Molecular Determinants of the Interaction between the Inositol 1,4,5-Trisphosphate Receptor-associated cGMP Kinase Substrate (IRAG) and cGMP Kinase Iβ* 210

Cyclic GMP-dependent protein kinase I (cGKI) affects the inositol 1,4,5-trisphosphate (InsP3)-dependent release of intracellular calcium by phosphorylation of IRAG (inositol 1,4,5-trisphophate receptor-associated cGMP kinase substrate). IRAG is present in a macromolecular complex with the InsP3 receptor type I (InsP3RI) and cGKIβ. The specificity of the interaction between these three proteins was investigated by using the yeast two-hybrid system and by co-precipitation of expressed proteins. The amino-terminal region containing the leucine zipper (amino acids 1–53) of cGKIβ but not that of cGKIα or cGKII interacted with the sequence between amino acids 152 and 184 of IRAG in vitroand in vivo most likely through electrostatic interaction. cGKIβ did not interact with the InsP3RI, but co-precipitated the InsP3RI in the presence of IRAG indicating that IRAG bound to the InsP3RI and to cGKIβ. cGKIβ phosphorylated up to four serines in IRAG. Mutation of these four serines to alanine showed that cGKIβ-dependent phosphorylation of Ser696 is necessary to decrease calcium release from InsP3-sensitive stores. These results show that cGMP induced reduction of cytosolic calcium concentrations requires cGKIβ and phosphorylation of Ser696 of IRAG.

Signal transduction via NO/cGMP/cGKI 1 is involved in a variety of cellular mechanisms including smooth muscle contractility and platelet aggregation (1)(2)(3)(4). cGKI affects smooth muscle tone by either decreasing the release of calcium from InsP 3 -sensitive stores (5)(6)(7)(8)(9) or by reducing calcium sensitivity of the contractile elements (10,11). During the last years, several mechanisms were proposed for the action of cGKI mediating these effects. A decrease of the cytosolic calcium con-centration by cGKI might involve reduced InsP 3 synthesis (12)(13)(14)(15), enhanced calcium re-uptake by intracellular stores via CaATPase (16), or inhibition of calcium release via the InsP 3 R (17). The molecular mechanisms for these different possible intracellular calcium regulation pathways were only partly resolved up to now.
Recently, we identified a 125-kDa cGKI substrate protein which was designated as inositol 1,4,5-trisphophate receptorassociated cGMP kinase substrate (IRAG). IRAG which is phosphorylated by cGKI is associated in a macromolecular complex with cGKI␤ and InsP 3 RI in smooth muscle (9). The observed perinuclear localization of heterologously expressed IRAG suggested the potential role of IRAG as a modulator of calcium release from intracellular stores. Indeed, functional studies revealed that IRAG inhibits InsP 3 -induced calcium release after activation of cGKI␤ with 8-pCPT-cGMP in COS-7 cells (9). However, the precise mechanism by which IRAG influences calcium release is still unknown.
In the present study we investigated the molecular determinants for the interaction of IRAG and cGKI. It is shown that IRAG interacts specifically with the amino-terminal region containing the leucine zipper of cGKI␤. Phosphorylation of Ser 696 of IRAG is essential for the inhibition of InsP 3 -induced calcium release.
Antibodies-A polyclonal antibody specific for IRAG, raised in rabbits against recombinant IRAG 53-499 expressed in bacteria, was used for Western blot analysis at a dilution of 1:1000. Further antibodies were directed against cGKI (8) and InsP 3 RI (ABR Biochemicals).
Plasmid Construction of Baits and Preys-The bovine IRAG cDNA was the template of all IRAG constructs (baits) which were synthesized by PCR. The generated IRAG amplicons were purified and inserted into the BamHI/EcoRI-digested expression plasmid pEG202 (20) in-frame with the DNA-binding domain of LexA, yielding pEG202-LexA/IRAG. The full-length and truncated PCR amplicons of bovine cGKI␣ and cGKI␤ were ligated into the EcoRI/XhoI sites of the pJG4-5 expression plasmid (20) and used as preys. The inserts were fused in-frame to the "acid loop" DNA activation domain of pJG4-5, yielding pJG4-5/cGKI. As cGKII Prkg2 gene from mouse contains an internal XhoI site it was subcloned as EcoRI fragment into pJG4-5. The correct orientation was checked by sequence analysis. The full-length rat InsP 3 RI (S1Ϫ/S2Ϫ) without the channel forming domain was divided into five different fragments. These inserts were amplified by PCR and ligated into pJG4-5 via EcoRI and XhoI in order to obtain various pJG4-5/InsP 3 RI constructs. cGKI␤ was generated as bait through EcoRI/XhoI digestion * This work was supported by the Deutsche Forschungsgemeinschaft, Fond der Chemischen Industrie, and Wilhelm Sander-Stiftung. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. of pJG4-5/cGKI␤ and subsequent cloning of the isolated cGKI␤ fragment into pEG202 digested with the same enzymes. Recombinant plasmids were transformed in Escherichia coli XL1-Blue. The correct reading frame was verified by sequence analysis. All used primer pairs are listed at the supplement.
Two-hybrid Screen-Growth, maintenance, and screening of the yeast cells has been performed as described (20). The reporter plasmid pSH18-34 (20), the different pEG202-LexA/bait plasmids, and their respective pJG4-5/prey plasmids were co-transformed by lithium acetate (21) in the yeast strain EGY48. The expression of IRAG as protein was assayed by standard lysis of cells expressing appropriate constructs followed by SDS-PAGE and immunoblot analysis with antibodies directed against IRAG. Transformants were selected on synthetic dropout agar plates lacking tryptophan, leucine, and histidine. As defined protein interaction sites were investigated in this study, the performed screen was identical with the rescreen described by Ausubel et al. (20). Interaction results in activation of the reporter genes on selective media. Positive clones were identified by their blue color on 5-bromo-4chloro-3-indoyl ␤-D-galactoside (X-Gal) plates and their ability to grow on media lacking leucine in the presence of galactose.
Generation of IRAG Phosphorylation Mutants by Overlap Extension PCR-Mutagenesis of IRAG phosphorylation sites from serine to alanine was performed by overlap extension PCR according to Ho et al. (22). For the PCR-based insertion of point mutations two complementary synthetic oligonucleotide primers carrying the mutation and a DNA fragment flanking primer pair are required. In a first PCR round, two corresponding DNA fragments were generated which were used together as template in a second PCR round with the flanking primer pair. The primer pairs used for the mutation of single phosphorylation sites (S1-S4) and the flanking primers (AJ1-AJ3) for the amplification of the complete IRAGa cDNA are listed in Table III of the supplement. Generation of multiple mutations was performed using single mutant plasmids as template DNA.
Cloning of IRAG-AJ1/AJ2-generated IRAG amplicons were purified and ligated into EcoRI/BamHI-restricted pcDNA3.1 vector (Invitrogen) after digestion with the same enzymes and used for the in vitro phosphorylation assays. For construction of IRAG-GFP fusions used for Fura-2 AM calcium measurements AJ1/AJ3-amplified PCR products were cloned into the EcoRI/BamHI-restricted pEGFP-N3 vector (CLONTECH). All introduced mutations were confirmed by sequence analysis (ABI Prism TM Sequence analyzer, PerkinElmer Life Sciences).
Cell Transfection-COS-7 cells grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) were transiently transfected with IRAG constructs cloned into pcDNA3.1 (for phosphorylation experiments) or pEGFP-N3 (for calcium measurements) by using the calcium phosphate method. After 48 h of incubation at 37°C, cells were harvested for lysate preparation or directly used for calcium measurements, respectively. COS-7 Cell Lysate Preparation and Phosphorylation-COS-7 cells transiently transfected with the different IRAG-pcDNA3.1 constructs, cGKI␤ cloned in pMT3-vector or InsP 3 RI in pcDNA3 were washed twice with phosphate-buffered saline buffer and then harvested in phosphate-buffered saline using a cell scraper. Cells were centrifuged at 600 ϫ g, washed with phosphate-buffered saline buffer, and then suspended in hypoosmotic lysis buffer (10 mM potassium phosphate, pH 7.4). Cells were then freeze-thawed, lysed using a syringe with a 4-gauge needle, and then frozen in aliquots at Ϫ70°C. Lysates from COS-7 cells were assayed for IRAG, cGKI␤, and InsP 3 RI content using standard immunoblot analysis with specific antibodies. Cell lysates containing equal amounts of IRAG and cGKI␤ were used for phosphorylation reactions. Cells (20 -30 g of protein) were incubated in 50 mM Mes, pH 6.9, 10 mM NaCl, 1 mM MgAc, 0.4 mM EGTA, 0.1% Lubrol-PX (Sigma), 0.1 mM [␥-32 P]ATP (2,000 cpm/pmol, Amersham Pharmacia Biotech) in the presence or absence of 8-pCPT-cGMP (3 M, Biolog) for 15 s up to 2 min at 30°C. The reactions were stopped by addition of Laemmli buffer. Proteins were separated by SDS-PAGE and blotted to polyvinylidene difluoride membrane (Whatman). Incorporated radioactivity was visualized by autoradiography and phosphoimage analysis (BAS-1500, Fuji, Raytest). Quantification of phospholuminescence was performed using the AIDA 2.0 image analysis program.
Measurement of Intracellular Calcium-COS-7 cells were loaded with Fura-2 AM (5 M, ICN) before measuring [Ca 2ϩ ] i by the dualwavelength microfluorescence technique (9). Analysis of the area under the curve of calcium transients was performed using the software program Microcal TM Origin 6.0.

IRAG Interacts with the Amino-terminal Region
Containing the Leucine Zipper of cGKI␤-IRAG is a new compound of the NO/cGMP signaling pathway which, in association with InsP 3 R and cGKI␤, negatively regulates InsP 3 -induced calcium release. It has been shown that IRAG and InsP 3 R are phosphorylated after addition of 8-pCPT-cGMP by cGKI. IRAG, InsP 3 R, and cGKI associate to a multimeric complex which has been purified from microsomal membranes of bovine tracheal smooth muscle (9). To analyze the interaction and identify the interaction sites of IRAG with cGK, a yeast two-hybrid screen was performed using an IRAG derivative (IRAG 53-845 ) lacking the putative NH 2 -and COOH-terminal transmembrane domains as bait and cGKI variants and cGKII as preys (Fig. 1). Strong interaction was observed between IRAG and cGKI␤ but no detectable interaction between IRAG and cGKI␣ or cGKII (Fig. 1, A and B). Further analysis showed that cGKI␤ interacted with its amino terminus with IRAG. This is of interest since the I␣ and I␤ isozymes of cGKI differ in their first 100 amino acids (23). Within the NH 2 -terminal region of cGKI␤, only the part containing the leucine zipper (cGKI␤ 1-53 : cGKI␤-NT (1)) but not the linker (cGKI␤ 54 -104 : cGKI␤-NT (2)) showed association with IRAG (Fig. 1B). Next, we analyzed which part of IRAG interacts with cGKI␤. For this study different IRAG variants were used as baits and full-length cGKI␤ or cGKI␤ variants as prey in the yeast two-hybrid system (Fig.  1, A and C). The amino-terminal region containing the leucine zipper of cGKI␤ interacted with the peptide sequence between amino acids 152 and 184 of IRAG. This sequence of IRAG contains 33 amino acids of which 16 amino acids are charged (Asp, Glu, Lys, and Arg). The corresponding peptide of cGKI␤ has 22 charged amino acids located between the leucine/isoleucine residues of the cGKI␤ leucine zipper (Fig. 1D). Therefore, interaction between these two sites might be mediated by electrostatic interactions. Interestingly, the putative coiled-coil domain present in the IRAG protein is not involved in the complex formation between IRAG and cGKI␤ suggesting that this site might mediate the assembly of IRAG with the InsP 3 RI.
IRAG Mediates the Assembly of cGKI␤ and InsP 3 RI in a Macrocomplex-The interpretations of the two-hybrid system experiments were supported by the results obtained after in vivo expression of the various proteins in COS-7 cells. The two different cGKI isoforms, cGKI␣ and cGKI␤, were separately coexpressed with full-length IRAG and/or InsP 3 RI. Using partially purified or pure proteins as standard (for example, see Fig. 2a in Ref. 9), we estimated that cGKI␤, IRAG, and/or InsP 3 RI were expressed at approximately equal amounts. Phosphorylation of IRAG and co-precipitation of the proteins was studied in cell lysates (Fig. 1E). IRAG was phosphorylated heavily in the presence of cGKI␤ and very weak or not at all in the presence of cGKI␣. Furthermore, cGMP-agarose co-precipitated only IRAG and cGKI␤ but not IRAG and cGKI␣. The inability of cGKI␣ to interact with IRAG was caused by its inability to bind the IRAG protein, since regular Western blots showed that each protein was expressed to a similar level in the COS-7 cells. From these results we concluded that phosphorylation of IRAG requires interaction of the amino-terminal region containing the leucine zipper of cGKI␤ with IRAG.
Next we investigated the association of cGKI␤ with the InsP 3 RI, since the InsP 3 RI is phosphorylated by cGKI (24,25) and is present in the cGKI⅐IRAG macrocomplex. The InsP 3 RI was coexpressed with cGKI␤ in the absence of IRAG in COS-7 cells. cGMP-agarose did not co-precipitate InsP 3 RI with cGKI␤ (Fig. 1E). This result agreed with a two-hybrid screen using cGKI␤ as bait and different InsP 3 RI fragments as preys. In this screen no interaction between these proteins could be detected (data not shown). Therefore, these results indicate that cGKI and InsP 3 RI are not stably associated with each other. However, when IRAG was expressed together with cGKI␤ and InsP 3 RI all three proteins were co-precipitated (Fig. 1E). 90% or more of the solubilized coexpressed proteins were bound to the cGMP-agarose in the presence of cGKI␤. These results suggest that IRAG could mediate the assembly of these proteins in a macrocomplex.
cGKI␤ Phosphorylates IRAG Predominantly at Ser 696 -Several potential cGKI phosphorylation sites have been identified previously within the IRAG protein ( Fig. 1C) (9). In the present study, the functional role of the phosphorylation of these serine residues by cGKI was analyzed. In control experiments, both native IRAG and expressed IRAGa which is ϳ10 kDa larger than the native IRAG (see Ref. 9) were phosphorylated by cGKI␤ in the presence of 8-pCPT-cGMP in the cell lysates ( Fig.  2A). Similar to the expression level, the stoichiometry for cGKI␤ and IRAG was approximately equal in COS-7 cells. This phosphorylation reaction was specific for cGK, since cGMP-dependent phosphorylation of IRAGa was only observed when cGKI␤ was coexpressed with IRAG ( Fig. 2A). The COS-7 expression system was used next to study the phosphorylation efficiency of the single phosphorylation sites. For this purpose, the various serine phosphorylation sites (at positions 118, 629, 683, and 696) were mutated to alanine (S118A ϭ S1A; S629A ϭ S2A; S683A ϭ S3A; S696A ϭ S4A) by PCR. The single mutants and several multiple mutants were transiently expressed in COS-7 cells together with cGKI␤ and phosphorylation of these mutant proteins by cGKI␤ was analyzed compared with wild type IRAG (Fig. 2B). Mutation of Ser 696 to alanine (S4A) diminished significantly cGMP-dependent phosphorylation of IRAG to 53.1 Ϯ 5.9% (n ϭ 5) when compared with the phosphorylation of the wild type protein (Fig. 2B). In contrast, the phosphorylation efficiency of the triple mutant S123A was not affected being 126.1 Ϯ 12.9% (n ϭ 5). As expected, a time course of the phosphorylation indicated that the phosphorylation reaction was maximal between 1 and 2 min (Fig. 2C). The IRAG S4A mutant protein incorporated about half the amount of phosphate compared with the wild type or the IRAG S123A mutant protein supporting the notion that Ser 696 is a major cGKI␤ phosphorylation site in IRAG.
The reduced phosphorylation of IRAG S4A was neither caused by a decreased expression of the mutant IRAG protein nor by a reduced affinity of cGKI␤ for the mutant proteins (Fig.  3). The extent of co-precipitation of cGKI␤ and IRAG was analyzed using cGMP affinity chromatography followed by phosphorylation and immunoblot analysis. No difference could be observed in the amount of IRAG wild type or mutant proteins co-precipitated with cGKI␤, indicating that the mutation of the various serine residues did not affect the interaction of IRAG with cGKI␤. This result was expected since the interaction site of IRAG did not include any of the mutated serines. However, phosphorylation was significantly reduced in the S4A mutant protein (Fig. 3). These results suggest that Ser 696 is the predominant cGKI␤ phosphorylation site.
Phosphorylation of Ser 696 Mediates cGMP-dependent Inhibition of Calcium Release-Phosphorylation of IRAG by cGKI␤ inhibits bradykinin-and InsP 3 -induced calcium release from intracellular stores in COS-7 cells (9). The phosphorylation site(s) responsible for the inhibition were not identified. We anticipated that phosphorylation of the above mutated serines were involved in this functional effect of cGKI␤. Therefore, we tested whether or not cGMP-dependent inhibition of the calcium release was observed after transfection of the mutated IRAG constructs into COS-7 cells. Western blot analysis indicated that wild type IRAG and all mutated proteins were expressed to the same level in COS-7 cells. Fig. 4, A and B, show representative traces and the statistics, respectively, of bradykinin-induced calcium transients of COS-7 cells transfected with various IRAG mutants (S4A, S123A, and S1234A) and cGKI␤. cGKI␤ was activated by addition of 8-pCPT-cGMP before the second stimulation with bradykinin. In cells expressing IRAG mutants containing the S4A mutation (S4A and S1234A), cGKI␤ was unable to inhibit the calcium release. In contrast, active cGKI␤ decreased to a similar extent the second Ca 2ϩ peak in cells expressing IRAG containing the mutated serines S123A or wild type IRAG (Fig. 4).
The calculated area under the curve ratio of second to first bradykinin-induced calcium transient in COS-7 cells transfected with cGKI␤ and IRAG single or multiple mutants is illustrated in Fig. 4B. While S1A, S2A, and S3A single mutants showed similar values as IRAG wild type (about 50 -60% decrease after activation of cGKI␤), the reduced calcium release was abolished in mutant S4A. The same result was obtained when multiple-mutated IRAG variants were measured. The IRAG triple mutant containing simultaneously S123A mutations still mediated a decreased area under the curve ratio like wild type, whereas in multiple variants containing the S4A mutation (e.g. S34A and S1234A) no reduction of second to first calcium transient could be observed (Fig. 4B). It can be concluded from these results that phosphorylation of Ser 696 by cGKI␤ is essential for the inhibitory effect of IRAG on InsP 3induced calcium release. inhibition of platelet aggregation, cell motility and cell proliferation, secretion of intestinal fluid, bone growth, renin secretion, guidance of nerve fibers, and the development of synaptic plasticity (1)(2)(3)(4)26). Part of these functions are regulated by different cGK enzymes as shown by deletion of the gene for cGKII and cGKI (8,27). Only few proteins have been identified that are phosphorylated by cGKI or -II and are involved in the regulation of the proposed functions. These proteins include cGMP hydrolyzing phosphodiesterase 5 (28), the large subunit of the maxi-K Ca channel (29,30), CRP2 (31), telokin (32), VASP (33), CFTR (27,34), the cerebellar G-substrate (35,36) and IRAG (9). There are additional proteins such as the myosin binding subunit of phosphatase I (37,38) and RhoA (39) that may be in vivo substrates for cGKI and affect smooth muscle tone in the absence of elevated cytosolic Ca 2ϩ concentrations.
IRAG has been identified recently as the substrate for cGKI that mediates cGMP-dependent relaxation of smooth muscle by decreasing the calcium release from InsP 3 -sensitive stores (9). IRAG was co-purified together with InsP 3 R and cGKI␤. In this study, we show that IRAG interacts only with the amino-terminal region containing the leucine zipper of cGKI␤ and not with that of cGKI␣ or cGKII. The IRAG sequence interacting with the amino-terminal region containing the leucine zipper of cGKI␤ does not contain a phosphorylation site supporting the notion that interaction between cGKI␤ and IRAG is independent of the mechanism allowing substrate binding. This interaction does not require activation of the kinase domain and should be stable in relaxed and contracted cells. Localization of cGKI␤ to IRAG explains the observed phosphorylation specificity for the I␤ isozyme and the inability of cGKI␣ to modify IRAG.
The results of this study contribute to earlier observations that cGMP kinases recognize their in vivo substrates by interaction of amino-terminal kinase sequences with the substrate protein. This mechanism has been identified not only for the co-localization of cGKI␤ and IRAG, but was found also with cGKII and cGKI␣. Myristoylation of the first glycine of cGKII is required for membrane localization of the enzyme and the phosphorylation of CFTR (34,40). Furthermore, it has been shown that cGKI␣ binds with its unique amino-terminal leucine zipper to the myosin binding subunit of phosphatase I (38), the skeletal muscle troponin T (41), and the male germ cell-specific 42-kDa protein GKAP42 (42). In each case it appears that the kinase binds to a substrate sequence that is outside the phosphorylation site. The permanent localization of the kinase to its substrate allows a rapid response after increases in cGMP. In comparison to the cAMP kinase system it is obvious that the subcellular organization of the cGMP kinases does not require extra anchoring proteins as necessary for the cAMP kinase signaling system (43).
IRAG, cGKI␤, and the InsP 3 RI could be co-purified together, whereas cGKI␤ and InsP 3 RI were not co-precipitated and did not interact in the two-hybrid screen. It is therefore plausible to assume that the assembly of the triple complex required expression of IRAG. In support of this hypothesis is the finding that the InsP 3 RI phosphorylation was evident only when cGKI␤ was coexpressed with the InsP 3 RI. The differential possibility of cGKI to phosphorylate the InsP 3 RI is in good agreement with previous in vivo findings. cGKI phosphorylated in vivo the smooth muscle InsP 3 RI (24) and affected calcium release in megakaryocytes (17). This is not in contrast to the above results, since smooth muscle and platelets express high levels of cGKI␤ and IRAG which apparently allows cGKI␤-dependent phosphorylation of IRAG and the InsP 3 RI. Possibly, additional substrates exist for cGKI␤ in smooth muscle since ϳ50% of cGKI␤ immunreactivity has been found in the cytosolic fraction (44).
Phosphorylation of the InsP 3 RI by cGKI␤ apparently did not contribute to the decrease in calcium release. This effect depended on the phosphorylation of Ser 696 of IRAG. Mutation of Ser 696 to alanine abolished complete the modulatory effect of cGKI␤ on calcium release without disruption of the triple complex. The phosphorylation of the InsP 3 RI by cAMP kinase may be important to modulate the calcium release in vivo (45) and may be an independent possibility to regulate smooth muscle contractility. Additional experiments are necessary to clarify whether or not cAMP kinase modulates smooth muscle contraction via phosphorylation of the InsP 3 R and requires the presence of IRAG. It is quite possible that modulation of the InsP 3 -dependent calcium release is specifically caused by cGKI␤, since cAMP analogs were unable to modulate the calcium release in murine aortic wild type and cGKI negative smooth muscle cells (8).
Together with the functional data obtained by intracellular calcium measurements our results clearly demonstrate that phosphorylation of Ser 696 is indispensable for the inhibitory effect of IRAG and cGKI on InsP 3 -induced calcium release. The data presented here add a further functional element how the NO/cGMP/cGKI signaling pathway is involved in calcium regulation. With the identification of the functional phosphorylation site of the cGKI␤ substrate protein IRAG we threw more light on the molecular mechanism by which cGKI inhibits InsP 3 -dependent calcium release in smooth muscle. Subsequent work must aim to elucidate the role of InsP 3 R in this process more precisely.