Dimerization of Guanylyl Cyclase-activating Protein and a Mechanism of Photoreceptor Guanylyl Cyclase Activation*

Ca2+-binding guanylyl cyclase-activating proteins (GCAPs) stimulate photoreceptor membrane guanylyl cyclase (retGC) in the light when the free Ca2+concentrations in photoreceptors decrease from 600 to 50 nm. RetGC activated by GCAPs exhibits tight dimerization revealed by chemical cross-linking (Yu, H., Olshevskaya, E., Duda, T., Seno, K., Hayashi, F., Sharma, R. K., Dizhoor, A. M., and Yamazaki, A. (1999) J. Biol. Chem. 274, 15547–15555). We have found that the Ca2+-loaded GCAP-2 monomer undergoes reversible dimerization upon dissociation of Ca2+. The ability of GCAP-2 and its several mutants to activate retGC in vitro correlates with their ability to dimerize at low free Ca2+ concentrations. A constitutively active GCAP-2 mutant E80Q/E116Q/D158N that stimulates retGC regardless of the free Ca2+ concentrations forms dimers both in the absence and in the presence of Ca2+. Several GCAP-2/neurocalcin chimera proteins that cannot efficiently activate retGC in low Ca2+concentrations are also unable to dimerize in the absence of Ca2+. Additional mutation that restores normal activity of the GCAP-2 chimera mutant also restores its ability to dimerize in the absence of Ca2+. These results suggest that dimerization of GCAP-2 can be a part of the mechanism by which GCAP-2 regulates the photoreceptor guanylyl cyclase. The Ca2+-free GCAP-1 is also capable of dimerization in the absence of Ca2+, but unlike GCAP-2, dimerization of GCAP-1 is resistant to the presence of Ca2+.

Photoexcitation of photoreceptors results in hydrolysis of cGMP and closure of the cGMP-gated Na ϩ /Ca 2ϩ channels, thus causing both hyperpolarization of the photoreceptor membrane and decrease in the intracellular free Ca 2ϩ concentrations (1)(2)(3)(4). Lowering of the intracellular Ca 2ϩ concentrations caused by illumination stimulates cGMP synthesis by guanylyl cyclase (retGC) 1 that contributes to recovery from photoexcitation and to light adaptation of photoreceptors (5,6). Recoverin-like Ca 2ϩ -binding proteins, GCAP-1 and GCAP-2, mediate Ca 2ϩ sensitivity of the cyclase in vertebrate retinas (7)(8)(9)(10) so that GCAPs activate the cyclase when the free concentrations of Ca 2ϩ decrease below 100 nM (a characteristic of light-adapted photoreceptors), but they do not stimulate the cyclase in the dark when the free Ca 2ϩ concentrations exceed 500 nM (6,11). GCAP-2 is highly abundant in mammalian rods, while GCAP-1 is highly expressed in cones (9,10,12,13). The cDNA for the third homologue of GCAPs has been recently cloned from a human retinal cDNA library, while it was absent form retinal cDNA libraries of other vertebrate species (14).
GCAP-1 and GCAP-2 interact with the photoreceptor membrane guanylyl cyclases, retGC-1 and retGC-2 (also known as ROSGCs or GC-E and GC-F in Refs. 8 and 15-18), via cyclase intracellular domains (19 -21). Several regions in GCAPs that contain amino acid sequences that are specific for GCAPs function as cyclase regulators have been identified using site-directed mutagenesis (22)(23)(24). Although the exact mechanism for retGC activation by GCAPs remains undetermined, it apparently involves dimerization of the cyclase. Consistently with other membrane guanylyl cyclases being active as dimers (25)(26)(27), retGCs in photoreceptors also form homodimers (28). Moreover, retGC subunits dimerize or at least come into closer contact when stimulated by GCAPs so that they can be chemically cross-linked (29). Here we report the evidence that GCAP-2 undergoes Ca 2ϩ -sensitive dimerization and that GCAP-2 dimerization correlates with its ability to activate retGC. We propose that the dimer of the Ca 2ϩ -free GCAP-2 acts as an adapter that controls dimerization of retGC. Similarly to GCAP-2 a substantial fraction of GCAP-1 also forms a dimer in the absence of Ca 2ϩ , but unlike GCAP-2 this dimerization apparently is not reversed by Ca 2ϩ binding.

EXPERIMENTAL PROCEDURES
Recombinant wild type GCAP-2 (30), its constitutively active mutant (E80Q/E116Q/D158N) (31), and GCAP-2/neurocalcin chimeras (III, IV, XIII, and XIX, Ref. 22) were expressed in Escherichia coli and purified as described previously in detail (22,30,31). Recombinant myristoylated GCAP-1 was produced in E. coli and purified using chromatography on Sephacryl S-100 as described previously (23,32). Neurocalcin was expressed in E. coli and purified using a phenyl-Sepharose chromatography column (22,33). Based on SDS-PAGE, purity of the recombinant proteins was at least 90%. Recombinant proteins were injected in a volume of 200 l into a Superdex 200 HR10/30 column (Amersham Pharmacia Biotech) using an automated fast protein liquid chromatography system and eluted at 0.5 ml/min in buffer A (20 mM Tris-HCl, 50 mM KCl, 10 mM NaCl, 10 mM MgCl 2 , 1 mM dithiothreitol, pH 7.5) containing either 400 M EGTA or 300 M CaCl 2 . The column was pre-equilibrated with 2 volumes of corresponding buffer between the runs. Free Ca 2ϩ concentrations in the samples were adjusted prior to injection by adding EGTA or CaCl 2 , respectively. The molecular weight standards for gel filtration included blue dextran (2,000 kDa), ␣-amylase (200 kDa), alcohol dehydrogenase (150 kDa, Stokes radius: 45.5 Å), bovine serum albumin (66 kDa, Stokes radius: 36.1 Å), carbonic anhydrase (29 kDa, Stokes radius: 20.1 Å) and cytochrome c (12.4 kDa) (Sigma). The column was calibrated prior to the analysis and reproducibility of the calibration was tested between series of the experiments. It was virtually identical in the presence of either EGTA or CaCl 2.
analyzed using TLC on polyethyleneimine-cellulose plates (Merck) essentially as described in full detail in our previous publications (9,30).

RESULTS AND DISCUSSION
The three-dimensional structure of Ca 2ϩ -bound GCAP-2 was recently determined by Ames et al. (34) using NMR spectroscopy. The main chain conformation in GCAP-2 molecule is very similar to that of neurocalcin or recoverin (35)(36)(37), although unlike Ca 2ϩ -loaded neurocalcin GCAP-2, is a monomer. Consistently with the previous report by Vijay-Kumar and Kumar (37) on crystal structure of neurocalcin, in the presence of Ca 2ϩ , neurocalcin elutes from the Superdex 200 column as a dimer (molecular mass ϳ46 kDa, Fig. 1), while it becomes a monomer in the presence of EGTA (molecular mass ϳ22 kDa, Fig. 1). However, Ca 2ϩ has an opposite effect on the chromatographic behavior of GCAP-2. We have found that in the absence of Ca 2ϩ , GCAP-2 can dimerize, while the presence of Ca 2ϩ strongly inhibits its dimerization (Fig. 2).
GCAP-2 has a K d for Ca 2ϩ near 300 nM (34), and activation of retGC by GCAP-2 is also inhibited by Ca 2ϩ (EC 50 200 -300 nM; Refs. 8, 9, and 32). Therefore, at 300 M free Ca 2ϩ GCAP-2 is fully Ca 2ϩ -bound and completely incapable of activating the cyclase (see also inset in Fig. 3). At the saturating free Ca 2ϩ concentration myristoylated wild type GCAP-2 elutes from the column as a protein with a Stokes radius of ϳ22 Å and apparent molecular mass of ϳ31 kDa (Fig. 2), approximately 7 kDa larger than its actual molecular mass (23,808 Da) determined by electrospray mass spectrometry (30).
In addition to the "31-kDa" monomer peak (marked "M" in Fig. 2), the Ca 2ϩ -free GCAP-2 elutes as a peak that corresponds to its dimer (molecular mass ϳ57 Ϯ 6 kDa, Stokes radius: ϳ34 Å, marked "D" in Fig. 2). The GCAP-2 dimer is relatively stable, otherwise it would have completely dissociated during the chromatography (ϳ29 min). Dimerization of GCAP-2 in the presence of EGTA may also be one of the reasons why the Ca 2ϩ -free GCAP-2 exhibits a poor NMR spectrum (34). However, the dimer is not irreversible even in the presence of EGTA, because when the peak of dimer is collected and subjected to the second round of chromatography, GCAP-2 elutes again as two peaks, a dimer and a monomer (data not shown). Also, the dimer collected in the absence of Ca 2ϩ dissociates during the subsequent round of chromatography in the presence of Ca 2ϩ (not shown). Conversely, the Ca 2ϩ -loaded GCAP-2 monomer (as in Fig. 2A, line a) forms a dimer when rechromatographed in the presence of EGTA ( Fig. 2A, line b).
Dimerization and dissociation of the GCAP-2 complex is a function of Ca 2ϩ concentrations and cannot be attributed to differences in ionic strength, because concentrations of CaCl 2 and EGTA are insignificant compared with the total salt concentration in the elution buffer. It can neither be attributed to a nonspecific effect of divalent cations, because both elution buffers contain 10 mM MgCl 2 . Importantly, we can also rule out a possibility that either of the peaks shown in Fig. 2A, line b, represents merely a misfolded inactive form of GCAP-2. When tested in retGC activation assay at low free Ca 2ϩ , the specific activity of GCAP-2 in each peak is similar (Fig. 2B). This demonstrates that each peak contains active GCAP-2 capable of reversible dimerization such that at low Ca 2ϩ concentrations the equilibrium between the dimer and the monomer is shifted toward the formation of the dimer, and at high free Ca 2ϩ it shifts toward the dissociation of the complex into Ca 2ϩ -loaded monomers.
The ability of the Ca 2ϩ -free GCAP-2 to dimerize gives rise to a possibility that GCAP-2 dimerization is directly involved in regulation of retGC. In order to verify this hypothesis, we compared several mutants of GCAP-2 that demonstrated distinct differences in their properties as retGC regulators. We have found that the ability of these mutants to activate retGC at low Ca 2ϩ concentrations correlates with their dimerization (Figs. 3 and 4).
A constitutively active mutant of GCAP-2 (E80Q/E116Q/ D158N, Ref. 31) that stimulates retGC in both low and high free Ca 2ϩ also forms the dimer regardless of the free Ca 2ϩ concentrations (Fig. 3A). The elution profiles of GCAP-2 (E80Q/ E116Q/D158N) in the presence of 0.4 mM EGTA and in 300 M Ca 2ϩ are almost identical. Myristoylation of GCAP-2 that is apparently nonessential for regulation of retGC (30) is also nonessential for dimerization of GCAP-2, because a minor fraction of nonmyristoylated GCAP-2 commonly present in preparations of recombinant fatty acylated GCAP-2 (30) appears in peaks of GCAP-2 dimer as well as monomer (marked "ϪMyr" in Fig. 3B).
In our recent study we constructed several chimera mutants of GCAP-2 (22). We have found that either the substitution of a large N-terminal fragment of GCAP-2 (chimera mutant XIX, see Ref. 22 for details) or the substitution of a region proximal to EF-hand 4 (chimera mutant XIII, Ref. 22) with the corresponding neurocalcin fragments interferes with activation of retGC. When tested for their ability to dimerize in the absence of Ca 2ϩ , neither of these chimeras exhibit a distinct dimer peak in the presence of EGTA (Fig. 4B). In our experiments the elution time for GCAP-2 dimer is approximately 29 min; therefore, even if these mutants were capable of dimerization, their dimers completely dissociated during the chromatography. Thus, our results demonstrate that the ability of these GCAP-2 mutants to dimerize at low Ca 2ϩ concentrations is either completely lost or at least dramatically reduced compared to the wild type GCAP-2. Importantly, these GCAP-2 mutants retain their ability to bind retGC in the presence of Ca 2ϩ , because they are both capable of inhibiting retGC at high free Ca 2ϩ concentrations (22).
Another GCAP-2 chimera protein whose central part was substituted with the corresponding neurocalcin fragment and  (lines a-b) at 10 nM free Ca 2ϩ compared with the basal activity in the absence of GCAPs (filled in black), assayed as described in Ref. 22. B, myristoylated recombinant wild type GCAP-2 (line a) and or its chimera mutants (lines b-e) were analyzed in the presence of 0.4 mM EGTA by highresolution gel filtration as described under "Experimental Procedures" and in the legend to Fig. 2. Peaks corresponding to the wild type GCAP-2 dimer or monomer are marked D and M, respectively. Plus or minus on the right side of each protein construct indicates its ability to activate retGC at low Ca 2ϩ concentrations. C, unlike GCAP-2, GCAP-1 forms stable dimers both in the presence and in the absence of Ca 2ϩ . Recombinant myristoylated GCAP-1 was loaded onto the column in the presence of either Ca 2ϩ (trace a) or EGTA (trace b). The analysis was performed as described under "Experimental Procedures." that fails to activate retGC at low free Ca 2ϩ concentrations (chimera III, Ref. 22) also fails to dimerize in the presence of EGTA (Fig. 4A, line d). However, if we insert into this chimera the exiting helix of EF-2 together with the entering helix of EF-3 derived from GCAP-2 (mosaic chimera IV, Ref. 22), that simultaneously restores the ability of the mosaic GCAP-2 chimera to activate retGC at low Ca 2ϩ concentrations (22) and to dimerize in the presence of EGTA similarly to the wild type GCAP-2 (Fig. 4A, line e).
We propose that dimerization of GCAP-2 is what controls dimerization of retGC and thus contributes to the cyclase regulation by Ca 2ϩ (Fig. 5). This model is based on the following observations: (i) catalytic domains of guanylyl cyclases are very similar to those of membrane adenylyl cyclase (38, 39) that form dimers (40), and peptide receptor membrane guanylyl cyclases also function as dimers (25). Likewise, retGCs form homodimers in photoreceptors in vivo (28). (ii) Activation of retGC by GCAPs stimulates cyclase dimerization or at least closer contacts between the subunits in retGC dimers (29). (iii) GCAP-2 is known to bind to retGC both in the presence and in the absence of Ca 2ϩ (31,41), but only the Ca 2ϩ -free GCAP-2 activates retGC, while the Ca 2ϩ -loaded GCAP-2 inhibits it. (iv) The ability of GCAP-2 and its mutants to activate retGC at low free Ca 2ϩ concentrations correlates with its ability to dimerize, indicating that not only GCAP-retGC interaction, but also GCAP-GCAP interaction, is important for the regulation of the cyclase.
Our general hypothesis is that retGC becomes activated because GCAP dimer acts as an adapter for dimerization of the cyclase. GCAP-2 can be always bound to the cyclase through interaction with its kinase homology and/or catalytic domains regardless of the free Ca 2ϩ concentrations. However, when GCAP-2 is loaded with Ca 2ϩ in the dark, it is unable to promote the formation of retGC dimers. Conformational changes in GCAP-2 caused by dissociation of Ca 2ϩ result in strong GCAP-GCAP interaction such that a Ca 2ϩ -free GCAP-2 dimer brings the two subunits (or parts of the subunits) of retGC closer together and stabilizes the interaction between the catalytic domains in retGC subunits. Dimerization is reversed when the free Ca 2ϩ concentrations increase in the dark. It cannot be excluded that cyclase subunits may also undergo reversible spontaneous dimerization in the absence of GCAPs in vitro, which may account for relatively high basal activity of retGC in washed photoreceptor membranes in the absence of GCAPs (8,9). If Ca 2ϩ -loaded GCAP-2 monomers attached to the cyclase subunits interfere with spontaneous dimerization of retGC, it would explain why a fully Ca 2ϩ -loaded GCAP inhibits basal activity of retGC in vitro (31,32).
It remains unclear whether or not such "dimer adapter" mechanism is applicable to retGC activation by GCAP-1. Similarly to GCAP-2, GCAP-1 also forms dimers in the absence of Ca 2ϩ (Fig. 4C) and is apparently able to interact with similar peptide fragments derived from the cyclase sequence (42). Moreover, activation of retGC by GCAP-1 also results in tight dimerization of the cyclase or at least closer interaction between its subunits (29). Nevertheless, the ability of GCAP-1 to dimerize alone would not be sufficient to account for the activation of the cyclase. There is an obvious difference between GCAP-1 and GCAP-2 at high Ca 2ϩ concentrations. The retention time for the Ca 2ϩ -free dimer of GCAP-1 is slightly different from that of the Ca 2ϩ -loaded dimer of GCAP-1, which suggests that Ca 2ϩ affects conformation of the whole dimer, but apparently does not change the GCAP-1 dimer stability (Fig. 4C). Conformational changes in Ca 2ϩ -loaded versus Ca 2ϩ -free GCAP must play a critical role in the interaction between the GCAPs and the cyclase. However, only in case of GCAP-2 such changes also strongly affect stability of GCAP/GCAP dimer. That may account for the different specificity of GCAP-1 versus GCAP-2 relative to different isoforms of the cyclase (14,20,43) and may also contribute to the fact that similar mutations in GCAP-1 and GCAP-2 can result in different biochemical phenotypes (22-24, 31, 44).
We outline in the model described in Fig. 5 the simplest hypothesis that a dimer of GCAP activates RetGC catalytic domains by causing dimerization of the cyclase. It may be a complete dissociation/association of the cyclase subunits that regulates its activity. However, there are also indications that membrane guanylyl cyclases in photoreceptors may always exist as homodimers (28), and therefore it is possible that dimers of GCAPs can also affect the conformation of retGC dimers rather than the equilibrium between association and dissociation of the cyclase. Based on chemical cross-linking experiments (29), GCAP-activated retGC can either undergo association from two completely separate cyclase monomers or equally likely change the conformation of a pre-existing retGC dimer so that it results in a closer contacts between the two cyclase subunits. Conformational changes and dimerization of a Ca 2ϩ -free GCAP-2 (or just conformational changes in GCAP-1 dimer) may provide the necessary energy to cause either association of retGC subunits or the intramolecular rearrangements within the pre-existing "flexible" dimer of the cyclase. Also, GCAP dimer does not necessarily act as an immediate "bridge" between the two catalytic domains of the cyclase. Instead, it may cause conformational changes in other parts of the cyclase subunits, and these changes could be then transduced through the structure of cyclase subunits to their catalytic domains.
The overall three-dimensional structure of the Ca 2ϩ -bound monomer of GCAP-2 is similar to that of recoverin or neurocalcin (34 -38). Neurocalcin dimers formation results from close FIG. 5. Putative mechanism of retGC activation via dimerization of GCAP-2. GCAP-2 can bind to the cyclase both when in Ca 2ϩloaded or Ca 2ϩ -free form. However, in the presence of Ca 2ϩ the affinity between the two Ca 2ϩ -loaded GCAP-2 subunits is low; therefore the interaction between the catalytic subunits in retGC homodimer is weak and the basal activity of the enzyme is low. Dissociation of Ca 2ϩ causes conformational changes in GCAP-2 that increase the affinity of GCAP-2-GCAP-2 complex. Dimerization of the Ca 2ϩ -free GCAP-2 molecules bound to the cyclase subunits promotes tighter dimerization of the cyclase subunits and thus activates cGMP synthesis. The process is reversed when the free Ca 2ϩ concentrations increase to the level characteristic of dark-adapted photoreceptors, and GCAP-2 becomes Ca 2ϩ -bound.
interactions between EF-hands 2, 3, and 4 in its subunits. Hence, one could assume that the corresponding residues on the surface of GCAP-2 may also be involved in dimerization. However, there is an obvious difference between GCAP-2 and neurocalcin, since GCAP-2 forms relatively stable dimers only in the absence of Ca 2ϩ . Besides, the dimers of Ca 2ϩ -free GCAP-2 are less stable compared with Ca 2ϩ -loaded neurocalcin, because they partially dissociate during the time of chromatography (compare Figs. 1 and 2). Further study will have to determine what sites in GCAP-2 may play critical role in its dimerization and what sites are directly involved in interaction with the cyclase. It will also require additional study to determine how Ca 2ϩ binding changes the kinetic parameters of the GCAP-2 dimerization.