The Amino-Terminus of Nitric Oxide Sensitive Guanylyl Cyclase α1 Does Not Affect Dimerization but Influences Subcellular Localization

Background Nitric oxide sensitive guanylyl cyclase (NOsGC) is a heterodimeric enzyme formed by an α- and a β1-subunit. A splice variant (C-α1) of the α1-subunit, lacking at least the first 236 amino acids has been described by Sharina et al. 2008 and has been shown to be expressed in differentiating human embryonic cells. Wagner et al. 2005 have shown that the amino acids 61–128 of the α1-subunit are mandatory for quantitative heterodimerization implying that the C-α1-splice variant should lose its capacity to dimerize quantitatively. Methodology/Principal Findings In the current study we demonstrate preserved quantitative dimerization of the C-α1-splice by co-purification with the β1-subunit. In addition we used fluorescence resonance energy transfer (FRET) based on fluorescence lifetime imaging (FLIM) using fusion proteins of the β1-subunit and the α1-subunit or the C-α1 variant with ECFP or EYFP. Analysis of the respective combinations in HEK-293 cells showed that the fluorescence lifetime was significantly shorter (≈0.3 ns) for α1/β1 and C-α1/β1 than the negative control. In addition we show that lack of the amino-terminus in the α1 splice variant directs it to a more oxidized subcellular compartment. Conclusions/Significance We conclude that the amino-terminus of the α1-subunit is dispensable for dimerization in-vivo and ex-vivo, but influences the subcellular trafficking.


Introduction
Nitric oxide sensitive guanylyl cyclase is the physiological receptor for nitric oxide (NO) and nitric oxide releasing drugs. Its second messenger cyclic GMP is crucial for vasodilatation, penile erection, platelet disaggregation and neurotransmission. The heterodimeric enzyme is formed by either an a 1 -or an a 2 -subunit and a b 1 -subunit. Dimerization of the enzyme is a prerequisite for its catalytic activity, because both the aas well as the b 1 -subunit provide essential residues for the conversion of GTP to cGMP [1]. There is conflicting evidence which parts of the subunits are mandatory for quantitative heterodimerization. Previous studies [2,3] show inconsistent results with respect to the requirement of the a 1 subunit amino-terminus which is missing in the naturally occurring C-a 1 splice variant [4]. Since the occurrence of this splice variant has been linked to differentiation of human embryonic stem cells [5], the capacity of the C-a 1 splice variant to heterodimerize and thus form a functionally active enzyme, is of biological importance.
Because of the controversial nature of the question, we investigated the dimerization capacity using two independent experimental approaches. First we used a purification method of the NOsGC b 1 -subunit and looked for co-purification of the a 1 variants. With the second experimental approach we examined fluorescence resonance energy transfer (FRET) based on fluorescence lifetime imaging (FLIM) in intact cells using fusion proteins of the respective subunits with fluorescent proteins. While analyzing the fluorescence lifetimes of the respective NOsGCsubunits, we serendipitously discovered that the C-a 1 splice isoform, shows a unique subcellular distribution. Using a fluorescent tagged marker for the endoplasmic reticulum (ER), we conclude, that the C-a 1 isoform is located at the ER. As it is well known, that the redox state of the ER is relatively more oxidized than the cytosol [6], we performed a ratiometric analysis of the redox properties of the C-a 1 protein by using a redox sensor (Grx1-roGFP2) and identified, that the protein is not only distributed in a different manner than the wild type, but also located in a more oxidized environment. Co-expression of the b 1 -subunit restored the cytosolic localization of the C-a 1 splice isoform, but led to a nuclear localization that was not found for the canonical a 1 /b 1 heterodimer. We could show that the C-a 1 splice isoform retains its ability to heterodimerize quantitatively with the b 1 -subunit ex vivo and in intact cells, despite lack of aminoterminal amino acids that were thought to be important for heterodimerization [2]. In addition, we observed that the fluorescent fusion of the C-a 1 splice subunit is directed to a more oxidized subcellular compartment, while the respective C-a 1 /b 1 heterodimer shows a cytosolic and nuclear localization.
Results a 1 DN 259 will be expressed by the C-a 1 splice form Analyzing the C-a 1 splice-variant described by Sharina et al. [4] and designated a 1 DN 240 shows that the first initiation codon after splicing would either form an a 1 DN 236 or an a 1 DN 259 variant because these represent the only methionines with an open reading frame in the human sequence ( Fig. 1 and Data S2). In a recently published review by Sharina et al. 2011, the authors have adapted our numbering [7]. In a previous study by Koglin and Behrends 2003 [3], we investigated intensely the a 1 DN 236 truncation in comparison to the a 1 DN 259 truncation. Because neither enzyme activity, substrate-dependency (GTP), nor dose-effect-curve of nitric oxide differed significantly from a 1 DN 259 , it was suggested by the reviewers to remove the redundant data for the a 1 DN 236 truncation. Since there was no difference in molecular weight between the a 1 DN 236 and a 1 DN 259 variant (Fig. 2), we now assume that the recognition of the translation initiation site of ATG 259 is dominant over ATG 236 in the human sequence. We thus suggest that the C-a 1 splice variant leads to the formation of a subunit with an a 1 DN 259 deletion.

Establishment of a novel one-step-purification protocol
We performed a new one-step-purification of the soluble guanylyl cyclase using the Strep Tag II, which results in 1 mg purified NOsGC from 1,000 ml Sf-9 culture medium. The yield calculated by using the total activities of cytosol and purified protein is <25 %. The purification factor is <125-fold and results in a more than 95 % purity of the enzyme determined by Coomassie stained gels.
Analysis of the heterodimerization of C-a 1 by co-purification with b 1 -S Using a carboxy-terminally Strep tagged b 1 -subunit of NOsGC, we could show that both a 1 -isoforms dimerize with the b 1 -subunit ( Fig. 3). Densitometric analysis and graphical representation as described in [2] shows that C-a 1 forms heterodimers not less but more effectively with a relative value of 166.2 % (612.2 %) compared with a 1 WT (Fig. 4). Alternatively the ratio of the band intensities of the a 1 WT variants to the b 1 -bands were 0.7560.11 for C-a 1 /b 1 and 0.4360.06 for a 1 /b 1 . The excess of the b 1 subunit is due to the fact that purification was performed with a Strep Tag II attached to the b 1 -subunit as described in [2]. Because our findings contradict the central message of the paper by Wagner at al. 2005 although the experimental approach was very similar [2], we tested more subtle differences. We expressed Sf-9 cells without any supplement, only with hemin (4 mg/l) and with both hemin and lipid supplement as in [2], the latter being used to reduce the shear forces during Sf-9 cell culture. After expression under these conditions, we purified the enzyme complex and performed a SDS-PAGE and the gel was stained with Coomassie Blue according to the protocol by Kang et al. [8]. There was no negative effect of either of the supplements on the heterodimerization of C-a 1 /b 1 S (Fig. 5) confirming our finding that the C-a 1 splice-variant retains its ability to heterodimerize under different experimental conditions.
Guanylyl cyclase activity of the purified C-a 1 /b 1 and a 1 /b 1 heterodimers was measured under basal conditions, in the presence of NO (100 mM DEA/NO) or in the presence of 10 mM cinaciguat. Consistent with the formation of more functional heterodimers for C-a 1 /b 1 versus a 1 /b 1 in Sf-9 cell culture, enzyme activity was higher for the splice variant under all experimental conditions (Fig. 6). Spectroscopic analysis shows that both C-a 1 /b 1 than a 1 /b 1 contain a significant amount of heme ( Fig. 7) which is consistent with their responsiveness to nitric oxide (see Fig. 6).

Analysis of the heterodimerization of C-a 1 by FLIM-FRET
In order to test whether the unexpected formation of more functional heterodimers with the b 1 subunit for C-a 1 versus the canonical a 1 subunit is also seen in intact cells we used fluorescence resonance energy transfer (FRET) based on fluorescence lifetime imaging. The a 1 -subunit and C-a 1 -subunit were fused with ECFP and the b 1 -subunit with EYFP in analogy to a previous FRET study by our group [9]. Analysis of the respective combinations in HEK-293 cells showed that the fluorescence lifetime was significantly shorter for a 1 /b 1 and C-a 1 /b 1 than the negative control (Fig. 8A). This indicates that heterodimerization of both a 1 variants with the b 1 subunit bring the respective fused fluorescent proteins in a physical distance below 80 Å [10]. As an additional control, FLIM-FRET experiments were performed with constructs where the fluorescent proteins were exchanged (Fig. 8B). Again there was evidence that the C-a 1 splice variant and the canonical a 1 -subunit heterodimerize equally well in intact cells.
The results are expressed as means 6 SEM of at least three independent experiments. For minimum 20 cells were analyzed each. All results were controlled for their statistical significance by Student's t-test. A value of p,0.01 was considered to be statistically significant.
The magnitude of the change of the fluorescence-lifetime (<0.3 ns) was expected to prove an interaction.

Subcellular distribution of C-a 1
While performing expression controls for the FLIM-FRET experiments, we noticed a different subcellular localization of the fluorescent fusion proteins of the C-a 1 splice variant versus the canonical a 1 subunit each expressed in the absence of the b 1subunit. Because of the granular perinuclear appearance of the splice variant, we co-expressed human heme oxygenase 1 (HO-1) with an amino-terminal ECFP as a well-known marker for the endoplasmic reticulum [11] (Fig. 9A and B). While the C-a 1 splice variant showed an exact co-localization with HO-1, the canonical a 1 subunit showed a diverging homogenous distribution in the cytosol (see Fig. 9A and B). Co-expression of untagged b 1 -subunit led to a homogenous wild type like distribution of the C-a 1 splice variant with an additional nuclear signal in some (Fig. 9C, red arrow) but not all cells (Fig. 9C, white arrow). Co-expression of CFP-tagged b 1 -subunit (and omission of CFP-HO1) demonstrates that expression of substantial amounts of the b 1 -subunit induces this translocation to the nucleus (Fig. 9D, red arrow). In contrast, no nuclear expression was detected for the canonical a 1 subunit in the presence of b 1 (Fig. 9E).
Because Sharin et al. [5] have shown that the C-a 1 splice variant is ''uniquely resistant to oxidative protein degradation'' we analyzed the redox-state of both a 1 variants by fusion with the redox-sensor Grx1-roGFP2. Ratiometric analysis showed no difference between Grx1-roGFP2 (Fig. 10A) and a 1 -Grx1-roGFP2 (Fig. 10B). In contrast, C-a 1 -Grx1-roGFP2 showed a higher ratiometric signal indicative of a more oxidized subcellular environment (Fig. 10C). This effect was attenuated by coexpression of the untagged b 1 -subunit (Fig. 10D). In contrast, co-expression of untagged b 1 -subunit with a 1 -Grx1-roGFP2 led to no change ( Fig. 10E in comparison to Fig. 10B).

Discussion
It has recently been shown that the a 1 -subunit of the nitric oxide receptor NOsGC undergoes splicing regulation in differentiating human embryonic cells [5]. The results indicated high levels of an amino-terminally deleted C-a 1 -splice form in differentiating cells that showed a different intracellular distribution in comparison to the canonical full-length a 1 -subunit. It has been published that the amino-terminus of the a 1 -subunit is important for quantitative dimerization with the b 1 -subunit [2]. This was in contrast to a study where we had shown co-purification of the b 1 -subunit with a 1 DN 259 [3]. To explain the discrepancy, Wagner et al. [2] raised the argument that we had applied a purification method, where we pooled the fractions with catalytic activity [3]. Due to this approach, it was conceivable, that only a small fraction which has formed active heterodimers is considered, while non-heterodimerizing subunits are discarded. The present study circumvents this potential problem by using a purification procedure based on an affinity StrepTag attached to the b 1 -subunit in analogy to the study of Wagner et al. [2]. Our results obtained by using copurification of the a 1 -subunit and the amino-terminally deleted Ca 1 -splice form with the StrepTagged b 1 -subunit indicate that the C-a 1 -splice form heterodimerizes quantitatively. Because of the  controversial nature of the question, whether dimerization of nitric oxide sensitive guanylyl cyclase requires the a 1 amino-terminus, we used an additional method based on fluorescence resonance energy transfer to demonstrate heterodimerization of the C-a 1splice form with the b 1 -subunit in intact cells. This corroborated our finding that lack of the amino-terminus of the a 1 -subunit does not influence quantitative dimerization with the b 1 -subunit. This is also consistent with a number of other studies [4,12,13,14]. Our finding that a 1 -variants heterodimerize equally well in intact cells, but the C-a 1 splice form seems to lead to more intact heterodimers upon purification is probably due to the greater stability of the C-a 1 splice variant [5]. This may also be the reason for the higher enzyme activity of the C-a 1 /b 1 heterodimer versus wild type. Alternatively, it is conceivable, that the amino-terminal region absent in C-a 1 has a negative regulatory influence on basal activity. This would be in line with a proposed regulatory domain-scale mechanism of the amino-terminal domains of NOsGC [15] that are in proximity to the catalytic domain [9]. The splice variant would thus represent a disinhibited isoform under basal conditions.
Using fluorescent fusion proteins, we noticed a peculiar localization of the C-a 1 -splice form compared to the wild type. Comparison with fluorescent tagged heme oxygenase-1 which is  . Coomassie staining of cytosolic fractions and purified proteins. 50 mg Cytosol of cytosolic fraction or 5 mg of purified enzyme was electrophoretically separated by SDS-PAGE and stained with Coomassie Blue (a representative gel is shown). Lane A, cytosolic fraction of C-a 1 /b 1 S expressed without heme-supplement and without lipid medium supplement. Lane B, purified C-a 1 /b 1 S expressed without heme-supplement and without lipid medium supplement. Lane C, cytosolic fraction of C-a 1 /b 1 S expressed only with heme-supplement and without lipid medium supplement. Lane D, purified C-a 1 /b 1 S expressed with heme-supplement and without lipid medium supplement. Lane E, cytosolic fraction of C-a 1 /b 1 S expressed only with heme-supplement and with lipid medium supplement. Lane F, purified C-a 1 /b 1 S expressed with heme-supplement and with lipid medium supplement. doi:10.1371/journal.pone.0025772.g005 known to be attached to the endoplasmic reticulum outer membrane [16], revealed a highly similar subcellular distribution. As the endoplasmic reticulum shows a relatively oxidizing thioldisulfide milieu [17], we employed a novel method to measure the glutathione redox potential based on a fusion of the a 1 -subunit in its different splice forms with glutaredoxin-1 and roGFP2 [18]. This confirmed a more oxidizing thiol-disulfide milieu for the C-a 1 splice variant compatible with localization at the endoplasmic reticulum. Co-expression of the non-tagged b 1 -subunit changed the subcellular localization as well as the thiol-disulfide milieu signal for the C-a 1 -splice form but had no effect on the a 1 -subunit. The finding that the a 1 /b 1 -heterodimer showed the expected diffuse cytosolic localization, while the C-a 1 /b 1 -heterodimer was also present in the nucleus, is interesting in the context of reports suggesting a role for the b 1 -subunit in the nucleus: Baltrons and colleagues have demonstrated that the b 1 -subunit translocates to the nucleus for proteosomal degradation in rat astrocytes after treatment with bacterial endotoxin [19]. In a subsequent report the b 1 -subunit was shown to be peripherally associated to chromosomes during mitosis and to play a role in chromatin condensation and cell cycle progression in rat C6 glioma cells [20]. The mechanism by which the b 1 -subunit enters the nucleus is unknown since the protein lacks a recognizable nuclear localization signal [19]. However, it is conceivable that regions of the b 1subunit that may be shielded in the classical a 1 /b 1 -heterodimeric enzyme are exposed in the C-a 1 /b 1 -heterodimer and the b 1homomer and are free to interact with proteins that regulate nuclear import [21].
Sharin et al. have analyzed the subcellular distribution of the native C-a 1 splice form in differentiating human embryonic cells  [5]: With biochemical methods the C-a 1 splice form could be detected in small amounts in the fraction containing nuclei and to a greater extent in the cytosolic fraction [5]. This is consistent with our findings for the overexpressed fluorescent tagged C-a 1 /b 1heterodimer. Immunocytochemical analysis of differentiating human embryonic cells at day 12 with an antibody that recognizes both the a 1 -subunit and the C-a 1 splice form showed a majority (88 %) of cells with a diffuse staining indicative of a predominant localization in the cytosol [5]. Again this is in line with our results for the fluorescent tagged C-a 1 /b 1 -heterodimer and the fluorescent tagged a 1 /b 1 -heterodimer. In a minority of cells (12 %), a filamentous staining was observed in human embryonic cells using an antibody that recognizes both the a 1 -subunit and the C-a 1 splice form. These cells were not analyzed for the b 1 -subunit expression. So it remains a possibility that these cells do express no or less b 1 -subunit. This would be consistent with our finding that the fluorescently tagged C-a 1 splice form shows a different subcellular localization and redox environment alone than in the presence of the b 1 -subunit.
The endoplasmic reticulum is involved in heme trafficking, heme degradation and heme insertion into hemoproteins in eukaryotic cells [22]. Biochemical analysis in native human embryonic cells showed small amounts of the canonical a 1 -subunit in the fraction containing the endoplasmic reticulum. The function of the H-NOX domain of the a-subunits that is absent in the C-a 1 splice variant is not clear. Very recently it has been suggested based on overexpression and purification of the H-NOX-a 1fragment in E. coli that it may bind heme via a non-covalent interaction [23]. It is possible that the lack of this domain in the Ca 1 protein leads to trapping of the splice variant at the site of heme insertion, while the canonical a 1 -subunit or the a 1 /b 1 -heterodimer interacts only transiently with the endoplasmic reticulum e.g. during heme insertion or maturation.
Sharina et al. showed that the C-a 1 splice variant is more stable in intact cells in the presence of ODQ, a well-known oxidant [24] while the canonical a 1 -subunit is more rapidly degraded under these conditions. We show in the current paper that fluorescent labeled overexpressed C-a 1 protein is targeted to a more oxidized environment in comparison to the canonical a 1subunit. Thus the higher stability of the C-a 1 protein in the presence of oxidants may represent an adaption to the redox environment at the endoplasmic reticulum because it is well known to be a more oxidized compartment than the cytosol [6]. NOsGC activators, like cinaciguat, which act at the H-NOX domain of the b 1 -subunit enhance the stability of the protein through inhibition of proteosomal degradation which can be induced by oxidants like ODQ in the cell [25,26]. It is thus conceivable that heme or ligand free H-NOX domains not only of b 1 but also of a 1 subunits provide a signal that leads to more rapid degradation.
In summary, we present evidence that the amino-terminal H-NOX domain of the a 1 -subunit does not preclude quantitative dimerization with the b 1 -subunit to form a stable active, heme containing NOsGC-heterodimer but is important for subcellular localization.  Analytic, Brunswick, Germany. All primers used for site directed mutagenesis, were obtained in HPLC purity grade from biomers.net, Ulm, Germany.

Cloning of a 1 Deletion Mutants
For construction of the a 1 DN 236 mutant, a PvuII/HindIII fragment of the a 1 full-length (see ref. [3]) clone was ligated StuI/HindIII into pFastBac TM 1vector (Invitrogen, Karlsruhe, Germany). Cloning of the human amino-terminal deletion mutant a 1 DN 259 has been described previously [3].
Cloning of carboxy-terminal Strep Tag II b 1 -subunit (Sf-9 system) For construction of the carboxy-terminal Strep Tag II with the b 1subunit a PvuI/StuI fragment of the conjoined NOsGC construct b 1 a 1 -Strep [27] was cloned using PvuI/SmaI into the pFastBac TM 1 vector. Cloning of a 1 WT and C-a 1 fused with fluorescent proteins for determination of fluorescence lifetime For construction of a 1 full-length in pECFP-N1 (Clontech, Mountain View, CA, USA) the a 1 cDNA in pcDNA3.1/V5/His-TOPO described by Haase et al. [27] was cloned HindIII/XhoI into HindIII/SalI pECFP-N1. Cloning of the C-a 1 (rat a 1 DN 258 ) was done by introduction of suitable restriction sites by site directed mutagenesis into this construct, restriction and religation. At the same time the Kozak consensus sequence [28] was optimized and a start-methionine had to be introduced. For construction of C-a 1 -ECFP the following primer pair was used: 59-GAA CCA GCC CTA TTT GCT CGA GTC GGT CGC CAT GGA GAG CAC CAA GCC TTC TCT-39 and 59-AGA GAA GGC TTG GTG CTC TCC ATG GCG ACC GAC TCG AGC AAA TAG GGC TGG TTC-39. The initiation codons or complementary sequence are bold type and the modified nucleotides are underlined. Using the respective restriction sites in pECFP before the insert XhoI digestion and religation led to C-a 1 -ECFP. Due to a slight difference in the amino acid sequence rat a 1 DN 258 correspond to human a 1 DN 259 (Data S1).

Cloning of b 1 fused with EYFP for determination of fluorescence lifetime
The cloning has been described previously [27].

Exchange of ECFP and EYFP
The carboxy-terminal ECFP-fusions of a 1 full-length and C-a 1 were exchanged for EYFP and the carboxy-terminal EYFP-fusion of b 1 was exchanged for ECFP using the pEYFP-N1 or pECFP-N1 vector (Clontech) and AgeI/BsrGI as restriction enzymes.
Restriction of C-a 1 -EYFP and a 1 -Grx1-roGFP2 with Eco47III/ BstEII led to an exchange of the full-length construct for the splice variant.

Control of sequences
All cloned constructs were verified by sequencing (GATC Biotech, Konstanz, Germany).

Baculovirus Generation
Recombinant baculoviruses of respective subunits were generated using the Bac-to-BacH Baculovirus Expression System (Invitrogen).

Sf-9 Cell Culture, Expression of Recombinant Guanylyl Cyclase Subunits and Preparation of Cytosolic Fraction
Sf-9 cells were cultured in Sf-900 TM II serum-free medium (Invitrogen) supplemented with 1 % penicillin/streptomycin (PAA Laboratories, Coelbe, Germany) and 10 % fetal bovine serum (Foetal Bovine Serum Gold, EU-approved, PAA Laboratories). Spinner cultures were grown at 27uC at 140 rpm shaking on a 50 mm orbit platform and diluted to 2 N 10 6 cells/ml for infection. 500 ml of cell solution were infected with the respective recombinant baculovirus stock with the multiplicities of infection (MOI) of 1. The MOI was determined using Sf-9ET cell line according to the recent published paper by Hopkins and Esposito [30]. After 72 h cells were harvested by centrifugation (4,000 g for 1 min at 4uC). All following steps were performed on ice. The cell pellet was resuspended in 30 ml of lysis buffer containing 50 mM TEA-HCl, 1 mM EDTA, 10 mM dithiothreitol (DTT), 250 nM Avidin, pH 7.4, and complete EDTA-free Protease Inhibitor Cocktail Tablets (Roche, Mannheim, Germany). The cells were lysed by sonication. Cytosolic fractions were obtained by centrifugation for 180 min at 15,000 g at 4 uC. The cytosolic fractions were filtered through a 0.2 mm syringe filter (Sartorius, Goettingen, Germany). 2 ml aliquots of cytosolic fractions were kept as reference for experiments to monitor the purification at all steps. For investigation of the influence of heme supplement and additionally of lipid medium supplement 4 mg/l hemin or/and 1 % lipid medium supplement were added.

One-Step-Purification of NOsGC
The purification was performed on ice or at 4uC. The chromatographic step was performed on an Ä KTApurifier 100 system (GE Healthcare, Munich, Germany). Cytosolic fractions were immediately applied to a Strep-TactinH SuperflowH high capacity resin (2 ml Volume in a Tricorn TM 10/20 column) at 1 ml/min [31]. Buffer W contained 100 mM Tris-HCl, 1 M NaCl, 1 mM EDTA, 1 mM benzamidine, 10 mM DTT, pH 8.0. Buffer E was prepared by adding 2.5 mM D-desthiobiotin to buffer W. The column was washed ad 1 ml/min with buffer W for 5 column volumes (CV). With 5 CV buffer E the elution of the NOsGC was performed. Monitoring the absorbance at 254 nm, 280 nm and 430 nm showed a single peak in the fraction which contained NOsGC. Regeneration of the column was performed by 15 CV buffer R (100 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 1 mM HABA, pH 8.0) at 1 ml/min and 8 CV of buffer W at 1 ml/min The pooled fractions (approximately 15 ml) were concentrated to a volume of about 500 ml using an Amicon Ultra-15 centrifugal filter unit with 30 kDa cut-off (Millipore, Schwalbach, Germany).

Determination of Protein Concentration and Guanylyl Cyclase Activity Assay
Protein concentrations were determined by the Warburg-Christian method [32] using a NanoPhotometer TM (Implen, Munich, Germany). Guanylyl cyclase activity was measured as described previously [3]. Purified NOsGC was diluted with 50 mM TEA-HCl, 10 mM DTT, 1 mM EDTA, 0.5 mg/ml bovine serum albumin (Roth, Karlsruhe, Germany), pH 7.4, quick-frozen in liquid nitrogen with 10 % (v/v) glycerol and stored at -80uC. Enzyme activity of purified protein (50 ng of protein per assay tube) were determined by incubation for 10 min at 37uC in the presence of 1 mM cGMP, 0.5 mM [a-32 P]GTP (about 0.2 mCi), 3 mM MgCl 2 , 50 mM TEA-HCl, pH 7.4, 0.25 g/liter creatine kinase, 5 mM creatine phosphate, and 1 mM 3-isobutyl-1methylxanthine in a total volume of 100 ml as described by Schultz and Boehme [33]. Reactions were started by the addition of protein and incubation at 37uC. All experiments were stopped by ZnCO 3 precipitation, and purification of the enzyme-formed cGMP was performed as described previously. Basal enzyme activity measurements were performed in the absence of NO or cinaciguat. Measurements of stimulated enzyme were performed in the presence of the NO donor DEA/NO or cinaciguat. DEA/NO was dissolved in 10 mM NaOH, which did not affect the enzyme activity (data not shown). Cinaciguat was dissolved in 100 % DMSO and then diluted in distilled water to a final concentration of 10 mM so that the final DMSO concentration in the enzyme assay did not exceed 2.5% (v/v). At this concentration no effects of DMSO on enzyme activity were observed (data not shown).

SDS-polyacrylamide electrophoresis gels and immunoblot analysis
Aliquots of 50 mg (cytosolic fractions) or 5 mg (purified enzyme) protein were heated for 3 min at 105uC with lidheat of 120uC (PCR-Cycler) in a modified Laemmli sample buffer (50 mM Tris-HCl, 1% SDS, 100 mM DTT, 30% Glycerol, pH 7.5). After the heat-incubation 1 ml of a blue sample puffer [10% (m/w) bromophenol blue solved in the modified Laemmli buffer] were added and the probes were resolved on 10% slab gels. Proteins were stained according to Kang et al. 2002 [8]. As protein markers predominantly PageRuler TM Prestained Protein Ladder and PageRuler TM Unstained Protein Ladder from Fermentas (St. Leon-Rot, Germany) were used. For immunoblotting, protein fractions were transferred electrophoretically to a nitrocellulose membrane [Amersham TM Hybond TM ECL (GE Healthcare)]. The membrane was reversibly stained with Ponceau S to evaluate the protein transfer. Unspecific binding sites were saturated by immersing the membrane for 1 h at room temperature in TBST (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.1% (v/v) Tween 20) containing 5% nonfat dry milk. The following antibodies against the two NOsGC subunits were used for detection: anti-a 1 (1:5,000) (Sigma, G4280) or a 1 -1200 (1:1,000; as described in [3]) and anti-b 1 (1:4,000) (Sigma, G4530). Antibodies were incubated for 1 h in TBST-buffer at room temperature. The membranes were washed three times for 10 min with TBST and subsequently incubated for 1 h with horseradish-peroxidase-conjugated anti-rabbit IgG antibody (1:2,000/Cell Signaling Technology, distributor: New England Biolabs, Frankfurt am Main, Germany or 1:4,000/Sigma). After three washes with TBST-buffer the membranes were processed with the enhanced chemiluminescence western blotting detection system according to the recommendations of the manufacturer (Roche) and the signals were detected with a charge-coupled device camera (Intas, Goettingen, Germany) or the membranes were processed with ECL Western blotting detection system according to manufacturer's recommendations (Amersham Pharmacia Biotech, Piscataway, USA).

Quantification of heterodimerization
For quantification of subunit interaction, the Coomassie stained gels were either captured using a white trans-illuminator and a charge-coupled device camera (HighRes, Intas) or scanned using ScanMaker i900 (Microtek, Willich, Germany). The analysis was done using LabImage 1D (Kapelan, Leipzig, Germany). For the wild-type of the human a 1 -subunit and C-a 1 the densitometric values of a 1 full-length and the truncation were normalized to the value of b 1 in the same lane and dimerization is given as a percentage of a 1 WT dimerization, which was set to 100%. Western blots were analyzed accordingly with similar results (data not shown).
Fluorescence Lifetime Imaging (FLIM) using a confocal laser scanning microscope Determination of fluorescence lifetime was done as described previously [9]. For microscopy HEK-293 cells were seeded in 24well imaging plates with special glass bottom (zell-kontakt, Noerten-Hardenberg, Germany; distributed by PAA Laboratories, Coelbe, Germany) and transfected with the cDNA coding for the respective constructs for the expression of fluorescent tagged NOsGC subunits using Lipofectamin TM LTX (Invitrogen). 48 h post transfection cells were imaged at 37uC on a Nikon Ti-E microscope equipped with an incubation chamber (Okolab, Naples, Italy) using a 60-fold immersion objective (NA 1.4, Nikon). For live cell imaging culture medium was removed, cells were washed twice and supplemented with Hank's balanced salt solution.
Fluorescence decays were measured in cells expressing the FRET donor alone (ECFP) using the vector pECFP-N1 (Clontech) as a negative control. The rat a 1 -subunits, wild-type and amino-terminal truncation are carboxy-terminally tagged with ECFP as FRET donor and the rat b 1 -subunit is carboxy-terminally tagged with FRET acceptor EYFP using the vector pEYFP-N1 (Clontech) or vice versa. Images were collected using a 405 nm pulsed laser. Emitted fluorescence signals were selected using a CFP bandpass filter (475/20 nm). FLIM images in the time domain from fluorescent cells were recorded with a 4 channel time gated detection system (LiMo module, Nikon). The images were 2566256 pixels in size and the acquisition time was 5 min.
Localization of a 1 constructs by confocal laser scanning microscope The cells were prepared as described above. Amino-terminally ECFP-tagged heme oxygenase 1 (human, HO-1) was used as a marker for the endoplasmic reticulum [11].
Analyzing the redox state of the C-a 1 and the a 1 wild type protein Using fusion proteins of the respective a 1 isoforms with the redox sensor Grx1-roGFP2 we performed a ratiometric measurement. The cells were simultaneously excited with a 405 nm and a 488 nm laser. The emission was collected by a GFP bandpass filter (525/ 50 nm). The ratio of the 405/488 emission corresponds to the redox state. A high value means a more oxidized state and vice versa.

Statistical analysis
The results are expressed as means 6 SEM of at least three independent experiments. All results were controlled for their statistical significance by Student's t-test. A value of p,0.01 was considered to be statistically significant.