Rhodopsin kinase inhibition by recoverin. Function of recoverin myristoylation.

Recoverin is a Ca-binding protein that may play a role in vertebrate photoreceptor light adaptation by imparting Ca sensitivity to rhodopsin kinase. It is heterogeneously acylated (mostly myristoylated) at its amino-terminal glycine. Recent studies have shown that recoverin myristoylation is necessary for its Ca-dependent membrane association and cooperative Ca binding. We have addressed several issues concerning the role of recoverin myristoylation with respect to inhibition of rhodopsin kinase. We find that 1) myristoylation of recoverin is not necessary for inhibition of rhodopsin kinase, 2) myristoylation of recoverin induces a cooperative Ca-dependence for rhodopsin kinase inhibition, and 3) each Ca-binding site on the nonmyristoylated recoverin partially inhibits rhodopsin kinase. The available data suggest that the functions of recoverin myristoylation in the living rod are to induce a sharp Ca dependence of rhodopsin kinase inhibition and to bring this dependence into the rod's physiological Ca concentration range.

Recoverin is a Ca 2؉ -binding protein that may play a role in vertebrate photoreceptor light adaptation by imparting Ca 2؉ sensitivity to rhodopsin kinase. It is heterogeneously acylated (mostly myristoylated) at its aminoterminal glycine. Recent studies have shown that recoverin myristoylation is necessary for its Ca 2؉ -dependent membrane association and cooperative Ca 2؉ binding. We have addressed several issues concerning the role of recoverin myristoylation with respect to inhibition of rhodopsin kinase. We find that 1) myristoylation of recoverin is not necessary for inhibition of rhodopsin kinase, 2) myristoylation of recoverin induces a cooperative Ca 2؉ -dependence for rhodopsin kinase inhibition, and 3) each Ca 2؉ -binding site on the nonmyristoylated recoverin partially inhibits rhodopsin kinase. The available data suggest that the functions of recoverin myristoylation in the living rod are to induce a sharp Ca 2؉ dependence of rhodopsin kinase inhibition and to bring this dependence into the rod's physiological Ca 2؉ concentration range.
In the vertebrate photoreceptor recoverin may provide a Ca 2ϩ -dependent feedback system involved in light adaptation by binding Ca 2ϩ in the dark, when Ca 2ϩ levels are high, and releasing it when Ca 2ϩ levels drop upon illumination (1)(2)(3)(4)(5). Ca 2ϩ -recoverin inhibits RK, 1 and release of this inhibition would accelerate the inactivation of Rho* when Ca 2ϩ levels drop. In a previous study (4) we have demonstrated that Ca 2ϩ bound recoverin acts through a direct interaction with RK to decrease its catalytic activity (see also Refs. 2 and 5).
Recoverin belongs to a family of proteins post-translationally modified by the covalent addition of myristate to their aminoterminal glycine residues (3). Myristoylation of recoverin is necessary for its Ca 2ϩ -dependent association with photoreceptor membranes (6 -8). This membrane binding serves to bring inhibition of RK by recoverin into the physiological Ca 2ϩ range under in vivo conditions (4,6). Recently it was demonstrated that the myristoylation of recoverin induces cooperative Ca 2ϩ binding to recoverin (10). A previous study has asserted that only one Ca 2ϩ -binding site on recoverin is involved in the inhibition of Rho* phosphorylation, and that the other is involved in membrane binding (8). We have studied recoverin myristoylation with regard to inhibition of RK in order to further define its function and have addressed the following issues. 1) Is myristoylation of recoverin necessary for RK inhibition? 2) Is myristoylation of recoverin necessary for the inhibition of RK to be cooperative with respect to Ca 2ϩ concentration? 3) Do both Ca 2ϩ -binding sites on recoverin act to promote the inhibition of RK? ROS Preparation and RK Extraction-Bovine ROS were purified under infrared illumination as described elsewhere (11). The same method of bovine ROS isolation was used for RK extracts, but all sucrose solutions were prepared in 10 mM Tris, 5 mM MgCl 2 , pH 7.5, and the procedure was performed in room light. RK was extracted as previously described (12).
Expression and Purification of Recombinant Bovine Recoverin-Recombinant bovine myristoylated and nonmyristoylated recoverins were produced using a bacterial expression system (7) that was a generous gift from Dr. J. Hurley. For the myristoylated form of recoverin, the extent of myristoylation was determined by electrospray ionization mass spectrometry and found to be greater than 95%. Both forms of recoverin were purified by a published procedure (13) and stored at Ϫ70°C. The concentrations of each form of recoverin were determined by absorbance at 280 nm using a molar extinction coefficient of 36,400 (4).
Calcium Buffering-A set of 4 ϫ stock solutions with different CaCl 2 concentrations and a fixed dibromro-BAPTA concentration in standard buffer were prepared as described in Klenchin et RK Activity Assay-RK activity was measured as described previously (4). Briefly, urea-treated ROS membranes were used as substrate (10 M rhodopsin, final concentration). Urea-treated ROS membranes, a ROS extract containing RK, a CaCl 2 /dibromo-BAPTA stock solution, and recoverin if necessary were mixed in 15 l, final volume, in the dark, the suspension was illuminated bleaching all Rho, and 5 l of [␥-32 P]ATP (0.2-0.4 mM) were added. After 1-2 min the reaction was stopped by addition of 100 mM EDTA, 100 mM KF, pH 7.5. 32 P i incorporation was measured by filtration of samples through nitrocellulose filters.

RESULTS AND DISCUSSION
Only the Ca 2ϩ bound form of recoverin is able to inhibit RK. The potency of inhibition of RK by myristoylated and nonmyristoylated recoverin can thus be compared when all Ca 2ϩ -binding sites are occupied. The dependence of the inhibition of RK on the concentration of myristoylated and nonmyristoylated recoverin at saturating Ca 2ϩ levels is shown in Fig. 1 (4), and at sufficiently high concentration are able to fully suppress RK activity (I max ϳ 96%, Fig. 1). Thus, myristoylation of recoverin does not affect the ability of recoverin to bind and inhibit RK. Several laboratories have reported that myristoylation of recoverin is necessary for its Ca 2ϩ -dependent association with membranes (6 -8), suggesting that it might be involved in recoverin inhibition of RK. Our data provide direct evidence against this idea and are consistent with previous conclusions (4,8) that the association of recoverin with membranes is not necessary for its inhibitory activity toward RK. Myristoylation of recoverin has been shown to radically change the way recoverin binds Ca 2ϩ . From direct Ca 2ϩ binding measurements Ames et al. (10) conclude that nonmyristoylated recoverin has two independent Ca 2ϩ -binding sites with affinities of 110 nM and 6.9 M, whereas the presence of the myristoyl group results in cooperative Ca 2ϩ binding with an apparent K D ϭ 17 M. If this is the case and if the hypothesis proposed earlier by us is correct (4) (namely, that binding of recoverin to RK is necessary and sufficient for the RK inhibition and is a strict function of Ca 2ϩ ), then the difference in Ca 2ϩ binding by recoverin should be observed in the Ca 2ϩ dependence of RK inhibition for the respective forms of recoverin. Indeed, the Ca 2ϩ dependence of RK inhibition is strikingly different for myristoylated and nonmyristoylated recoverin (Fig. 2). Fitting the data with the Hill equation consistently shows a Hill coefficient (n) less than 1 and K1 ⁄2 of ϳ0.35 M for the nonmyristoylated recoverin, while the best fit for the myristoylated form gives n ϭ 1.9 and a much higher K1 ⁄2 (5.3 M; Fig. 2). Thus, myristoylated recombinant recoverin behaves much as the native form with respect to positive cooperativity with n ϳ 2 and a K1 ⁄2 in the micromolar range (4).
Nonmyristoylated recoverin inhibits RK activity with apparent negative cooperativity (n ϳ 0.8). This data, considered with the direct Ca 2ϩ binding study (10), suggests an independent and additive action of the two Ca 2ϩ -binding sites on nonmyristoylated recoverin with respect to RK inhibition rather than true negative cooperativity. The data for nonmyristoylated re-coverin can be fit with the following equation where P is the proportion of maximal phosphate incorporation by RK activity, P max , at the indicated Ca 2ϩ concentration, a and b are amplitude factors that reflect the contribution of the two Ca 2ϩ -binding sites, and K 1 and K 2 are constants that reflect the Ca 2ϩ range of RK inhibition upon the occupancy of these sites (see legend to Fig. 2). If the occupancy of only the first site is sufficient for RK binding and inhibition, then in experiments of the sort shown in Fig. 2 one would expect to find an apparent K1 ⁄2 slightly less than the K D for the high affinity Ca 2ϩ binding site (110 nM) (10). Similarly, if both sites have to be occupied by Ca 2ϩ to observe full inhibition, the observed K1 ⁄2 would be close to the K D of the lower affinity site, in the micromolar range (10). In both cases, the Hill coefficient should equal 1. None of these predictions is seen experimentally, indicating that occupation of each Ca 2ϩ -binding site on recoverin leads to a partial inhibition of RK activity. Our data and conclusions differ, in part, from those of Kawamura et al. (8). Based on Ca 2ϩ titration experiments analogous to the ones shown in Fig. 2, they also conclude that there is essentially no difference in the inhibition of RK by myristoylated and nonmyristoylated recoverin. Such experiments are not sufficient to prove this point. What is required are recoverin titration experiments of the sort shown in Fig. 1. Our data on the Ca 2ϩ dependence of RK inhibition do unequivocally show a difference between myristoylated and nonmyristoylated recoverin. Closer examination of the data of Kawamura et al. reveals a qualitatively similar difference for the two forms of recoverin (Ref. 8, Fig. 2). Possible errors in Ca 2ϩ buffering (see Ref. 4 for discussion) and large error bars may have led the authors to ignore this small difference. Kawamura et al. take their data, that there is no difference in Ca 2ϩ -dependent inhibition of RK by myristoylated and nonmyristoylated recoverin and that nonmyristoylated recoverin does not bind to membranes, to suggest that there are two distinct Ca 2ϩ -binding sites on recoverin. One of these sites is responsible for the inhibition of RK, and the other is responsi- ble for membrane association. Our data (Fig. 2) rule out this model by showing that occupation of each Ca 2ϩ -binding site on recoverin leads to a partial inhibition of RK activity.
The influence of recoverin myristoylation on its Ca 2ϩ -dependent inhibition of RK has also been recently addressed by Chen et al. (5) who report that Ca 2ϩ titration curves of recoverin inhibition by nonmyristoylated and myristoylated recoverins have the same apparent K1 ⁄2 of 3 M free Ca 2ϩ and also that myristoylated recoverin is a much more potent inhibitor of RK than the nonacylated form (apparent K1 ⁄2 is 0.8 M versus 8 M recoverin at saturated Ca 2ϩ ). These results are entirely at variance with our own. We do not attempt to explain this, but note that our findings are consistent with all previous studies of Ca 2ϩ /recoverin/RK interaction (1,2,4,10). It has been shown that nonmyristoylated recoverin binds Ca 2ϩ better (10) and that Ca 2ϩ binding to recoverin is sufficient for binding to and inhibition of RK (1,2,4,5). It follows then that the Ca 2ϩ dependence of nonmyristoylated recoverin inhibition of RK should have a K1 ⁄2 lower than that for myristoylated recoverin. This is exactly what we observe (Fig. 2). Also, similar results have been obtained by Kawamura et al. (8) (see above).
Most of the recoverin found in the retina is acylated by myristate or closely related fatty acids (3). Thus, recombinant myristoylated recoverin should be similar to the native protein.
We find this to be the case both with respect to Ca 2ϩ titration of RK inhibition and recoverin titration at saturated Ca 2ϩ (4; present report). The high half-saturating recoverin concentration that we find for the recombinant protein (ϳ3.5 M) is consistent with what we and others observe for native bovine recoverin (5-7 M (2) and 3.4 M (4)) and for the frog protein (ϳ7.5 M) (1); this contrasts with the K1 ⁄2 of 0.8 M found by Chen et al. (5). Our finding that the myristoyl group has little or no effect on recoverin interaction with RK ( Fig. 1) is consistent with our observation that RK inhibition by Ca 2ϩ -recoverin is not sensitive to relatively high detergent concentrations (as high as 0.4% Tween 80), but Ca 2ϩ /myristoyl-dependent binding of recoverin to membranes is diminished by Tween 80. 2 This suggests that hydrophobic interaction is not of crucial importance for recoverin binding to RK.
It is interesting to note that myristoylation of recoverin has opposing consequences on recoverin function. On one hand, it induces cooperative Ca 2ϩ binding, a feature that allows efficient detection of changes in Ca 2ϩ concentrations in photoreceptors. On the other hand, it significantly increases the Ca 2ϩ range over which inhibition of RK by recoverin occurs in vitro (Fig. 2), bringing it far from the reported in vivo range of free Ca 2ϩ concentrations (200 -600 nM in the darkness and much lower in bright light) (14 -17). In a previous report we point out that a K1 ⁄2 for recoverin inhibition of RK in the micromolar free Ca 2ϩ range under dilute, in vitro conditions is expected, because Ca 2ϩ -dependent membrane association of recoverin effectively reduces this K1 ⁄2 to about 270 nM Ca 2ϩ under more concentrated in vivo conditions (4). This effect arises because Ca 2ϩ -bound myristoylated recoverin binds to membranes and no longer participates in free solution equilibrium. Thus, a higher total concentration of Ca 2ϩ -recoverin results.
From the available data to date we propose two roles for recoverin myristoylation relevant to photoreceptor physiology. First, it imparts a sharp calcium sensitivity range, allowing it to act as more of a "switch" in sensing Ca 2ϩ ; second, it induces Ca 2ϩ -dependent association of recoverin with photoreceptor membranes, allowing recoverin to act in the physiological Ca 2ϩ range of the photoreceptor.