAP5 might be a different mechanism for turning off GC activity in the inner segment, which could serve to protect zebrafish photoreceptor cells from retinal degeneration.

Abbreviations

EGTA

ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid

HSQC

heteronuclear single quantum coherence

ITC

isothermal titration calorimetry

NMR

nuclear magnetic resonance

GC

retinal guanylate cyclase

SEC

size exclusion chromatography