Recoverin Binds Exclusively to an Amphipathic Peptide at the N Terminus of Rhodopsin Kinase, Inhibiting Rhodopsin Phosphorylation without Affecting Catalytic Activity of the Kinase*

Recoverin is a calcium-dependent inhibitor of rhodopsin kinase. It prevents premature phosphorylation of rhodopsin until the opening of cGMP-gated ion channels causes a decrease in intracellular calcium levels, signaling completion of the light response. This calcium depletion causes release of recoverin from rhodopsin kinase, freeing the kinase to phosphorylate rhodopsin and to terminate the light response. Previous studies have shown that recoverin is able to bind to a region at the N terminus of rhodopsin kinase. In this study we map this interaction interface, showing that residues 1-15 of the kinase form the interaction site for recoverin binding. Mutation of hydrophobic residues in this region have the greatest effect on the interaction. The periodic nature of these residues suggests that they lie along one face of an amphipathic helix. We show that this region is essential for recoverin binding, as a catalytically active kinase lacking these residues is unable to bind recoverin. In addition, we show that neither the N-terminal deletion nor the presence of recoverin inhibits the overall catalytic activity of the kinase, as measured by light-independent autophosphorylation. Finally, we observe that a kinase mutant lacking the N-terminal recoverin binding site is unable to phosphorylate light-activated rhodopsin. Taken together, these data support a model in which recoverin prevents rhodopsin phosphorylation by sterically blocking a region of kinase essential for its interaction with rhodopsin, thereby preventing recognition of rhodopsin as a kinase substrate.

Recoverin is a calcium-dependent inhibitor of rhodopsin kinase. It prevents premature phosphorylation of rhodopsin until the opening of cGMP-gated ion channels causes a decrease in intracellular calcium levels, signaling completion of the light response. This calcium depletion causes release of recoverin from rhodopsin kinase, freeing the kinase to phosphorylate rhodopsin and to terminate the light response. Previous studies have shown that recoverin is able to bind to a region at the N terminus of rhodopsin kinase. In this study we map this interaction interface, showing that residues 1-15 of the kinase form the interaction site for recoverin binding. Mutation of hydrophobic residues in this region have the greatest effect on the interaction. The periodic nature of these residues suggests that they lie along one face of an amphipathic helix. We show that this region is essential for recoverin binding, as a catalytically active kinase lacking these residues is unable to bind recoverin. In addition, we show that neither the N-terminal deletion nor the presence of recoverin inhibits the overall catalytic activity of the kinase, as measured by light-independent autophosphorylation. Finally, we observe that a kinase mutant lacking the N-terminal recoverin binding site is unable to phosphorylate light-activated rhodopsin. Taken together, these data support a model in which recoverin prevents rhodopsin phosphorylation by sterically blocking a region of kinase essential for its interaction with rhodopsin, thereby preventing recognition of rhodopsin as a kinase substrate.
Light-dependent activation of rhodopsin in rod cells initiates a G protein-mediated signaling cascade that hyperpolarises the rod and inhibits glutamate release at its synaptic terminal. Inactivation of rhodopsin is essential for termination of this signaling event and for resetting rod cells to receive future stimuli. The inactivation process depends on phosphorylation of the C terminus of light-activated rhodopsin through the action of rhodopsin kinase (1,2). The subsequent binding of arrestin to this phosphorylated peptide, and to cytoplasmic loops of the receptor, blocks access of the G protein, transducin, and prevents further signaling (3).
Rhodopsin kinase is one of seven G protein-coupled receptor kinases (GRKs) 2 (4). These contain a common domain structure for the N-terminal 520 residues and a divergent C-terminal region (5). Rhodopsin kinase (also known as GRK1) shows many similarities to the more ubiquitous ␤-adrenergic receptor kinase (GRK2), sharing two major domains, the N-terminal regulator of G protein signaling (RGS) homology domain and the classical bilobal catalytic domain. The two kinases have 34% sequence identity, with 44% identity in the catalytic domain (6). The crystal structure of GRK2 in complex with G protein subunits, G␤␥, shows the structure of these major domains (7). The N-terminal RGS homology domain comprises nine ␣-helices, spanning residues 30 -185, and two additional ␣-helices from the C terminus of the protein (residues 513-547). The N terminus (residues 1-29) was not observed in the crystal structure. The kinase domain (residues 186 -512) shows a classic bilobal structure and adopts a slightly "open" conformation characteristic of an inactive protein kinase with a disordered nucleotide gate (7). Rhodopsin kinase has two major substrates in the rod outer segment: it interacts with and phosphorylates light-activated rhodopsin (8) and undergoes autophosphorylation (9,10). A series of seven Ser and Thr residues on the C-terminal peptide of light-activated rhodopsin form the substrate (11). However, the activity of rhodopsin kinase for isolated peptide substrates derived from this region is around 1000-fold lower than for intact, light-activated receptor (11). In addition, no clear consensus sequence immediately surrounds the phosphorylated residues to provide such high selectivity (11). This suggests that further contacts between kinase and rhodopsin are involved in selection of activated rhodopsin as a substrate. Residues in all three cytoplasmic loops of rhodopsin have been implicated in this interaction (12)(13)(14) as have residues at the N terminus (15) and C terminus of kinase (16).
Recoverin is a small myristoylated calcium binding protein containing four EF hands (17,18). Structures in the presence (19) and absence (20) of calcium show a considerable calcium induced conformational change (21,22). While in the calciumfree form, the myristoyl group is held within a hydrophobic cleft (19), addition of calcium leads to its release (21,22) and to membrane binding (23,24). Addition of calcium also leads to tight binding of recoverin to rhodopsin kinase (25,26) with residues 1-25 of kinase showing capacity to interact (27). This interaction prevents phosphorylation of light-activated rhodopsin (25,28). Therefore, the presence of recoverin increases the lifetime of the light response (29,30), while recoverin deficiency shortens the response (31,32). Kinase is released from recoverin when cGMP-mediated gating of ion channels causes a decrease in intracellular calcium levels (33). This delays phosphorylation of rhodopsin and inactivation of the signaling process until the light response is complete.
In this study, we investigate the mechanism of inhibition of rhodopsin kinase by recoverin. Previous work has shown that recoverin is able to bind to a peptide corresponding to residues 1-25 of rhodopsin kinase (27). Here we show, using a catalytically active truncated kinase, that this N-terminal peptide is necessary for the binding of recoverin. We investigate the role of each residue from this region on recoverin binding and present data suggesting that the N-terminal 15 residues form an amphipathic helix that interacts with recoverin through the nonpolar face. Finally we show that this peptide is not essential for catalytic function of the kinase domain, as measured by autophosphorylation, but is needed for phosphorylation of light-activated rhodopsin. The same effect is seen in the presence of recoverin, with recoverin bound kinase able to undergo autophosphorylation, but not to phosphorylate light-activated rhodopsin. This supports a mechanism in which the N terminus of rhodopsin kinase is essential for recognition of rhodopsin as a substrate and is sterically blocked by recoverin, thereby preventing rhodopsin phosphorylation without altering the catalytic activity of the kinase.

EXPERIMENTAL PROCEDURES
Expression and Purification of Recoverin-The recoverin gene (residues 1-202) was amplified by PCR from a bovine retinal cDNA library and inserted into BamHI-XhoI cleaved pGex4T1 (GE Healthcare), generating an N-terminal glutathione S-transferase (GST) fusion of recoverin with an intervening thrombin cleavage site. Protein was expressed by transformation into BL21 RIL Codon-Plus cells (Stratagene), growth to an absorbance at 600 nM of 1.0, induction with 1 mM isopropyl ␤-D-thiogalactopyranoside, and growth overnight at 25°C. Recoverin was purified by breaking cells in PBS (20 mM sodium phosphate, pH 7.2, 150 mM NaCl) with 1% Triton X-100 using a French press and pelleting at 125,000 ϫ g. Supernatant was incubated with glutathione-Sepharose (Amersham Biosciences) for 45 min, washed with 500 ml of PBS with 1% Triton X-100, 1 liter of PBS with 1 M NaCl, and 200 ml of thrombin cleavage buffer (20 mM Tris, pH 8.3, 150 mM NaCl, 2.5 mM CaCl 2 ). 75 units of thrombin (Sigma) were added and incubated overnight at 4°C. Cleavage efficiency was in excess of 90%. Beads were removed by centrifugation, and the protein was concentrated using a Centricon Plus-20 to a volume of less than 1 ml. Gel filtration on a Superdex 200 gel filtration column, with a buffer containing 50 mM NaCl, 20 mM Tris, pH 8.0, resulted in a single peak with a molecular weight consistent with a monomer of recoverin. Protein concentration was determined using absorbance at 280 nM and the BCA assay (Pierce).
Rhodopsin Kinase Expression, Purification, and Mutagenesis-Residues 1-562 of rhodopsin kinase were amplified by PCR from a bovine retinal cDNA library and inserted into the BamHI site of pBacPAK8 (Clontech). A baculovirus was generated through cotransfection of the plasmid with linearized BaculoGold DNA (Pharmingen) into Sf9 cells (Invitrogen). The virus was plaque-purified and amplified through standard methods (43) and expression tested using a rhodopsin kinase 1a/b antibody (Abcam). For large scale expression, Sf9 cells were grown to 2 ϫ 10 6 cells/ml in TNM-FH medium (Sigma) with 10% fetal bovine serum. Virus was added to a multiplicity of infection of 2.0. Cells were diluted to 1 ϫ 10 6 cells/ml with fresh medium and incubated for 72 h before harvesting.
Rhodopsin kinase was purified using recoverin affinity chromatography and gel filtration. Infected Sf9 were broken using a Dounce homogenizer into purification buffer (20 mM MOPS, 60 mM KCl, 30 mM NaCl, 2 mM MgCl 2 , 1 mM CaCl 2 , pH 7.2) and clarified by centrifugation at 125,000 ϫ g for 30 min. Recoverin resin was prepared by incubating GST-recoverin with glutathione-Sepharose as described above. Recoverin resin was washed with purification buffer and incubated with the supernatant from infected Sf9 cells for 60 min at 4°C. Beads were washed with purification buffer containing 0.5 M NaCl. Rhodopsin kinase was eluted from the resin by addition of 5 mM EGTA and concentrated using a Centricon Plus-20 to a volume less than 1 ml. Gel filtration on a Superdex 200 gel filtration column, with 1 mM CaCl 2 , 50 mM NaCl, 20 mM Tris, pH 8.0, resulted in a single peak with a molecular weight consistent with a monomer of rhodopsin kinase. Protein concentration was determined using absorbance at 280 nM and the BCA assay (Pierce).
Expression and Purification of Kinase Mutant 32-562-A truncated rhodopsin kinase gene, consisting of residues 32-562 linked to a C-terminal His 6 tag, was generated by PCR and inserted into the BamHI site of pBacPAK8 (Clontech). Baculovirus production and expression were carried out as described above. Protein was purified using Ni 2ϩ affinity, heparin sulfate affinity chromatography, and gel filtration. Infected Sf9 cells were broken using a Dounce homogenizer into PBS with 10 mM imidazole. Supernatant was clarified by centrifugation at 125,000 ϫ g for 30 min and incubated with Ni 2ϩ -nitrilotriacetic acid resin for 60 min. Resin was washed with PBS containing 0.5 M NaCl and 10 mM imidazole. Protein was eluted with PBS containing 200 mM imidazole and desalted using a PD10 column (Amersham Biosciences) into 100 mM NaCl, 1 mM CaCl 2 , 20 mM Tris, pH 8.0. The sample was loaded onto a heparin sulfate column (Apbiotech) (44,45), washed, and eluted with a 0 -0.5 M KCl gradient in 20 mM Tris, pH 7.4, and concentrated with a Centricon Plus-20 to a volume of less than 1 ml. Gel filtration on a Superdex 200 gel filtration column, with 1 mM CaCl 2 , 50 mM NaCl, 20 mM Tris, pH 8.0, resulted in a single peak with a molecular weight consistent with a monomer. Protein concentration was determined using absorbance at 280 nM and the BCA assay (Pierce).
Expression and Purification of GST Fusion Proteins-To make GST fusion proteins with an N-terminal GST, genes were inserted into the BamHI-XhoI site of pGex4T-1 (GE Healthcare). To make fusion proteins with rhodopsin kinase peptides attached to the N terminus of GST, residues 1-210 of GST were amplified by PCR from pGex4T-1 and inserted into the NcoI-BamHI site of pET15b (Novagen), generating a construct with an NcoI site at the 5Ј-end of GST. Inserts encoding rhodopsin kinase peptides were made by annealing synthetic oligonucleotides, generating double-stranded products with single-stranded overhangs to anneal into NcoI cleaved vector. Oligonucleotides were combined, treated with polynucleotide kinase (New England Biolabs), heated to 100°C, and cooled at 1°C per min. The product was gel-purified (Qiagen) and ligated into NcoI cut vector to generate an N-terminal fusion to GST.
Proteins were expressed by transformation of the plasmid into BL21 Codon-Plus cells (Stratagene), growth to an absorbance at 600 nM of 1.0, induction with 1 mM isopropyl ␤-D-thiogalactopyranoside, and growth for 4 h at 30°C. Cells were broken in PBS with 1% Triton X-100 using a French press. Supernatant was clarified by centrifugation at 125,000 ϫ g and incubated with glutathione-Sepharose for 45 min. Resin was washed with PBS with 1% Triton X-100 and PBS with 0.5 M NaCl. Protein was eluted with 50 mM reduced glutathione in 50 mM Tris, pH 8.0, and concentrated using a Centricon Plus-20 to a volume of less than 1 ml. Gel filtration on a Superdex 200 gel filtration column, with buffer containing 1 mM CaCl 2 , 50 mM NaCl, 20 mM Tris, pH 8.0, resulted in a single peak with molecular weights consistent with monomers. Protein concentration was determined using absorbance at 280 nM and the BCA assay (Pierce).
Phosphorylation Assays-Rod outer segment membranes were produced as in (46)  Pulldown Experiments-Pulldown experiments were performed using purified proteins and glutathione resin. 25 l of glutathione-Sepharose (Amersham Biosciences), with capacity of 5 mg of GST per ml, was incubated with 20 M GST fusion protein and 20 M interaction partner. Interaction buffer (20 mM MOPS, 60 mM KCl, 30 mM NaCl, 2 mM MgCl 2 , 1 mM CaCl 2 , pH 7.2) was added to a total volume of 100 l. After incubation for 60 min, resin was pelleted by centrifugation at 13,000 rpm for 2 min and washed twice by resuspension and pelleting with 1-ml interaction buffer with 0.5 M NaCl. Pelleted protein was eluted by incubation with 50 l of interaction buffer containing 10 mM EGTA and was analyzed by SDS-PAGE and silver staining (Pierce).
Biacore Analysis-Measurements were performed on a Biacore 2000 instrument with a constant flow rate of 5 l/min. Channel 1 was blocked as a control channel with 50 mM L-cysteine. Purified recoverin was coupled to channel 2 as described in Ref. 34 using the thiol coupling protocol recommended by the manufacturer (Biacore). Around 1000 response units of recoverin was immobilized using this procedure. Channels 1 and 2 were equilibrated with HBS-MCP buffer (20 mM HEPES, pH 7.4, 150 mM NaCl, 0.1 mM CaCl 2 , 1 mM MgCl 2 , 0.005% Tween 20). Purified rhodopsin kinase and kinase truncations were concentrated, and buffer was exchanged using PD10 columns (Amersham Biosciences) for HBS-MCP buffer. A dilution series (62.5 nM, 125 nM, 250 nM, 500 nM, 1 M, 2 M, 3 M, 5 M, and 10 M) was injected over channels 1 and 2 with the subtraction 2-1 giving the specific binding to recoverin. Data were analyzed using BIAevaluation software to obtain maximum response values. These were plotted against the concentration of kinase using Prism and fitted to a single site binding model, allowing determination of the K D .
To analyze binding to GST fusion peptides, anti-GST antibody was immobilized to all channels using the amide coupling protocol as recommended by the manufacturer (Biacore) and washed with HBS-MCP buffer. Antibody was loaded with GST and GST fusion proteins by injection of 120 l of a 5 g/ml solution of GST fusion in HBS-MCP buffer. Channel 1 was loaded with GST to act as a control, while channels 2-4 were loaded with fusion proteins. The subtractions of responses, 2-1, 3-1, and 4-1 gave specific binding to channels 2, 3, and 4, respectively. Solutions of increasing concentrations of recoverin ( were injected through the channels and difference response recorded. Data were analyzed as above, plotting the maximum response against the concentration of recoverin and fitting to a one-site binding model. For mutants with decreased affinity, extrapolation of data was required to obtain K D measurements. This is reflected in the standard deviation of the measurements reported in Table 1.

Recoverin Inhibits Light-dependent Rhodopsin Phosphorylation but Not Light-independent Rhodopsin Kinase
Autophosphorylation-To investigate the role of recoverin in the inhibition of rhodopsin kinase we incubated purified kinase with membranes from bovine rod outer segments and [␥-32 P]ATP in the presence of 1 mM CaCl 2 (Fig. 1). In the dark a band running at around 65 kDa is observed due to kinase autophosphorylation. This band also occurs when purified rhodopsin kinase is incubated with [␥-32 P]ATP (data not shown). In the presence of light, kinase autophosphorylation is unaltered and an additional band at 35 kDa appears, together with higher molecular mass species. Western blotting with an antibody specific to rhodopsin confirmed that both the 35-kDa species and the higher molecular mass specifies correspond to rhodopsin (data not shown), showing that rhodopsin phosphorylation is light-dependent.
These experiments were repeated in the presence of 100 M purified recoverin (Fig. 1). The addition of recoverin had no effect on light-independent kinase autophosphorylation but prevented light-dependent phosphorylation of rhodopsin. This effect was only observed in the presence of calcium, confirming previous observations that calcium is required for the inhibition of light-dependent rhodopsin phosphorylation (25,26). This suggests that recoverin does not inhibit the catalytic function of rhodopsin kinase but instead prevents recognition of rhodopsin as a substrate for the kinase, presumably by preventing the interaction between kinase and rhodopsin.
Recoverin Binds to the N-terminal Peptide of Rhodopsin Kinase-To characterize the recoverin binding site of rhodopsin kinase we divided the kinase into functional regions. Guided by alignment between rhodopsin kinase and GRK2, four regions were selected: residues 1-32 (disordered in the GRK2 structure), residues 32-184 (the RGS domain), residues 185-522 (the kinase domain), and residues 522-562 (the C terminus). These were expressed in bacteria as fusions to the C terminus of GST and were tested for ability to pellet recoverin. Of these constructs, only GST fusions containing residues 1-32 could cause recoverin to pellet on glutathione resin ( Fig. 2A).
The exclusive binding of recoverin to these residues of kinase was further investigated by peptide competition experiments. Here, the ability to pellet full-length rhodopsin kinase by a GST-recoverin fusion was studied. In the presence of calcium, GST-recoverin pelleted kinase on glutathione beads (Fig. 2B). The addition of a synthetic peptide corresponding to residues 1-32 of rhodopsin kinase prevented this interaction, showing that residues 1-32 of the kinase are sufficient to compete with recoverin for binding to full-length rhodopsin kinase.
In addition, rhodopsin kinase was expressed in the absence of the N-terminal 31 residues (32-562). This protein no longer pellets on glutathione resin in the presence of GST-recoverin (Fig. 2C), showing that the N-terminal 31 residues of rhodopsin kinase are essential components of the recoverin binding site.
To confirm this observation under equilibrium binding conditions, Biacore measurements were made. Recoverin was coupled via its unique Cys residue (Cys 39 ) (34) to a CM5 Biacore chip, while a control channel was reacted with cysteine. Rhodopsin kinase was flowed over the chip and showed clear specific binding to the recoverin immobilized surface. Data showing the response of the Biacore chip to different kinase concentrations fitted to a one site-binding model with a K D of 37 Ϯ 9 M (Fig. 3). In contrast, 32-562 kinase showed no specific binding to recoverin under these conditions (Fig. 3). Therefore the N-terminal peptide of rhodopsin kinase is essential for recoverin binding at physiologically relevant kinase concentration.

Residues 1-15 of Rhodopsin Kinase Form a Putative Amphipathic Helix That Interacts with Recoverin
Primarily through the Hydrophobic Surface-To further delineate the recoverin binding site, GST fusions were generated in which peptides from the N terminus of rhodopsin kinase were fused to the N terminus of GST. To determine the extent of the binding site, a series of truncations were generated and used in recoverin pelleting experiments. First, the C terminus of the peptide was truncated, showing that, while residues 1-15 are sufficient to pellet recoverin, residues 1-10 are not sufficient (Fig. 4A). Truncation experiments with N-terminal deletions showed that removal of residues 1-5 also prevented binding (Fig. 4B).
Next, GST fusions were made with each residue in 1-15 mutated to Ala. Each of these mutations caused a decrease in the quantity of recoverin in the pellet (Fig. 4C). The extent of this decrease in binding followed a periodic pattern with mutations F3A, L6A, V9A, V10A, S13A, and F15A causing a near complete loss of recoverin pelleting, while other mutations decreased the interaction by a lesser degree. If residues 1-15 of rhodopsin kinase are plotted as a helical wheel representation (Fig. 4D), the residues that cause complete loss of binding (labeled red), with the exception of Phe 15 , lie on one, predominantly hydrophobic face of the helix.
To estimate the affinity of these mutants for recoverin, Biacore measurements were made. Peptide-GST fusions were immobilized to the chip surface through an anti-GST antibody and recombinant GST was attached to the control channel. Increasing concentrations of recoverin were flowed over a cell with immobilized wild type 1-15 GST, and the response could be fitted to a one-site binding model, with a K D of 41 Ϯ 2 M   JULY 14, 2006 • VOLUME 281 • NUMBER 28 (Fig. 4E). This compares well with the K D of 37 Ϯ 9 M for binding of full-length kinase to recoverin, supporting a model in which residues 1-15 of kinase are sufficient for the high affinity interaction between recoverin and kinase.

The Mechanism of Control of Rhodopsin Kinase by Recoverin
The same experiment was performed for GST fusions containing mutations of the 1-15 peptide from kinase. All mutations tested reduced binding to recoverin (Fig. 4E and Table 1). Data were again fitted to a one-site binding model, with extrapolation to higher concentrations necessary to allow K D estimation for those mutants with dramatically decreased affinities.
The smallest effect was the ϳ5-fold increase in K D of the A11R mutation, while the largest change was the ϳ250-fold increase in K D caused by the L6A mutation. Most mutations increased the K D by 20 -65-fold. Once again, a periodic pattern was observed, with large effects from F3A, L6A, V9A, V10A, S13A and F15A. In this case, D2A and T8A mutations also had   shows no response. B, a plot of the maximum response for a series of rhodopsin kinase injections. While data for full-length rhodopsin kinase fit a curve with a K D of 37 Ϯ 9 M, no binding is observed for truncated rhodopsin kinase. a significant effect, suggesting that these hydrophilic residues make a significant contribution to the binding interface.
The N-terminal Peptide of Rhodopsin Kinase Is Essential for Phosphorylation of Rhodopsin-To study the role of the N-terminal peptide of rhodopsin kinase in phosphorylation of rhodopsin, we investigated incorporation of radiolabeled phosphate into rhodopsin by full-length kinase and by the N-terminal deletion mutant, 32-562. Kinases were incubated with membranes from rod outer segment in the presence of [␥-32 P]ATP in the presence or absence of light. Both full-length and 32-562 kinases showed light-independent autophosphorylation, showing the presence of a functional catalytic domain (Fig. 5). However, while full-length kinase phosphorylated rhodopsin in the light, 32-562 showed no rhodopsin phosphorylation (Fig. 5). Therefore the N-terminal peptide is needed for phosphorylation of rhodopsin but is not necessary for catalytic activity of the kinase domain. Indeed, a previous study showed that the E7A point mutation of rhodopsin kinase inhibits phosphorylation of rhodopsin but not phosphorylation of peptide substrates (15).

DISCUSSION
Previous studies have shown that residues 1-25 of rhodopsin kinase form a binding site for recoverin (27). We have characterized this recoverin binding site and investigated the role of these residues in recoverin binding and rhodopsin phosphorylation. We have shown the high affinity recoverin-kinase interaction to be exclusively mediated through this peptide, as a catalytically active kinase lacking these residues is unable to bind to recoverin. In addition, a peptide corresponding to residues 1-15 of the kinase binds recoverin with an affinity comparable with that for full-length kinase. Finally, we show that this truncated kinase is unable to phosphorylation rhodopsin. These data support a steric model for the function of recoverin in which recoverin prevents rhodopsin phosphorylation by binding to a region of rhodopsin kinase essential for the recognition of rhodopsin as a kinase substrate.
Mapping the Interface between Recoverin and Rhodopsin Kinase-We have used pulldown assays and surface plasmon resonance measurements to investigate the binding of recoverin to GST fusion proteins containing the N terminus of rhodopsin kinase. Pulldown experiments show that the first 15 residues of rhodopsin kinase are sufficient for this interaction, and five-residue deletions from either end of this peptide prevent binding. Through mutagenesis studies of this kinase peptide, we show that the interface between the kinase and recoverin is extensive, with all mutations causing a reduction in binding. Moreover, the degree to which binding is inhibited shows a periodic pattern, with mutation of hydrophobic side chains having the greatest effect. This suggests the N-terminal peptide of the kinase to form an amphipathic ␣-helix with a hydrophobic face generated by residues F3A, L6A, V9A, and V10A. Mutation of residue Leu 6 , in the center of this face, leads to the greatest loss of recoverin affinity, suggesting that it lies at the core of the interaction interface.
A similar binding interaction has been proposed for other recoverin homologues. These proteins, which are members of the neuronal calcium sensor family, include the Kv channel interacting proteins (KChIPs) and frequentin (35). KChIP proteins interact directly with the N terminus of Shal-type voltagegated potassium channels to modulate their surface expression and functional properties (36,37). In a crystal structure of a fusion between the C terminus of KChIP1 and the N-terminal 30 residues of Kv4.2, the N-terminal end of Kv4.2 forms an ␣-helix, which lies in a deep hydrophobic pocket of KChIP1, interacting primarily through non-polar and aromatic residues that lie along one face of the helix (38). In addition, frequentin is an neuronal calcium sensor protein that binds to the phosphatidylinositol 4-kinase isoform, Pik1. In this case, hydrophobic residues within a 28-residue peptide from Pik1 (residues 145-172) form an amphipathic helical binding site for frequentin (39,40).
The data presented in this study suggest that the interaction between recoverin and rhodopsin kinase also occurs through the binding of the hydrophobic face of an amphipathic ␣-helix at the N terminus of rhodopsin kinase to a hydrophobic surface on recoverin. Indeed, a hydrophobic groove is present in the calcium-bound form of recoverin in an equivalent position to the Kv4.2 binding groove of KChIP (20) but is not observed in the inactive calcium-free state of recoverin (22). Mutations made around the surface of this groove in the frog recoverin homologue, S-modulin, decrease rhodopsin kinase binding