Functional domains of soluble guanylyl cyclase.

Soluble guanylyl cyclase is a heterodimer consisting of an alpha and beta subunit and stimulation occurs upon binding of NO to a prosthetic group. Little is known about the localization of catalytic and regulatory domains within the subunits of soluble guanylyl cyclase. We used deletion mutagenesis to identify the regions of alpha 1 and beta 1 subunits that are responsible for cGMP production or NO-heme-mediated activation. The amino terminus of the beta 1 subunit was necessary for NO stimulation since deletion of the 64 NH2-terminal amino acids resulted in a mutant with intact basal activity but complete loss of NO activation. The amino terminus of the alpha 1 subunit also appeared to be essential for NO sensitivity since deletion of 131 NH2-terminal amino acids of alpha 1 led to markedly reduced NO activation. These results suggest that NH2-terminal regions of alpha 1 and beta 1 are involved in NO-heme-mediated signal transduction. The NH2 terminally truncated beta 1 subunit exerted a dominant negative effect exclusively on the NO-stimulated activity of the wild type enzyme, further underlining that the regulatory domain is located within the NH2 terminus of the enzyme. Aside for the structural implications, the mutant represents a powerful tool to investigate nitric oxide-sensitive signaling pathways. Coexpression of the COOH-terminal halves of alpha 1 and beta 1 were sufficient for basal cGMP production while either of the halves expressed alone was inactive. Therefore the COOH-terminal regions appear to contain sufficient information for dimerization and basal enzymatic activity. Thus, we provide the first evidence that the regulatory and catalytic properties of soluble guanylyl cyclase can be attributed to different regions of the subunits and that the catalytic domain can be functionally expressed separately from the NH2-terminal regulatory domain. Taken together with findings on the membrane bound enzyme form, guanylyl cyclases, appear to resemble fusion proteins where different regulatory domains have been joined with a common cGMP-forming segment.

Soluble guanylyl cyclase is a heterodimer consisting of an ␣ and ␤ subunit and stimulation occurs upon binding of NO to a prosthetic group. Little is known about the localization of catalytic and regulatory domains within the subunits of soluble guanylyl cyclase. We used deletion mutagenesis to identify the regions of ␣ 1 and ␤ 1 subunits that are responsible for cGMP production or NO-heme-mediated activation.
The amino terminus of the ␤ 1 subunit was necessary for NO stimulation since deletion of the 64 NH 2 -terminal amino acids resulted in a mutant with intact basal activity but complete loss of NO activation. The amino terminus of the ␣ 1 subunit also appeared to be essential for NO sensitivity since deletion of 131 NH 2 -terminal amino acids of ␣ 1 led to markedly reduced NO activation. These results suggest that NH 2 -terminal regions of ␣ 1 and ␤ 1 are involved in NO-heme-mediated signal transduction. The NH 2 terminally truncated ␤ 1 subunit exerted a dominant negative effect exclusively on the NO-stimulated activity of the wild type enzyme, further underlining that the regulatory domain is located within the NH 2 terminus of the enzyme. Aside from the structural implications, the mutant represents a powerful tool to investigate nitric oxide-sensitive signaling pathways. Coexpression of the COOH-terminal halves of ␣ 1 and ␤ 1 were sufficient for basal cGMP production while either of the halves expressed alone was inactive. Therefore the COOH-terminal regions appear to contain sufficient information for dimerization and basal enzymatic activity.
Thus, we provide the first evidence that the regulatory and catalytic properties of soluble guanylyl cyclase can be attributed to different regions of the subunits and that the catalytic domain can be functionally expressed separately from the NH 2 -terminal regulatory domain. Taken together with findings on the membrane bound enzyme form, guanylyl cyclases, appear to resemble fusion proteins where different regulatory domains have been joined with a common cGMP-forming segment.
Guanylyl cyclases (GTP pyrophosphate-lyase (cyclizing); EC 4.6.1.2), the enzymes catalyzing the formation of cGMP, exist in membrane-bound and soluble forms. The membrane-bound enzymes are stimulated by different peptide hormones and belong to the group of receptor-linked enzymes with one membrane spanning region (1). Accordingly, their structure can be divided into three domains: an amino-terminal extracellular ligand-binding domain, an intracellular protein kinase-like domain, which has been proposed to function as a negative regulatory element (2), and the COOH-terminal catalytic region comprising about 250 amino acids responsible for the synthesis of cGMP, as shown by a deletion mutant of the membranebound guanylyl cyclase GC-A (3,4). In contrast to the membrane-bound enzymes which exist as homodimers or higher ordered structures, soluble guanylyl cyclase consists of two different subunits designated ␣ and ␤, which are both required for catalytic activity (5). Four subunits have been reported up to date (␣ 1 , ␣ 2 , ␤ 1 , and ␤ 2 ). The ␣ 1 ␤ 1 heterodimer corresponds to the enzyme purified from bovine lung, whereas the ␣ 2 and ␤ 2 subunits have been identified by homology screening but have not yet been detected on the protein level. All subunits show some homologies over the whole length of the polypeptide chains; the strongest homologies are found in the COOH-terminal regions which are also shared with the membrane-bound guanylyl cyclases and adenylyl cyclases (6).
Soluble guanylyl cyclase is a heme protein with spectral properties indicative of a 5-coordinate ferrous heme with histidine as the axial ligand (7), and it is the heme moiety which serves as the receptor for nitric oxide (NO), 1 the activator of the soluble enzyme (8). Recently, histidine 105 of the ␤ 1 subunit has been shown to be essential for the stimulation by NO since substitution by phenylalanine yielded an enzyme that was catalytically active but insensitive to NO (9).
It is unknown whether heme binding and catalytic activity can be attributed to distinct domains of soluble guanylyl cyclase. Here, we show that the COOH-terminal halves of the ␣ 1 and ␤ 1 subunits comprising the putative catalytic domain are sufficient for cGMP formation. Moreover, deletions of the poorly conserved NH 2 termini of the ␤ 1 (␤ 1 -⌬N 64 ) and ␣ 1 subunits (␣ 1 -⌬N 131 ) led to severe impairment of NO stimulation, thus underlining the importance of the NH 2 -terminal regions of both subunits for heme binding and/or transduction of the stimulatory binding signal to the catalytic domain. The NH 2terminally truncated ␤ 1 subunit specifically blocks only the stimulated activity of ␣ 1 ␤ 1 wild-type enzyme while leaving the non-stimulated activity unchanged. Therefore, this mutant offers a new and potentially powerful means by which to inhibit the NO stimulated but not basal catalytic activity of the soluble form of guanylyl cyclase. EXPERIMENTAL PROCEDURES NH 2 -terminal truncated mutants of ␣ 1 and ␤ 1 subunits were constructed by polymerase chain reaction (PCR). Fragments coding for shortened NH 2 termini were amplified and used to substitute the re-* This work was supported by Deutsche Forschungsgemeinschaft and Fonds der Chemischen Industrie. 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. spective fragments in the original cDNA clones. Primers used in the PCR had an average length of 22 bases matching the sequences used for amplification. In the sense oligonucleotides, the beginning of the coding region was preceded by the ribosomal binding site derived from the sequence of the ␣ 1 subunit, and a HindIII site was added to the 5Ј-end to aid subcloning of the amplified products. The positions of the antisense oligonucleotides were selected to cover an internal restriction site allowing convenient substitution in the original cDNA. PCR was carried out in 30 cycles (94°C for 60 s; 55°C for 150 s; 72°C for 150 s), using 1 g of cDNA as template, 80 pmol of each primer, and 5 units of Taq-Polymerase (Promega) under the conditions suggested by the manufacturer. After subcloning the PCR products, the truncated mutants were constructed as follows.
For construction of the ␤ 1 -⌬N 64 mutant, a HindIII/SphI fragment (151 base pairs) of the subcloned PCR product was used to substitute for the respective fragment in the cDNA of the ␤ 1 subunit subcloned in the HindIII/EcoRI sites of pUC BM20. The mutant was subcloned into the HindIII/SmaI sites of pCMV by cutting with EcoRI, filling the recessed 3Ј termini, and subsequently cutting with HindIII. For expression in Sf9 cells, the pCMV clone was cut with EcoRI/SspI and subcloned into the EcoRI/SmaI sites of pVL1392.
For construction of the ␤ 1 -⌬N 305 mutant, the cDNA of ␤ 1 subunit was cut with BamHI/EcoRI, and the resulting fragment coding for the COOH-terminal part of the subunit was ligated to the BamHI/EcoRIcut PCR product subcloned in pUC BM20. For expression in Sf9 cells, the pUC BM20 clone was cut with HindIII, the recessed 3Ј termini were refilled and, after a EcoRI cut, the mutant was subcloned into the SmaI/EcoRI sites of pVL1393.
For construction of the ␣ 1 -⌬N 131 mutant, the cDNA of ␣ 1 subunit cloned into the EcoRI site of pBR322 was cut with SacI/EcoRI, and the resulting fragment coding for the COOH-terminal part of the subunit was ligated to the SacI/EcoRI-cut PCR product coding for the shortened NH 2 terminus subcloned in pUC BM20. The resulting mutant was subcloned into the HindIII/SmaI sites of pCMV after cutting with EcoRI, filling of the recessed 3Ј termini and a HindIII cut. For expression in Sf9 cells, the pCMV clone was cut with EcoRI and SspI and subcloned into the EcoRI/SmaI sites of pVL1392.
For construction of the ␣ 1 -⌬N 366 mutant, the cDNA of ␣ 1 subunit cloned into the EcoRI site of pBR322 was cut with ApaI/EcoRI, and the resulting fragment coding for the COOH-terminal part of the subunit was ligated to the ApaI/EcoRI-cut PCR product coding for the shortened NH 2 terminus subcloned in pUC BM20. The mutant was subcloned into the HindIII/SmaI sites of pCMV using the EcoRI and HindIII sites of the mutant cDNA clone in pUC BM20. For expression in Sf9 cells, the NH 2 -terminal fragment of the mutant (BglII/NcoI) was used to substitute for the respective region (BamHI/NcoI) of the wild type cDNA of ␣ 1 in pVL1393. Construction of the wild type ␣ 1 and ␤ 1 subunits, generation of recombinant viruses, and cell culture of Sf9 cells were performed as described (9). To obtain cytosolic fractions containing the recombinant protein, spinner cultures of Sf9 cells grown to 1.5 ϫ 10 6 cells/ml were coinfected with combinations of the respective baculoviruses as the indicated apparent multiplicity of infection (m.o.i.). Cells were collected by centrifugation 42 h after infection and resuspended in 40 l (1 mM EDTA, 2 mM dithiothreitol, 0.2 M benzamidine, 50 mM triethanolamine hydrochloride, pH 7.4) per ml of the original cell suspension. After passing the cells 10 times through a 22-gauge needle, the homogenates were centrifuged for 30 min at 30,000 ϫ g. The resulting supernatants (50 -100 g) were assayed for cyclase activity as described previously (9) in the presence of 0.1 mM GTP, 3 mM Mg 2ϩ or Mn 2ϩ with or without 100 M S-nitrosoglutathione (GSNO) unless stated otherwise. Endogenous cGMP formation of Sf9 cells infected with only the ␤ 1 virus (up to 10 pmol of cGMP ϫ min Ϫ1 ϫ mg Ϫ1 ) was substracted as control from recombinant heterodimer activity. Data are shown for typical experiments performed in duplicate which were repeated at least once. Western blots were performed as described (10).

RESULTS
In order to be able to encircle functional domains of soluble guanylyl cyclase, we analyzed the homology of different regions of the subunits. As outlined above, each subunit contains a region homologous to the catalytic domain of the membranebound enzyme which is also conserved in the adenylyl cyclases. This COOH-terminal region is preceded by a central part of the polypeptide chains with less but still pronounced homology between guanylyl cyclase subunits. In contrast, the overall homology of the NH 2 -terminal parts of the subunits is low.
Notable is a stretch of about 100 amino acids within this NH 2 -terminal region revealing a significantly higher degree of identical amino acids between ␣ 1 and ␣ 2 or ␤ 1 and ␤ 2 than between ␣ and ␤ subunits. In one set of deletions, we omitted the amino acids preceding this stretch (131 and 64 amino acids of the ␣ 1 and ␤ 1 subunits, respectively) to investigate the potential function of these very low conserved NH 2 termini. In another set of mutations, the entire NH 2 -terminal half of either subunit was deleted so that the resulting truncated subunits (␤ 1 -⌬N 305 and ␣ 1 -⌬N 366 ) contained the putative catalytic regions. All deletion mutants were coexpressed with their corresponding wild type subunit in Sf9 cells. In order to obtain comparable amounts of expressed proteins, the m.o.i. used for the different viruses was varied, and the expression of the recombinant proteins was monitored in cytosolic fractions of infected cells in Western blots using antibodies directed against COOH-terminal peptides of the ␣ 1 and ␤ 1 subunit. Fig.  1 shows the results of a Western blot in which the m.o.i. used for the respective heterodimers were adjusted. Accordingly, the expression of the recombinant truncated mutants and the wild type subunits was quite similar. Like the wild type ␣ 1 subunit (calculated molecular mass of 78 kDa), all deletional mutants of the ␣ 1 subunit showed a slightly faster migration on SDS gels than expected from the predicted molecular masses.
Subsequently, enzyme activity of the modified heterodimers was determined under basal and NO-stimulated conditions in the presence of 3 mM Mg 2ϩ or 3 mM Mn 2ϩ as divalent cations (Table I). cGMP production by the ␣ 1 ␤ 1 wild type in the presence of Mg 2ϩ was 50 pmol of cGMP ϫ min Ϫ1 ϫ mg Ϫ1 ; addition of GSNO resulted in a 90-fold increase in catalytic activity. Among the modified guanylyl cyclases, only ␤ 1 -⌬N 64 revealed measurable basal activity (40 pmol of cGMP ϫ min Ϫ1 ϫ mg Ϫ1 ) in the presence of Mg 2ϩ . Whereas basal activity was comparable to that of the wild type enzyme, GSNO did not alter enzyme activity of this mutant. In contrast, ␣ 1 -⌬N 131 exhibited some sensitivity to NO in the presence of Mg 2ϩ as the enzyme showed cGMP formation (10 pmol of cGMP ϫ min Ϫ1 ϫ mg Ϫ1 ) under stimulated conditions whereas basal activity was below the detection limit. The use of Mn 2ϩ instead of Mg 2ϩ as divalent cation led to an increased basal activity of the wild type (400 pmol of cGMP ϫ min Ϫ1 ϫ mg Ϫ1 ), with the activator causing only a 10-fold stimulation. These findings are in accordance with earlier results (11). Under these conditions, all modified heterodimers were catalytically active but insensitive to NO. Basal activity of ␤ 1 -⌬N 64 (503 pmol of cGMP ϫ min Ϫ1 ϫ mg Ϫ1 ) was again in the range of the wild type enzyme, whereas cGMP formation of the other mutant enzymes was reduced to 9 -22% of wild type activity (see Table I). After establishing intact basal activity of ␤ 1 -⌬N 64 but insensitivity to NO, this mutant was further characterized. To compare the kinetic properties of this mutant with the wild type enzyme, cGMP formation was determined in the presence of increasing GTP concentrations. A Lineweaver-Burk plot of the data revealed similar apparent K m values of 106 and 110 M GTP for ␣ 1 ␤ 1 and ␣ 1 ␤ 1 -⌬N 64 , respectively (Fig.  2), suggesting proper folding of the catalytic domain.
Since the ␤ 1 -⌬N 64 heterodimer is not activated by the NOreleasing compound GSNO nor by protoporphyrin IX (data not shown), we tested whether the mutant could function as a dominant-negative protein inhibiting only the stimulated wild type enzyme while leaving the basal activity unchanged. For this purpose, Sf9 cells were infected with the viruses coding for wild type subunits as well as with increasing amounts of the virus coding for the ␤ 1 -⌬N 64 mutant. Parallel to the determination of the catalytic activity, we performed Western blots of the different samples to ensure that the loss of stimulated activity was not due to a decrease in expression of the wild type subunits (Fig. 3B). Catalytic activity of the coexpressed subunits in cytosolic fractions was determined under basal and stimulated conditions (100 M GSNO) in the presence of 3 mM Mn 2ϩ . As shown in Fig. 3A, increasing amounts of ␤ 1 -⌬N 64 virus led to a decline of the stimulated activity to 60, 37, 29, 23, and 20% of the original activity (5200 pmol of cGMP ϫ min Ϫ1 ϫ mg Ϫ1 ) of the wild type enzyme, whereas basal cGMP forma-tion remained constant over a range of 10 to 30 pmol of cGMP ϫ min Ϫ1 ϫ mg Ϫ1 .

DISCUSSION
In the present study, we demonstrate that coexpression of the COOH-terminal halves of the ␣ 1 and ␤ 1 subunits of soluble guanylyl cylcase, comprising the region conserved in all cyclases, yields an enzyme sufficient for the formation of cGMP but insensitive to NO. Preliminary results indicate that COOH-  I Guanylyl cyclase activity of deletion mutants expressed in Sf9 cells Sf9 cells were infected with viruses coding for the indicated subunits and mutants at the following m.o.i.; 0.5 and 1 for the ␣ 1 ␤ 1 ; 0.5 and 0.25 for ␣ 1 ␤ 1 -⌬N 64 ; 1 and 0.05 for ␣ 1 ␤ 1 -⌬N 305 ; 0.5 and 3 for ␣ 1 -⌬N 131 ␤ 1 ; 0.05 and 3 for ␣ 1 -⌬N 366 ␤ 1 ; 0.5 and 0.1 for ␣ 1 -⌬N 366 ␤ 1 -⌬N 305 . The NH 2 -terminally truncated subunits are schematically represented, the numbers give the amino acids present in the respective mutant. Guanylyl cyclase activities were determined with Mn 2ϩ or Mg 2ϩ as indicated in the absence or presence of 100 M GSNO. Further details are given under "Experimental Procedures." Functional Domains of Soluble Guanylyl Cyclase terminally truncated mutants are catalytically inactive, thus further underlining the importance of the catalytic consensus domain. Our results are in good agreement with reports on an NH 2 -terminally truncated GC-A receptor mutant exhibiting ligand-independent cGMP production. In the deletional mutant of the GC-A receptor, the cyclase catalytic domain is preceded by the so-called hinge region (47 amino acids) and a small portion (15 amino acids) of the carboxyl segment of the consensus protein kinase-like domain. These additional amino acids were shown to be essential for a properly folded catalytically active site, and it was suggested that these sequences may be required for dimerization (12). In a recent report, the 43 amino acids in front of the catalytic domain have indeed been shown to be required for dimerization (13). In analogy, the 80 additional amino acids NH 2 -terminal of the putative catalytic region present in the COOH-terminal halves of the ␣ 1 and ␤ 1 subunits may be required for proper folding or dimerization of the subunits. Dimerization of the subunits is a prerequisite for enzyme activity since expression of one subunit or a truncated mutant did not yield any cGMP forming activity. The various adenylyl cyclases also contain two cyclase homology domains which in contrast to soluble guanylyl cyclase are localized on one polypeptide chain. Similarily as in soluble guanylyl cyclase, both domains present in adenylyl cyclases are required for catalysis as the separate expression of either region results in a loss of enzyme activity (14). More-over, the ability of the NH 2 -terminally truncated GC-A receptor mutant to dimerize coincided with intact catalytic activity, further underlining the necessity of two catalytic consensus domains. The significance of the association of two different catalytic domains in soluble guanylyl cyclases and adenylyl cyclases as opposed to the existence of two identical catalytic domains in the membrane bound guanylyl cyclases, however, is unknown.
Besides the identification of the catalytic domain, we show that even the very low conserved NH 2 -terminal part of the ␤ 1 subunit (64 amino acids) is required for the stimulation by NO. Recently, we identified histidine 105 of the ␤ 1 subunit as a likely candidate to be the residue forming a linkage to the central iron atom of the prosthetic heme group, since substitution of this histidine 105 by phenylalanine yielded an NOinsensitive enzyme lacking the prosthetic heme group (9). Although histidine 105 of the ␤ 1 subunit is present in the mutant ␤ 1 -⌬N 64 , the loss of NO sensitivity suggests disturbance of either a configuration required for heme binding or lack of structures involved in the transduction of the stimulatory effect to the catalytic center. Deletion of the NH 2 -terminal part of the ␣ 1 subunit (131 amino acids) yielded an enzyme still activated by NO, although the non-responsiveness in the presence of Mn 2ϩ suggests a severe impairment of NO stimulation. As further deletion of the ␣ 1 subunit also destroys NO activation of the enzyme, we conclude that the NH 2 -terminal regions of the ␣ 1 and ␤ 1 subunits are responsible for the regulation of the enzyme. Characterization of the purified mutants will reveal whether those are still able to bind heme or whether the mediation of the stimulatory signal is impaired. When coexpressed with the wild type enzyme, the NH 2 -terminally truncated ␤ 1 subunit inhibits only the stimulated activity but leaves the basal activity unchanged. The ability of the mutant to compete with the wild type ␤ 1 subunit for dimerizing with the ␣ 1 subunit further emphasizes the NH 2 -terminal location of the regulatory domain. The dominant-negative effect of the mutant on the stimulated activity will allow the selective inhibition of nitric oxide-dependent cGMP formation in intact cells. Hence, the mutant will be a helpful tool in forthcoming studies of the NO-cGMP signaling system and, in this regard, the retained ability of basal cGMP formation is probably beneficial when compared to the recently identified dominant-negative mutants that knock out guanylyl cyclase activity completely (16).
Our results show that the two defining features of soluble guanylyl cyclase, namely the catalytic activity and the regulation by NO, can be attributed to different regions on the subunits and that the catalytic domain can be expressed separately. Similar results exist for the bacterial protein FixL, a dimeric hemoprotein kinase whose enzymatic activity is reversibly blocked by oxygen binding to the heme. There the heme binding domain and the phosphotransferase activity were also attributed to the NH 2 -terminal and COOH-terminal regions, respectively; they still contained their heme-dependent oxygenbinding and catalytic properties when expressed separately (15). These findings suggest that during evolution, regulatory segments have been connected to distinct catalytic domains. In the case of guanylyl cyclases, a common cGMP-forming catalytic unit occurs under the control of different regulatory domains; in that respect, the regulatory heme-containing domain of the soluble enzyme appears to represent the equivalent to the ligand-binding and kinase-like domains of the membrane bound guanylyl cyclases. Acknowledgment-We thank Jü rgen Malkewitz for excellent technical assistance.  ). A, cyclase activity was determined in the presence of 3 mM Mn 2ϩ as described under "Experimental Procedures." B, shown are immunoblots of the same cytosolic protein preparations that were used for the determination of enzymatic activity. Cytosolic proteins (30 g) were separated on SDS-polyacrylamide electrophoresis gels and blotted onto nitrocellulose, peptide antibodies against the COOH termini of the ␤ 1 and ␣ 1 subunits were used for detection of the wild type subunits and the ␤ 1 -⌬N 64 mutant.