Guanylyl cyclase/PSD-95 interaction: targeting of the nitric oxide-sensitive alpha2beta1 guanylyl cyclase to synaptic membranes.

The signaling molecule nitric oxide (NO) exerts most of its effects by the stimulation of the NO-sensitive guanylyl cyclase. Two isoforms of the NO receptor molecule exist: the ubiquitously occurring α1β1 and the α2β1 with a more limited distribution. As the isoforms are functionally indistinguishable, the physiological relevance of these isoforms remained unclear. The neuronal NO synthase has been reported to be associated with PSD-95. Here, we demonstrate the interaction of the so far unnoticed α2β1 isoform with PSD-95 in rat brain as shown by coprecipitation. The interaction is mediated by the α2 C-terminal peptide and the third PDZ domain of PSD-95. As a consequence of the PSD-95 interaction, the so far considered “soluble” α2β1 isoform is recruited to the membrane fraction of synaptosomes, whereas the α1β1 isoform is found in the cytosol. Our results establish the α1β1 as the cytosolic and the α2β1 as the membrane-associated NO-sensitive guanylyl cyclase and suggest the α2β1 isoform as the sensor for the NO formed by the PSD-95-associated neuronal NO synthase.

The signaling molecule nitric oxide (NO) exerts most of its effects by the stimulation of the NO-sensitive guanylyl cyclase. Two isoforms of the NO receptor molecule exist: the ubiquitously occurring ␣ 1 ␤ 1 and the ␣ 2 ␤ 1 with a more limited distribution. As the isoforms are functionally indistinguishable, the physiological relevance of these isoforms remained unclear. The neuronal NO synthase has been reported to be associated with PSD-95. Here, we demonstrate the interaction of the so far unnoticed ␣ 2 ␤ 1 isoform with PSD-95 in rat brain as shown by coprecipitation. The interaction is mediated by the ␣ 2 C-terminal peptide and the third PDZ domain of PSD-95. As a consequence of the PSD-95 interaction, the so far considered "soluble" ␣ 2 ␤ 1 isoform is recruited to the membrane fraction of synaptosomes, whereas the ␣ 1 ␤ 1 isoform is found in the cytosol. Our results establish the ␣ 1 ␤ 1 as the cytosolic and the ␣ 2 ␤ 1 as the membrane-associated NO-sensitive guanylyl cyclase and suggest the ␣ 2 ␤ 1 isoform as the sensor for the NO formed by the PSD-95-associated neuronal NO synthase.
NO-sensitive guanylyl cyclase (GC), 1 the enzyme that catalyzes the conversion of GTP to cGMP, acts as the effector molecule for nitric oxide, thereby playing a crucial role within the NO/cGMP signaling cascade (1,2). By binding to the prosthetic heme group of GC, NO leads to the up to 200-fold activation of the enzyme (3,4). The subsequent increase in cGMP formation results in the activation of cGMP effector molecules (cGMP-activated protein kinases, cGMP-regulated phosphodiesterases, cGMP-gated ion channels), which mediate the NOinduced effects. Besides its role in the cardiovascular system, NO acts as a freely diffusible transmitter in numerous pathways in the central and peripheral nervous systems (2,5) In addition to NO, carbon monoxide (CO) has been proposed as a neuronal messenger that activates GC (5, 6) particularly as CO stimulates GC effectively in the presence of the new GC modulator YC-1 (7).
The NO-sensitive GC is a heterodimeric enzyme consisting of an ␣ and a ␤ subunit (8,9). The ␣ 1 ␤ 1 heterodimer corresponds to the enzyme purified from lung; homology screening yielded the ␣ 2 and ␤ 2 subunits (10,11). In contrast to the ␤ 2 subunit that did not form a catalytically active enzyme with any other subunit in most experimental settings, the ␣ 2 subunit yielded a NO-sensitive cGMP-forming enzyme upon coexpression with the ␤ 1 subunit (11). On the protein level, the ␣ 2 subunit was shown to occur in placenta, and the ␤ 1 subunit was identified as the dimerizing partner (12). Northern blot analysis showed comparatively high amounts of ␣ 2 message in placenta, uterus, and brain (13). Recently, the wide distribution of ␣ 2 mRNA in rat brain with a particular abundance in cerebellum and hippocampus was demonstrated using in situ hybridization (14).
All GC subunits contain the so-called cyclase catalytic domain also conserved in the membrane-bound GC and in the adenylyl cyclase within the C-terminal regions; the N termini of the ␣ 1 and the ␤ 1 subunits have been shown to be required for proper coordination of the prosthetic heme group (15,16). With only 30% of identical amino acids, the N-terminal region of the ␣ 2 subunit differs considerably from the corresponding part of the ␣ 1 subunit. Although the catalytic domains are very much conserved, the very C-terminal amino acids are unrelated. Despite these differences in primary structure, extensive analysis of the purified isoforms of GC (␣ 1 ␤ 1 and ␣ 2 ␤ 1 ) did not reveal any regulatory or kinetic differences (12). Therefore, the physiological relevance of two similarly operating isoforms of GC remained unclear.
It has been suggested that the C terminus of the ␣ 2 subunit contains a sequence that allows the interaction with the PDZ domain of ␣1-synthrophin (17). PDZ domains, their name being derived from the first proteins in which they were discovered (PSD-95, Dlg, ZO-1) are protein modules of about 100 amino acids that apparently mainly occur in proteins of the cytoskeleton. They are supposed to function as molecular adaptors to form functional complexes by interacting with the C termini of their binding partners in most cases (for reviews see Refs. 18 -20). A possible GC-synthrophin interaction appears plausible as the neuronal NO synthase has been reported to interact with ␣ 1 -synthrophin (21). In addition, the neuronal NO synthase has been shown to interact with the PDZ domain-containing PSD-95 (postsynaptic-density-95) (21) that occurs in high concentrations in postsynaptic densities in the central nervous system (22). PSD-95 was the first member of a family of synaptic proteins containing typical protein-interaction motives: three N-terminal PDZ domains, one SH3 domain, and a C-terminal guanylate kinases homology region. Three other members of this family are known up to this date: SAP97 (23), SAP102 (24), and Chapsyn-110/PSD-93 (25,26).
We considered the PDZ domain-containing protein PSD-95 as a tempting candidate for a ␣ 2 ␤ 1 GC interaction. In the current report, we present multiple lines of evidence for an interaction of the ␣ 2 ␤ 1 isoform of GC with PSD-95. We show that the C-terminal peptide of the ␣ 2 subunit interacts with the PDZ domains of PSD-95 and related proteins (PSD-93, SAP97, SAP 102) in brain homogenates; vice versa, the PDZ domains of PSD-95 are shown to exclusively bind the ␣ 2 ␤ 1 and not the ␣ 1 ␤ 1 GC isoform both present in brain. The in vivo existence of ␣ 2 ␤ 1 -PSD-95 complexes is shown in coprecipitation experiments. In contrast to the ␣ 1 ␤ 1 isoform, the ␣ 2 ␤ 1 heterodimer is found in the membrane fraction of synaptosomes demonstrating that the PSD-95 interaction results in the membrane association of this so far soluble-considered GC isoform.

MATERIALS AND METHODS
Antibodies against GC-Antibodies directed against the C-terminal peptides of the ␣ 1 subunit (KKDVEEANANFLGKASGID), ␣ 2 subunit (KKVSYNIGTMFLRETSL), and ␤ 1 subunit (SRKNTGTEETEQDEN) were obtained as described in Ref. 27. Immunoaffinity purification of the antibodies was performed according to standard protocols using the respective cysteine-tagged antigenic peptides coupled to Sulfolink Coupling Gel (Pierce).
Determination of cGMP Forming Activity-GC activity was determined by incubation of the samples (containing 5-15 g of protein) for 10 min at 37°C with 1 mM cGMP, 250 M [␣-32 P]GTP (about 0.2 Ci), 3 mM MgCl 2 , 3 mM DL-dithiothreitol, 0.5 g/liter bovine serum albumin, 0.25 g/liter creatine phosphokinase, 5 mM creatine phosphate, 1 mM 3-isobutyl-1-methylxanthine, and 50 mM triethanolamine hydrochloride, pH 7.4, in a total volume of 100 l. The incubation was started by addition of diethylamine-NO (final concentration 100 M) and by transferring the tubes from ice to 37°C. Reactions were stopped by ZnCO 3 precipitation, and isolation of the enzymatically formed cGMP was performed as described by Schultz and Böhme (28).
Cloning and Expression of PDZ-GST Fusion Proteins-Rat cDNAs encoding for PDZ 1 (amino acids 62-156), PDZ 2 (amino acids 157-256), and PDZ 3 (amino acids 310 -404) of PSD-95 were amplified by polymerase chain reaction and cloned into pGEX2T vector (Amersham Pharmacia Biotech). The identity of the clones was verified by sequencing and expression was performed in DH5␣ cells (Life technologies). After induction of expression with 0.4 mM isopropyl-␤-D-thiogalactopyranoside for at least 6 h, cells were collected by centrifugation at 7700 ϫ g for 10 min and lyzed in phosphate-buffered saline by sonication. The lysate (1-3 mg/ml) was cleared by centrifugation at 50,000 ϫ g for 15 min.
Preparation of Brain Homogenates for GC Subunit Detection-Rat brains were homogenized in 5 volumes of phosphate-buffered saline containing protease inhibitors (0.5 mM phenylmethylsulfonyl fluoride, 0.2 mM benzamidine, 1 M pepstatin A). Subsequently, SDS was added to a final concentration of 1% to solubilize the membranes, and the homogenate was incubated (10 min, 4°C) and cleared by centrifugation (50,000 ϫ g, 20 min, 4°C). Protein concentrations were determined by the Bradford protein assay (Bio-Rad) following trichloroacetic acid precipitation. Proteins (40 g) were analyzed in Western blots.
Precipitations with the Immobilized ␣ 2 C-terminal Peptide-The cysteine-tagged ␣ 2 C-terminal peptide (C-FLRETSL-COOH) and, as control, a scrambled peptide (C-SLFTRLE-COOH) were coupled to Sulfolink Coupling Gel (Pierce) according to instructions of the manufacturer.
Rat brains were homogenized in 5 volumes of 0.5 M NaCl, 1 mM EDTA, 1% (v/v) Nonidet P-40 (Calbiochem), and 25 mM Tris/HCl, pH 7.4, and centrifuged for 30 min at 15,000 ϫ g. The resulting supernatant (1.8 ml containing ϳ18 mg of protein) was incubated with the respective immobilized peptides (50 l of resin) on an overhead rotator for 2 h at 4°C. After extensive washing, interacting proteins were eluted with Laemmli buffer, and aliquots (5%) of the samples were analyzed by Western blotting.
Cleared lysates (1.8 ml) of Escherichia coli expressing the three PDZ domains of PSD-95 as GST fusion proteins prepared as described above were incubated for 30 min with the immobilized peptides (25 l of resin). After washing, sample buffer was added, and aliquots (5%) of the samples were analyzed in Coomassie Brilliant Blue-stained gels. The identity of the fusion proteins was confirmed in Western blots using anti-GST antibodies (Sigma).
Preparation of Synaptosomal Cytosol and Membranes-Crude syn-aptosomes were prepared by combination of the methods described by Cohen et al. (29) and Mayer et al. (30). Rat brains were homogenized in 4 volumes of buffer A (0.32 M sucrose, 1 mM EDTA, and 10 mM HEPES, pH 7.4 containing 0.5 mM phenylmethylsulfonyl fluoride, 0.2 mM benzamidine, 1 M pepstatin A, and 2 mM DL-dithiothreitol) by 15 strokes with a glass-Teflon Potter-Elvehjem homogenizer. The homogenate was diluted 1:10 in buffer A, and a low speed pellet was removed by centrifugation (750 ϫ g, 10 min, 4°C). The resulting supernatant was recentrifuged (17,300 ϫ g, 10 min, 4°C) to pellet the crude synaptosomal fraction. The crude synaptosomes were lyzed by hypoosmotic shock and ultrasound in 1 mM EDTA and 50 mM triethanolamine hydrochloride, pH 7.4, containing protease inhibitors and DL-dithiothreitol as in buffer A. The lysate was kept 10 min at 4°C, and subsequently the NaCl concentration was increased to 150 mM to remove loosely associated proteins of the membrane. The lysate was then recentrifuged (32,000 ϫ g, 15 min, 4°C). The resulting synaptosomal cytosol and membranes were analyzed in Western blots, and cGMP-forming activity in the fractions was determined as described above. Protein concentrations were determined by the Bradford protein assay (Bio-Rad) after trichloroacetic acid precipitation of the proteins. Immunoprecipitation of PSD-95-Crude synaptosomes were prepared as described above and resuspended in buffer A. Triton X-100 was added to a final concentration of 1% to extract a fraction of PSD-95 and associated proteins. After a 10-min incubation at 4°C, the Triton X-100 extract was centrifuged (32,000 ϫ g, 10 min, 4°C) to remove the cell debris. PSD-95 immunoprecipitation was carried out as follows: the synaptosomal extract (4.2 ml containing ϳ10 mg of protein) was incubated for 4 h at 4°C with monoclonal anti PSD-95-antibody (15 l, Clone 7E3-1B8, Affinity Bioreagents) and 250 l of Dynabeads Panmouse IgG (Dynal); as control, the incubation was performed without the monoclonal anti PSD-95 antibody. Dynabeads were precipitated using a magnet, and the beads were extensively washed with buffer A. Precipitated proteins were eluted with 65 l of SDS-containing sample buffer, and 20 l of the samples were analyzed in Western blots.
Precipitation of GC with GST-PDZ Fusion Proteins-The three PDZ-GST fusion proteins prepared as described above were loaded onto glutathione-Sepharose CL4B (Amersham Pharmacia Biotech) as suggested by the manufacturer. The amounts of loaded GST fusion proteins were checked by Bradford protein assay (Bio-Rad). Rat brains were homogenized in 5 volumes of a buffer containing 150 mM NaCl, 1 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, 0.2 mM benzamidine, 1 M pepstatin A, 2 mM DL-dithiothreitol, and 50 mM triethanolamine hydrochloride, pH 7.4, and centrifuged for 1 h at 220,000 ϫ g to obtain a cytosolic fraction. Cytosolic proteins (10 mg of protein in 2.5 ml) were incubated with each of the three immobilized PDZ-GST fusion proteins (200 l of Sepharose with ϳ1.2 mg of GST fusion protein) for 30 min at 4°C. After extensive washing, 2.5 l of the Sepharose were used to detect bound GC by measuring NO-stimulated cGMP-forming activity as described above. For Western blots, PDZ interacting proteins were eluted with SDS-containing sample buffer and analyzed in Western blots (2.5 l of resin/lane).

RESULTS
Occurrence of the ␣ 2 Subunit in Rat Brain-To date, the existence of the ␣ 2 subunit of GC on protein level has been demonstrated only in human placenta (12). As the occurrence of the ␣ 2 subunit in brain was a prerequisite for a possible PSD-95/␣ 2 interaction, we studied the expression of this subunit in rat brain using immunoaffinity-purified antibodies generated against the C-terminal peptide of the ␣ 2 subunit. As can be seen on the Western blot (Fig. 1), a protein with the appropriate molecular mass was detected in brain homogenates showing the expression of the ␣ 2 subunit in rat brain. In addition, the presence of the ␣ 1 and ␤ 1 subunits was demonstrated with the respective affinity-purified antibodies. The antibody against the ␣ 1 subunit detected an additional high molecular mass band; however the nature of this band remains to be clarified.
Interaction of the ␣ 2 Peptide with PSD-95-The existence of the ␣ 2 subunit in brain encouraged us to investigate whether the C-terminal peptide of the ␣ 2 subunit was capable of interacting with the PDZ domain-containing protein PSD-95 and related proteins. For that reason, we incubated the immobilized C-terminal ␣ 2 peptide with brain extracts and subse-quently separated interacting proteins on SDS gels. Western blots performed with antibodies against members of the PSD-95 family revealed that PSD-95 and the related proteins PSD-93, SAP97, and SAP102 bound to the ␣ 2 peptide (Fig. 2). The reaction was specific as binding of the synaptic proteins to the immobilized ␣ 2 peptide was abolished in the presence of free ␣ 2 peptide (0.05 mg/ml, not shown) and a scrambled ␣ 2 peptide did not bind any of these proteins (see Fig. 2). Furthermore, these results indicate that the peptide corresponding to the C-terminal sequence of the ␣ 2 subunit was able to interact with members of PSD-95 family.
Subcellular Distribution of GC Subunits-In vivo interaction of ␣ 2 ␤ 1 GC and PSD-95 should lead to the colocalization of the ␣ 2 ␤ 1 isoform and PSD-95 in synaptic membranes. As PSD-95 has been shown to be enriched in membranes of synaptosomal preparations (22), we proceeded to study the subcellular distribution of GC subunits in this material. All GC subunits were detected in crude synaptosomes (not shown); however, after disruption of the synaptosomes, the ␣ 1 subunit was almost exclusively found in the synaptosomal cytosol, whereas the majority of the ␣ 2 subunit was detected in the membrane fraction (Fig. 3A). The ␤ 1 subunit was distributed almost equally between membranes and cytosol. This finding is in accordance with our former observation that the ␤ 1 subunit represents the dimerizing partner for both the ␣ 2 and ␣ 1 subunits. The immunological evidence of a cytosolic and a membrane-associated NO-sensitive GC was further supported by determination of NO-stimulated cGMP formation in the respective fractions (Fig. 3B). Enzyme activities in the presence of NO amounted to ϳ4 and ϳ1 nmol ϫ mg Ϫ1 ϫ min Ϫ1 in cytosolic and membrane fractions, respectively. Non-stimulated cGMP-forming activities were below the detection limit. These results show that the membraneassociated isoform of GC was responsible for a substantial amount of the NO-stimulated cGMP-forming activity in synaptosomes and established the ␣ 1 ␤ 1 GC as the cytosolic and the ␣ 2 ␤ 1 GC as the membrane-associated isoform of GC. Moreover, the results demonstrate the presence of the ␣ 2 ␤ 1 heterodimer in the PSD-95-containing compartment.
␣ 2 ␤ 1 GC Coprecipitates with PSD-95-To study the existence of in vivo formed ␣ 2 ␤ 1 /PSD-95 complexes in rat brain, we performed immunoprecipitation experiments. To be able to precipitate PSD-95, synaptosomes had to be extracted with Triton X-100 to solubilize at least a portion of the otherwise membrane-associated PSD-95. The Triton extracts were incubated with monoclonal antibodies against PSD-95, and precipitated proteins were analyzed in Western blots using immunoaffinitypurified antibodies against each of the GC subunits. In the PSD-95 precipitate, the ␣ 2 antibodies recognized a protein with the appropriate molecular mass clearly showing that the ␣ 2 subunit is associated with PSD-95 (Fig. 4). The dimerizing partner of the ␣ 2 subunit, the ␤ 1 subunit, was found in the PSD-95 precipitate, too. In contrast, the ␣ 1 subunit was not detected in the PSD-95 precipitate, indicating that only the ␣ 2 subunit but not the ␣ 1 subunit is able to interact with PDZ domains.
␣ 2 ␤ 1 GC Preferentially Interacts with the Third PDZ Domain of PSD-95-To study the molecular basis for the ␣ 2 ␤ 1 /PSD-95 interaction, we generated GST fusion proteins of each of the PDZ domains of PSD-95 and tried to precipitate these fusion proteins with the immobilized ␣ 2 C-terminal peptide. As can be seen on the Coomassie Brilliant Blue-stained gel (Fig. 5), all three PDZ domains were precipitated by the ␣ 2 peptide showing that the PSD-95/␣ 2 subunit interaction was mediated by the PDZ domains of PSD-95 and the C-terminal peptide of the ␣ 2 subunit. However, in this experimental setting no binding preference of the ␣ 2 peptide to one of the three PDZ domains was detectable. We attributed this lack of differential binding to the fact that only the minimal parts of both binding partners required for interaction were present in the experiment. Therefore, we used the immobilized PDZ-GST fusion proteins to precipitate the complete ␣ 2 ␤ 1 enzyme of brain cytosol and quantified the amount of PDZ domain-bound GC by measuring GC activity. Here, the third PDZ domain of PSD-95 precipitated substantially more NO-stimulated cyclic GMP-forming activity than the first and second PDZ domains (Fig. 6A) indicating a higher affinity of the complete ␣ 2 ␤ 1 enzyme to the third PDZ domain. To ensure that PDZ domain-bound cGMPforming activity was solely due to the ␣ 2 ␤ 1 isoform, we performed Western blot analysis of the PDZ precipitates. As shown in Fig. 6B, only the ␣ 2 and ␤ 1 subunits but not the ␣ 1 FIG. 1. Expression of GC subunits in rat brain. Rat brain homogenate (40 g per lane) was analyzed in Western blots using immunoaffinity purified antibodies against the three GC subunits. The most prominent signals with molecular masses of ϳ83 kDa, ϳ82 kDa, and ϳ66 kDa correspond to the ␣ 1 , ␣ 2 , and ␤ 1 subunits, respectively.

FIG. 2. Affinity precipitation of PSD-95 and PSD-95 related
proteins by the immobilized ␣ 2 C-terminal peptide. ␣ 2 C-terminal peptide or with a scrambled peptide as control. After washing, interacting proteins were separated on SDS-gels. The Western blots performed with antibodies against PSD-95, PSD-93, SAP102, and SAP97 demonstrated the specific precipitation of these proteins by the immobilized ␣ 2 C-terminal peptide. In the input, 6% of the brain extracts used for the precipitation experiments were applied. subunit were detected, again demonstrating that the ␣ 2 ␤ 1 but not the ␣ 1 ␤ 1 isoform interacted with PSD-95. DISCUSSION Nitric oxide has been implicated to be involved in synaptic transmission since the late 1980's, and a role in neuronal plasticity has been proposed. Although other targets for NO might exist, the NO-sensitive GC is the best studied NO receptor. However, little information about NO-sensitive GC in brain is available.
In the current report, we studied the so far unnoticed guanylyl cyclase isoform ␣ 2 ␤ 1 and present evidence that the special quality of this isoform is defined by the ␣ 2 C-terminal peptide which is able to mediate the interaction with PDZ domains. In particular, we studied the ␣ 2 ␤ 1 /PSD-95 interaction, thereby demonstrating for the first time the existence of the ␣ 2 ␤ 1 heterodimer in brain and addressing the so far unattended aspect in the GC research, i.e. the membrane association of the until now considered cytosolic enzyme.
PDZ domains are present in a plethora of proteins and have been reported to act as molecular adaptors to direct the subcellular distribution of proteins thereby mediating the formation of functional complexes. In brain, PSD-95 and related proteins have been proposed to play important roles e.g. by clustering NMDA receptors at excitatory synapses and by anchoring the receptor to downstream effector proteins like the neuronal NO synthase. Mice deficient in PSD-95 showed unaffected synaptic morphology and NMDA-receptor function but featured impaired learning and enhanced long-term potentiation (31). These results emphasize the importance of the formation of "transducisomes" downstream the NMDA receptor for synaptic plasticity and learning.
In most cases, PDZ domains bind to the C-terminal sequence of their binding partners, although in the case of the NO synthase an internal stretch of amino acids has been reported to account for the interaction with PSD-95 (32). A C-terminal consensus sequence (S/T)X(V/I) for proteins interacting with the PDZ domains of PSD-95 has been deduced from the known binding partners of PSD-95 (18). Despite the fact that the ␣ 2 peptide (RETSL) contains a leucine residue at the C-terminal position, we were able to show that the C-terminal ␣ 2 peptide does bind to the PDZ domains of PSD-95. The relevance of this finding is corroborated by the demonstration that the recombinantly expressed PDZ domains are able to precipitate the ␣ 2 ␤ 1 holoenzyme of brain extracts. As binding of the ␣ 2 ␤ 1 GC isoform to the third PDZ domain was most pronounced, this PDZ domain very likely represents the structure responsible for binding of the ␣ 2 C-terminal peptide in vivo.
The physiological occurrence of the ␣ 2 /PSD-95 interaction observed ex vivo was confirmed in coimmunoprecipitation experiments with PSD-95. Moreover, not only the ␣ 2 but also the ␤ 1 subunit was found in PSD-95 precipitates. These results clearly show that the ␣ 2 ␤ 1 isoform is associated with PSD-95 and support our former finding that the ␤ 1 subunit represents the physiological dimerizing partner of the ␣ 2 subunit.
Our data indicate targeting of a central component of the NO/cGMP cascade by PSD-95. PSD-95 as a molecular adaptor contains three PDZ domains; binding of NMDA-receptor has been shown to occur via the first and second PDZ domains (33,34) and the neuronal NO synthase has been reported to interact with the second PDZ motif (21). Moreover, the neuronal NO synthase has been shown to be selectively stimulated by Ca 2ϩ entering the cell through NMDA receptors while Ca 2ϩ provided by other channels is not as effective (35)(36)(37). The interaction of the ␣ 2 ␤ 1 GC with the PSD-95 third PDZ domain would lead to the formation of a signaling microdomain composed of Ca 2ϩpermeable NMDA receptors, Ca 2ϩ -dependent neuronal NO synthase, and the NO effector ␣ 2 ␤ 1 GC within the postsynaptic density. The proximity of the NO-generating and the NO-sensing enzyme suggests that NO concentrations differ within a cell, and a close contact between the neuronal NO synthase and the ␣ 2 ␤ 1 GC is required to transduce NO signals in certain cellular compartments without rising the overall NO concentration.
On the other hand, PSD-95 may not be the only interaction partner of ␣ 2 ␤ 1 GC since the C-terminal peptide of the ␣ 2 subunit precipitates other members of the PSD-95 family of proteins. Presynaptic targeting of ␣ 2 ␤ 1 GC by PDZ-containing proteins such as SAP97 also appears reasonable. In this case, NO produced in the postsynaptic nerve terminal could act as a retrograde messenger for presynaptically located GC. The current lack of suitable antibodies for immunohistochemistry addresses these questions to future studies. However, our results provide the necessary requirements for a synaptic localization of the ␣ 2 ␤ 1 isoform of the NO-sensitive GC, thereby enabling the enzyme to act as the mediator of the NO effects in synaptic transmission.
By the interaction with PSD-95, the ␣ 2 ␤ 1 heterodimer is recruited to the membrane as shown by Western blot analysis and determination of GC activity in cytosolic and membrane fractions of synaptosomes. Thus, in the current report, we show for the first time the association of a substantial portion of the NO-stimulated cGMP-forming activity to the membrane, which is not due to nonspecific binding of the soluble ␣ 1 ␤ 1 isoform but to the specific targeting of the ␣ 2 ␤ 1 GC by a PDZ domaincontaining protein. Further experiments have to show whether this membrane association of the ␣ 2 ␤ 1 isoform of GC via PDZ domains represents a general biological principle also found in other tissues. Up to this date, GCs have been divided in membrane-bound and soluble enzymes. This categorization may be the reason why this membrane-associated NO-sensitive GC has been overlooked for the past 20 years. The classification in membrane-bound and soluble guanylyl cyclase enzymes is incompatible with the physiological existence of receptor-linked membrane-bound GCs, a membrane-associated NO-sensitive GC (␣ 2 ␤ 1 ), and a soluble NO-sensitive GC (␣ 1 ␤ 1 ). Rather, we suggest the classification of the enzyme family according to the regulatory properties of their members, i.e. receptor-linked GCs and NO-sensitive GCs.
FIG. 6. ␣ 2 ␤ 1 GC preferentially binds to the third PDZ domain of PSD-95. Brain cytosol was incubated with GST fusion proteins of the three PDZ domains of PSD-95 (PDZ 1, PDZ 2, PDZ 3) bound to GSH-Sepharose. As control, an unrelated GST fusion protein was used. Shown are the NO-stimulated cGMP formation (A) and the Western blots of the GC subunits (B) in the respective precipitate. The input lane represents 50% of the cytosolic proteins used for precipitation.