Mapping Sites in Guanylyl Cyclase Activating Protein-1 Required for Regulation of Photoreceptor Membrane Guanylyl Cyclases*

Guanylyl cyclase activating protein (GCAP)-1 regulates photoreceptor membrane guanylyl cyclase, RetGC, in a Ca2+-sensitive manner. It contains four Ca2+-binding motifs, EF-hands, three of which are capable of binding Ca2+. GCAP-1 activates RetGC in low Ca2+ and inhibits it in high Ca2+. In this study we used deletion and substitution analysis to identify regions of GCAP-1 sequence that are specifically required for inhibition and activation. A COOH-terminal sequence within Met157 to Arg182 is required for activation but not for inhibition of RetGC. We localized one essential stretch to 5 residues from Arg178 to Arg182. Another sequence essential for activation is within the N-terminal residues Trp21 to Thr27. The region between EF-hands 1 and 3 of GCAP-1 also contains elements needed for activation of RetGC. Finally, we found that inhibition of RetGC requires the first 9 amino-terminal residues of GCAP-1, but none of the residues from Gln33 to the COOH-terminal Gly205 are specifically required for inhibition. The ability of GCAP-1 mutants to regulate RetGC was tested on total guanylyl cyclase activity present in rod outer segments. In addition, the key mutants were also shown to produce similar effects on recombinant bovine outer segment cyclases GC1 and GC2.

RetGC1 and RetGC2 are membrane guanylyl cyclases that catalyze the conversion of GTP into cyclic GMP (cGMP) in vertebrate rods and cones. RetGCs are implicated in restoring cGMP levels following activation of cGMP phosphodiesterase by light (1)(2)(3)(4)(5). The ability of RetGC to catalyze cGMP synthesis is sensitive to Ca 2ϩ but only in the presence of guanylyl cyclase activator proteins (GCAPs) 1 (6 -8).
GCAPs were identified and purified as Ca 2ϩ -binding proteins that impart Ca 2ϩ sensitivity to RetGCs in vitro (9,10). Two isoforms, GCAP-1 and GCAP-2, have been found. Both stimulate RetGC activity in homogenates of rod outer segments (ROS) at low free Ca 2ϩ concentrations (below 200 nM) and inhibit it at high free Ca 2ϩ concentrations (11,12). GCAP-1 and GCAP-2 share the following primary structural features: (i) four EF-hand motifs in the core of the protein, three of which bind Ca 2ϩ , (ii) an acylated NH 2 terminus, and (iii) a molecular mass of roughly 24 kDa. However, there are significant differences between GCAP-1 and GCAP-2. They display little sequence conservation in their NH 2 and COOH termini. Also, a naturally occurring point mutation (13) affects the two proteins differently (14,15).
The ability of GCAPs to regulate RetGC in a Ca 2ϩ -sensitive manner is well established (7,10,11,16). However, specific structures within GCAPs that are responsible for regulating RetGCs have not yet been clearly defined. Peptide competition experiments have suggested that three structures in GCAP-1 are involved in activation of RetGCs. The first is between residues Gly 2 and Glu 28 . The second one is contiguous with the first; it runs from Glu 28 to Glu 57 (9), and the third one is the EF-hand 3 motif (17).
In order to more precisely define sites in GCAP-1 that interact with RetGC, we constructed deletion mutants of GCAP-1 and chimeras of GCAP-1 with recoverin, a closely related Ca 2ϩbinding protein that does not regulate RetGC. Chimeras were used in cases where deletions were not desirable. For instance, deletions from the NH 2 terminus are especially likely to complicate folding. Moreover, mere deletions would change the length of the peptide, thus introducing another variable into the experiments. Chimeras with recoverin, on the other hand, allowed us to conserve the total length of the constructs and improve the chances of proper folding.
These assumptions were borne out by the fact that nearly all constructs displayed one or more kinds of assayable activity: (i) ability to stimulate RetGC in low Ca 2ϩ , (ii) ability to inhibit it in high Ca 2ϩ , (iii) ability to block activation of RetGC by wt GCAP-1 in low Ca 2ϩ , or (iv) ability to block activation by a Ca 2ϩ -insensitive GCAP-1 mutant in high Ca 2ϩ .
In the study described here we have addressed the following specific questions. Is regulation of RetGCs by GCAP-1 mediated by a single contiguous stretch of GCAP-1 sequence? Or, if multiple regions of the sequence contribute to the interaction and regulation, what are they and what are their roles in regulating RetGC? EXPERIMENTAL PROCEDURES DNA Constructs-All mutants were derived from bovine GCAP-1 and recoverin cDNA clones (18). All chimeras were generated by the polymerase chain reaction-based "splicing by overlap extension" method (19). Cloned Pfu polymerase (Stratagene) was used in all polymerase chain reactions. GCAP-1 truncations were generated by introducing a stop codon into the reverse polymerase chain reaction primers used to amplify the cDNA. It was found that the wt sequence of GCAP-1 does not provide for complete myristoylation of the protein in our expression system even at saturating myristate concentrations in the media and with overexpression of yeast NMT. The wt GCAP-1 sequence does not have a Ser in the sixth position. This residue is part of the myristoylation consensus (20). The GCAP-1 sequence was mutated to encode Ser in the sixth position. This substitution alone provided for complete myristoylation of GCAP-1 as confirmed by mass spectrometry. The properties of this D6S GCAP-1 in regard to activat-ing and inhibiting the cyclase were found to be indistinguishable from those of fully myristoylated wt GCAP-1. For the sake of brevity the D6S GCAP-1 is referred to as wt in the rest of the paper. A Ca 2ϩ -insensitive GCAP-1 mutant was produced by the following substitutions in EFhands 2, 3, and 4: E75Q, E111Q, D144N. The mutagenesis strategy is described in Ref. 11. The mutant protein activated RetGC in high Ca 2ϩ in our assay system. Some constructs were confirmed by DNA sequencing. The masses of all of the expressed proteins were confirmed to be correct by electrospray mass spectrometry. All constructs were at least 90% myristoylated.
Expression of GCAP-1 Mutants-The cDNA constructs were ligated into pET11d or pET11a vectors (Novagen) using the NcoI or NdeI sites respectively at the 5Ј and BamHI site at the 3Ј end. The expression plasmids were transformed into Escherichia coli (BL21 DE3pLysE) that harbored p88131 encoding yeast N-myristoyl transferase (NMT) and kanamycin resistance. Expression was carried out essentially as described in Ref. 11. 30 min prior to induction of expression with 1 mM isopropyl-1-thio-␤-D-galactopyranoside bacterial media were supplemented with free myristic acid. After expression (2-5 h) cells were collected and sonicated on ice in lysis buffer: 40 mM Tris-HCl (pH 7.5), 0.5 mM EDTA, 1 mM ␤-mercapthoethanol, 20 g/ml leupeptin, and 100 M phenylmethylsulfonyl fluoride. The resulting lysate was centrifuged at 30,000 ϫ g. Most of the expressed protein was recovered in the pellet. The pellet was washed twice with lysis buffer by resuspension and centrifugation at 30,000 ϫ g. It was then dissolved in 6 M urea and dialyzed 3 times against 1000 volumes of lysis buffer without protease inhibitors. The renatured fraction typically contained 50% or more of the expressed protein. Construction of the ⌬LQ truncation mutant required an introduction of non-endogenous TAA stop codon. The use of the endogenous stop codon in this truncation produced a read-through peptide of a higher than predicted molecular weight.
Expression of RetGC1 and RetGC2-Bovine GC1 and GC2 were expressed in the HEK293 cell line. The expression vector pCDNA3 (Invitrogen) containing corresponding cDNAs was a gift from Dr. R. Sharma (24). Cells at 80% confluency were transfected using calcium phosphate. 15 g of vector DNA was used per 100-mm dish. Cells were harvested 48 h after transfection and lysed by passaging three times through a 26-gauge needle in hypotonic buffer. A 500 ϫ g supernatant was collected and centrifuged at 400,000 ϫ g for 10 min. The resulting pellet was resuspended in the buffer containing 10 mM Tris (pH 7.5) and 10 mM ␤-mercapthoethanol to the concentration of 4 g/l of total protein as measured by the Bradford assay.
Circular Dichroism-All experiments were performed on circular dichroism spectrometer 62A DS from AVIV TM , Lakewood, NJ, in a 1-mm optical path cell. We used purified proteins at 20 -30 M in 10 mM phosphate buffer (pH 7.0) and 50 M EDTA. Denaturation curves were obtained by monitoring ellipticity at 222 nm. Ellipticities were normalized according to the formula: ⌰ MRW ϭ ⌰ o M r /lc, where ⌰ o is observed ellipticity in degrees, M r is the average molecular weight of an amino acid in the protein, l is the optical path length in mm, and c is protein concentration in grams/liter. GC Assays-The expressed proteins were assayed for their ability to regulate RetGCs in parallel with wtGCAP-1 and nonspecific protein (BSA or recoverin). The assays were carried out as described previously (10). In brief, rod outer segments were washed to remove endogenous GCAPs and were then assayed for GC activity under infrared illumination. Ca 2ϩ concentrations were controlled by 1 mM EGTA or EGTA/ Ca 2ϩ buffers. The substrate was 5 mM cold GTP and 0.1 Ci of [␣-32 P]GTP (Amersham). The reactions were carried out at 30°C for 20 min, and the products were analyzed by TLC and scintillation counting. Typically, synthesized cGMP was labeled to 1,000 -10,000 cpm. The background was typically 50 cpm. The amount of cGMP hydrolysis was controlled in every reaction by adding 25 mM cold cGMP and 20,000 dpm of 3 H-labeled cGMP. For recombinant cyclases each assay point contained membranes with 10 g of total membrane protein.

Recoverin Does Not Regulate RetGC in Our Experimental
System-In order to perform an interpretable analysis of chimeras we first established that recoverin is not a regulator of RetGC in our system. Bovine recoverin and GCAP-1 share roughly 30% amino acid sequence identity (Fig. 1A). Previous studies have shown that pure recoverin does not stimulate photoreceptor guanylyl cyclase (RetGC) (21). To confirm this result in our system and to determine whether or not recoverin inhibits RetGC we assayed GC activity in washed ROS membranes titrated with recombinant myristoylated recoverin (Fig.  1B). Recoverin did not stimulate RetGC even at concentrations Identical residues are boxed. EF-hands are underlined in dashed lines. B, purified recombinant myristoylated GCAP-1 and recoverin were added to washed ROS membranes, and RetGC activity was assayed in 1 mM EGTA. Ⅺ denotes GCAP-1, E denotes recoverin. Solid lines represent guanylate cyclase activity, broken and dashed lines represent cGMP levels. The left y axis is plotted in percent of maximal GC activity, the right y axis in percent unhydrolyzed cGMP recovered from the assay. C, GC activity was assayed in the presence of Ͼ10 M Ca 2ϩ . The data shown are the average of duplicate data points in one experiment. They are representative of two or more independent experiments. up to 30 M whereas GCAP-1 stimulated it 4-fold at 10 M concentration. Similarly, recoverin did not inhibit RetGC in Ͼ10 M Ca 2ϩ while GCAP-1 did (Fig. 1C).
These results confirmed that recoverin indeed does not regulate RetGC in our assay. None of the structural elements in our chimeras derived from recoverin are in themselves sufficient to regulate RetGC. This suggested that we could indeed use chimeric proteins to identify GCAP-1-specific structural elements that are responsible for activating and inhibiting RetGC.
Lack of Phosphodiesterase Activation in the Assay System-It is a formal possibility that the changing levels of cGMP in our assay system result from variations in PDE activity present in ROS preparations. We monitored hydrolysis of cGMP in all our assays as described under "Experimental Procedures." As evident from the dashed lines on Fig. 1, B and C, the level of cGMP hydrolysis did not depend on increasing concentrations of GCAP-1 and recoverin in low as well as high Ca 2ϩ . Similarly we observed no effect on cGMP hydrolysis by any of the mutants we produced (data not shown). We were able to conclude that the varying amounts of cGMP in our assay system result solely from varied guanylyl cyclase activity.
The Role of the COOH Terminus-For the purpose of this work we consider the COOH terminus of GCAP-1 as the residues from Phe 156 at the end of EF-hand 4 to the very COOHterminal Gly 205 . To study the role of the COOH terminus in regulating RetGC we constructed several truncation mutants and chimeras with the COOH terminus of recoverin (Fig. 2, A  and B). The truncation mutant that ended after Gly 159 (⌬VQ) had only 7% of the stimulatory activity of wt GCAP-1 (low Ca 2ϩ conditions) when its concentration in the assay was 25 M (Fig.  3A). The stimulatory effect of GCAP-1 saturated below 10 M (Fig. 1B).
The longer truncation mutants ⌬SL (ends at Arg 172 ) and ⌬RI (ends at Thr 176 ) also do not stimulate RetGC. In fact they suppress RetGC in low Ca 2ϩ below the basal level (Fig. 3A).
The truncation mutant ⌬LQ ends after Arg 182 . In contrast to the shorter deletion mutants it stimulated RetGC to 93% of the wt GCAP-1 level when assayed at 50 M. Essentially, only ⌬LQ of all truncation mutants described here is capable of activating RetGC to significant levels.
The truncation mutants that failed to activate RetGC do inhibit RetGC in high Ca 2ϩ (Ͼ10 M) as shown in Fig. 3B. They also block activation by wt GCAP-1 in low Ca 2ϩ in a competition experiment with a half-maximal effect reached at a molar excess of 35-100 (data not shown). Even the most extensive COOH-terminal truncation mutant, ⌬VQ, and the EF4 chimera inhibited the cyclase in high Ca 2ϩ . The EF4 chimera contains GCAP-1 sequence from the NH 2 terminus down to Phe 156 following EF-hand 4 (Fig. 2). The rest of the chimera consists of Ile 172 to Leu 202 of recoverin. The length of this chimera exceeds the lengths of ⌬VQ, ⌬SL, and ⌬RI.
Since the EF4 chimera does not activate RetGC (Fig. 3A) a specific sequence in the COOH-terminal region is required for activation, not simply any sequence of a suitable length. More precisely the presence of the sequence RIVRR flanked by Arg 177 and Arg 182 appears to be crucial for activation but not for inhibition. The residues COOH-terminal of Arg 182 are not essential for stimulating RetGC. The actual structural requirements provided by the RIVRR structure are not yet clear. Results of a preliminary alanine scanning mutagenesis study A minimum 2-fold difference from the negative control (BSA) was considered as a positive effect (؉). Less than 2-fold difference was considered as no effect (Ϫ). ⌬LQ is the shortest truncation mutant that can activate GC. The difference between ⌬LQ and the next shortest truncation ⌬RI is the sequence RIVRR. The data shown are the average of duplicate data points in one experiment. They are representative of two or more independent experiments. Basal GC activity in the presence of a nonspecific protein (BSA) was taken to be 0 and % maximal activation was calculated using the formula: %(X) ϭ ((activity X Ϫ basal activity)/ (maximal stimulated activity Ϫ basal activity))*100%. A saturating amount of GCAP-1 was taken as 100% activation (3-5-fold stimulation depending on the ROS preparation). Suppression of the activity below the basal level is represented as a negative value. B, inhibition of GC in high Ca 2ϩ by COOH-terminal mutants. e denotes ⌬VQ; ᭜, ⌬SL; f, EF4; ⅜, GCAP-1. The mutant proteins were tested in a GC assay using washed ROS membranes. All assays were performed in a 25-l volume in the presence of Ͼ10 M free Ca 2ϩ . The data shown are the average of duplicate data points in one experiment. They are representative of two or more independent experiments. Basal GC activity in the presence of a nonspecific protein (BSA) was taken as 100%. Saturation with wt GCAP-1 is reached at 2.5 M.
suggest that none of the specific residues within the RIVRR sequence are essential for RetGC regulation (data not shown).
A chimera, EF3-4 Ϫ , has the region between EF-hands 3 and 4 substituted with the corresponding recoverin sequence (see Fig. 7 and discussion on core sequences below). It can stimulate RetGC as shown in Fig. 8A. Based on the EF3-4 Ϫ chimera and the ⌬LQ truncation mutant we conclude that all elements essential for RetGC activation that lie in the COOH terminus are localized within residues Glu 155 and Arg 182 .
The Role of the NH 2 Terminus-We consider the NH 2 terminus of GCAP-1 as residues from Gly 2 to Thr 27 . It is 31% identical to the corresponding region of recoverin. Since recoverin does not regulate RetGC, we constructed and analyzed chimeras that have increasing portions of the GCAP-1 NH 2 terminus replaced by recoverin (Fig. 4). The "VEEL" chimera with sequence from the NH 2 terminus to Val 10 replaced by recoverin stimulates RetGC in low Ca 2ϩ (data not shown). The chimera referred to as "WYK" has recoverin sequence from the NH 2 terminus to Trp 21 . This chimera also stimulated RetGC as shown in Fig. 5A. However, replacing only 6 more residues of the native GCAP-1 sequence produced a chimera, "TEC," that was completely inactive (Fig. 5A). This can be because TEC lacks sequence elements necessary to activate RetGC. Alternatively, misfolding could cause TEC to be inactive.
In order to evaluate the folding state of the TEC chimera we used circular dichroism (Fig. 6). GCAP-1 displayed a spectrum with an ellipticity 1.5 times higher than that of TEC. However, the shape of the two spectra are virtually indistinguishable and both are characteristic of a folded protein (Fig. 6A). The ellipticity at 222 nm decreased as a function of temperature in a similar fashion for TEC and wt GCAP-1 (Fig. 6B). Ellipticity at this wavelength is indicative of the helical content of a protein.
It is routinely used to monitor temperature denaturation of proteins.
These data suggest the ␣-helical content of TEC is similar to that of GCAP-1 at room temperature. The smaller ellipticity of TEC may be explained by the tendency of the recoverin NH 2 terminus to stay unfolded (22,23). However, the CD spectra indicate that most, if not all, of TEC is indeed folded. Since WYK activates RetGC and TEC does not, we conclude that the GCAP-1 sequence from Trp 21 to Thr 27 , WYKKFMT, is required for activation. Even though this sequence is essential, other residues in the core also contribute to activation. This is apparent from the properties of core substitution mutants we describe in the following section. A summary of the NH 2 -terminal chimeras and their properties is presented in Fig. 4B.
An essential inhibitory structure also resides within the GCAP-1 NH 2 terminus. None of the recoverin/GCAP-1 chimeras VEEL, TEC, and WYK inhibit RetGC in high Ca 2ϩ (Fig.  5B). Despite its ability to stimulate, WYK did not block activation of Ret GC by a Ca 2ϩ -insensitive GCAP-1 mutant in high Ca 2ϩ at up to 30-fold molar access (data not shown). The NH 2 terminus is not in itself sufficient for inhibition, however. A chimera, "ECP," consisting of the complete NH 2 terminus from GCAP-1 up to EF-hand 1 and the rest of the sequence from recoverin fails to inhibit RetGC (Fig. 5B).
The Role of the Core Sequences-We consider the sequence between Glu 28 and Met 157 as the core of GCAP-1; it includes EF-hands 1 through 4 (Fig. 7A). We constructed chimeras that replaced native GCAP-1 core sequences with the corresponding FIG. 4. NH 2 -terminal constructs. A, an alignment of the NH 2terminal sequences of GCAP-1, GCAP-2, and recoverin. B, black denotes recoverin sequences, white denotes GCAP-1. A minimum 2-fold difference from the negative control with BSA was considered as a positive effect (؉). Less than 2-fold difference was considered as no effect (Ϫ). The sequence WYKKFMT appears to be critical for the ability to activate GC. sequences of recoverin. The "EF4" chimera was spliced at Phe 156 at the end of EF-hand 4 giving it the least recoverin and most GCAP-1 sequence. We also produced chimeras spliced after Glu 111 at the end of EF-hand 3 ("EF3") and at Val 77 at the end of EF-hand 2 ("EF2").
EF4, EF3, and EF2 inhibit RetGC in high Ca 2ϩ although the concentrations required for inhibition are higher than for wt-GCAP-1 (Fig. 8B). The EF2 chimera, which has the least GCAP-1 sequence, also blocked activation of RetGC in high Ca 2ϩ by a Ca 2ϩ -insensitive GCAP-1 mutant (data not shown). Out of EF1-2 Ϫ , EF2-3 Ϫ , and EF3-4 Ϫ , only EF3-4 Ϫ failed to inhibit RetGC (data not shown). This may suggest that the region between EF-hands 3 and 4 is involved in inhibiting RetGC. Alternatively, the recoverin sequence introduced into this chimera may interfere with the correct conformation required for inhibition.
None of the chimeras with the C terminus replaced by recoverin (EF4, EF3, and EF2) activate RetGC (Fig. 8A). This agrees with our finding described above that the COOH-terminal RIVRR structure is needed for activation. This sequence is not present in recoverin.
The chimera EF3-4 Ϫ activates RetGC in low Ca 2ϩ by 2-fold above the basal level. This constitutes 28% of the wt GCAP-1 level of activation in this experiment (Fig. 8A, inset). In this chimera the GCAP-1 sequence between Glu 111 and Phe 156 is replaced by recoverin. Since the conservation between GCAP-1 and recoverin is quite low here, we suggest that this region is not essential for RetGC activation.
The regions between EF-hands 1 and 2 and between 2 and 3 could not be replaced without complete loss of the ability to activate RetGC. It appears unlikely that this whole 71-amino acid stretch interacts with the cyclase. Rather, it may provide for the proper configuration of the activating elements that we identified in the NH 2 and COOH termini. Since both EF1-2 Ϫ and EF2-3 Ϫ can inhibit RetGC in high Ca 2ϩ it appears that they can bind to the cyclase but fail to activate it.
Effects of Key Mutants on Recombinant RetGC1 and RetGC2-To study the effects our mutants may have on the known retinal guanylyl cyclases we tested several mutant proteins on bovine recombinant RetGC1 and GC2 referred to as OS GC1 and OS GC2. Fig. 9 shows the effects of key mutants in low Ca 2ϩ . The key COOH-terminal truncations, ⌬LQ and ⌬RI, exhibited regulatory properties toward recombinant RetGC1 and GC2 that are similar to those of the total ROS guanylyl cyclase activity. The longer mutant ⌬LQ stimulated RetGC1 by 10-fold, while the shorter mutant ⌬RI which lacks the critical structure represented by "RIVRR" sequence had a less than 2-fold effect on GC1. For the less active recombinant RetGC2 the effects were: 2.6-fold for ⌬LQ and under 2-fold for ⌬RI. Similarly the key NH 2 -terminal chimeras WYK and TEC affected the recombinant cyclases much like the RetGC activity in ROS preparations. Both for expressed RetGC1 and GC2 WYK stimulated the activity and TEC did not have a significant effect (Fig. 9). We conclude that the essential GCAP-1 stimulatory sequences we identified in the native system are also essential for activation of recombinant RetGC1 and GC2. We were not able to assess the inhibitory effects of our mutants on recombinant RetGC1 and GC2 due to the very low basal activity of the bovine recombinant cyclases. DISCUSSION In this study we evaluated the ability of deletion mutants and GCAP-1/recoverin chimeras to regulate RetGC. The ability   FIG. 7. The core chimeras. A shows an alignment of the core sequences of GCAP-1, GCAP-2 and recoverin. Highlighting represents the region necessary for activation. B, EF2, EF3, and EF4 chimeras consist of an NH 2 -terminal GCAP-1 stretch (shown in black) and a COOHterminal recoverin stretch (shown in white). For stimulation and inhibition at least a 2-fold effect on the basal activity of GC was considered as positive (؉). Less than 2-fold difference was considered to be no effect (Ϫ).

FIG. 8. Effects of core chimeras on RetGC activity in low Ca 2؉ .
A, the chimeras were tested in a GC assay using washed ROS membranes in the presence of 1 mM EGTA. Basal GC activity in the presence of a nonspecific protein (BSA) was taken to be 0 and % maximal activation was calculated using the formula: %(X) ϭ ((activity X Ϫ basal activity)/(maximal stimulated activity Ϫ basal activity))*100%. The following concentrations were used: GCAP-1 at 1 M, EF1-2 Ϫ at 100 M, EF2-3 Ϫ at 50 M, EF3-4 Ϫ at 58 M, EF2 21 M, EF3 at 26 M. Chimeras EF3, EF2-3 Ϫ , and EF2 suppressed RetGC below the basal level. EF3-4 Ϫ displayed the ability to stimulate RetGC 2-fold above the basal level. The inset shows a titration of ROS with EF3-4 Ϫ protein in low Ca 2ϩ . ‚ denotes EF3-4 Ϫ , " denotes basal activity. The data shown are the average of duplicate data points in a one experiment. They are representative of two or more independent experiments. B, inhibition in high Ca 2ϩ by core chimeras. ‚ denotes EF2, q EF3, f EF4, ϩ denotes HINGE, Ⅺ GCAP-1. The chimeras were tested in a GC assay using washed ROS membranes. All assays were performed in the presence of Ͼ10 M free Ca 2ϩ . The data shown are the average of duplicate data points in one experiment. They are representative of two or more independent experiments. Basal GC activity in the presence of BSA was taken to be 100%.
of each protein to stimulate RetGC was studied at low free Ca 2ϩ levels buffered by EGTA. The ability to inhibit was assayed at Ͼ10 M free Ca 2ϩ . Here we correlate the effects of the mutants on RetGC activity with the presence of specific GCAP-1 sequences. We use this correlation to map GCAP-1 sequences critical for RetGC regulation. We do not distinguish here between sequences that directly interact with RetGC and those required for any other reason, e.g. for proper scaffolding of non-contiguous interacting side chains.
Experimental Strategy-In order to simplify the analysis of the mutants, we broke down the sequence into three major stretches: the NH 2 terminus (Gly 2 to Thr 27 ), the core (Glu 28 to Phe 156 ), and COOH terminus (Met 157 to Gly 205 ). In this discussion we take a qualitative approach to describing the ability of each protein to activate and inhibit RetGC. When a mutant was able to regulate RetGC we frequently found that its apparent affinity for RetGC was altered (Table I). This suggests that the chimeras may not reproduce all features of the wild type GCAP-1 conformation correctly. Nonetheless, the stimulation and inhibition of RetGC by these mutant proteins were reproducible. We only considered 2-fold or greater effects as significant. This cut-off clearly differentiated between a specific effect and background variation that we routinely observe with nonspecific proteins (e.g. BSA, recoverin) in our assay. The effect of these proteins on GC activity typically does not exceed 10% of the basal level.
The presence of two distinct guanylyl cyclases, RetGC1 and RetGC2, is established in ROS of humans and other species (1)(2)(3)(4)(5). These cyclases are referred to as ROS GC1 and ROS GC2 in the literature. At present it is not clear if they account for all cyclase activity in ROS or if other cyclases are also present. In this study we have focused on the regions of GCAP-1 which are essential for the interaction with cyclases. We therefore used ROS preparations to make all cyclases that are regulated by GCAP-1 in vivo available to the GCAP-1 mutants in our assays. However, we have also confirmed that the key COOH-terminal deletion mutants and NH 2 -terminal chimeras affect recombinant ROS GC1 and ROS GC2 the same way they affect GC activity in ROS preparations (Fig. 9).
The Role of the COOH Terminus-In the COOH terminus of GCAP-1 a structure represented by the sequence RIVRR appears crucial for activation. A mutant truncated immediately after this sequence activates RetGC whereas a truncation that stops immediately before it does not. Paradoxically, none of the residues in the RIVRR sequence seems essential for RetGC activation based on the results of a preliminary point mutagenesis study. Our results localize all elements essential for activation in the COOH terminus to residues from Glu 155 to Arg 182 . None of the structures in the COOH terminus of GCAP-1 are required to inhibit RetGC. All the truncation mutants as well as chimeras with the COOH terminus of recoverin inhibit the cyclase.
The Role of the NH 2 Terminus and the Core-As evident from the TEC chimera the NH 2 terminus is critical for activating RetGC. TEC displays CD spectra resembling those of wt GCAP-1 (Fig. 6, A and B) arguing that it is a folded protein. It does not, however, stimulate RetGC in low Ca 2ϩ (Fig. 5A) nor does it block stimulation by wt GCAP-1 (data not shown). Another chimera, WYK, that included only 7 more residues of GCAP-1 than TEC activates RetGC by over 2-fold. We conclude that these 7 residues, WYKKFMT, are essential for activation.
Replacing the NH 2 terminus of GCAP-1 with recoverin sequence to Ser 9 (as in the VEEL chimera, Fig. 5B) abolishes inhibition but not activation. This shows that a structure within GCAP-1 between Gly 2 and Ser 9 is specifically required for inhibition. It has been shown in a different study that an NH 2 -terminal peptide derived from GCAP-1 blocks activation of RetGC by GCAP-1 (IC 50 of 10 M) (4,8). The role of the NH 2 terminus is summarized in Fig. 4B.  We identified no GCAP-1-specific sequences within the core of the protein (Glu 28 to Phe 156 ) that are required for inhibition. The chimera EF2 with all GCAP-1 sequence from EF-hand 2 to the COOH terminus replaced by recoverin sequence inhibits RetGC (Fig. 8B). EF1-2 Ϫ with GCAP-1 sequence between EFhand 1 and EF-hand 2 replaced by the corresponding recoverin sequence also inhibits RetGC (data not shown). Based on the ability of the EF1-2 Ϫ chimera to inhibit we conclude that the region of GCAP-1 from Gln 33 to Val 77 is not specifically required for inhibition. To summarize, the first 9 amino acids of GCAP-1 are specifically required for inhibition of RetGC in high Ca 2ϩ . Other residues in the NH 2 terminus and the core, however, are likely to contribute to inhibition in a nonspecific way, e.g. by providing scaffolding for inhibitory structures. For example, chimera ECP that contains the whole NH 2 terminus of GCAP-1 down to Thr 27 , with the rest of it derived from recoverin, fails to inhibit RetGC in high Ca 2ϩ .
Chimeras EF1-2 Ϫ and EF2-3 Ϫ do not stimulate RetGC. That shows that the GCAP-1 region between EF-hands 1 and 3 is necessary for RetGC activation. Since this is a long stretch, it appears unlikely that all of it is involved in a direct contact with RetGC. This region of GCAP-1 may provide for the correct scaffolding of activating sequences, while the corresponding region of recoverin does not fulfill this role. The role of the core sequences is summarized in Fig. 7B.
Activation Versus Inhibition-A summary of our findings is shown in Fig. 10. The inhibitory and stimulatory effects of GCAP-1 on RetGC appear to require different GCAP-1 structures. Stimulation requires both the COOH-terminal RIVRR and the NH 2 -terminal WYKKFMT sequences, whereas inhibition appears to require the first 9 amino acids which are distinct from either of the stimulatory determinants. Moreover, structures between EF-hands 1 and 3 are required for activation but not for inhibition.
Comparison with GCAP-2-Both GCAP-1 and GCAP-2 inhibit and stimulate RetGC, yet there are substantial differences in their sequences. In particular the NH 2 and the COOH termini of the two proteins have few common primary sequence features. A parallel study using chimeras of GCAP-2 with neurocalcin (see accompanying article, Ref. 25) showed that a sequence near the COOH terminus of GCAP-2 is specifically required for activation. This correlates with our finding that a specific sequence in the GCAP-1 COOH terminus is essential for activation but not inhibition of RetGC.
The GCAP-2 study also identified a sequence flanking EF1 of GCAP-2 as important for activation and inhibition. According to our results a sequence, WYKKFMT, which flanks EF1 in GCAP-1 is necessary for activation of RetGC. In fact part of this stretch, WYKKF, is conserved between the two proteins.
There are also significant differences between the findings in the GCAP-1 and GCAP-2 studies. GCAP-1 but not GCAP-2 appears to require the 9 NH 2 -terminal residues for inhibition and the region between EF-hands 1 and 3 for activation of RetGC. We have produced a GCAP-1 chimera whose region between Ser 53 and Ile 122 is replaced by the corresponding recoverin sequence. A similar GCAP-2/neurocalcin chimera displayed reversed Ca 2ϩ sensitivity in the GCAP-2 study. GCAP-1/recoverin chimera we have constructed, however, does not exhibit a reversed Ca 2ϩ sensitivity. It inhibits RetGC in high Ca 2ϩ and does not stimulate it in low Ca 2ϩ (data not shown). These differences between GCAP-1 and GCAP-2 may reflect the divergence of the primary sequences of the two proteins. For instance, a mutation in GCAP-1 and GCAP-2 has been shown to affect their activity differently (13)(14)(15). Alternatively, the differences may arise from the experimental systems used to make chimeras in the two studies since recoverin has less homology to GCAP-1 than neurocalcin to GCAP-2. Further structural analyses will be required to clarify the precise functions of the regions identified in these studies for binding, activation, and inhibition of RetGCs.