Distinct Patterns of Bidirectional Regulation of Mammalian Adenylyl Cyclases*

The capacities of the (Y subunits of pertussis toxin- sensitive guanine nucleotide-binding regulatory proteins (G proteins) to inhibit different isoforms of mammalian adenylyl cyclases were assessed. Membranes from Sf9 cells infected with recombinant baculoviruses encoding either type I, 11, V, or VI adenylyl cyclase were reconstituted with purified G protein subunits. Types V and VI adenylyl cyclase are most sensitive to inhibition by Gim1, Gid, and Gius; type I adenylyl cyclase can be inhibited by these three Gi, proteins and by Go, as well. Type I1 adenylyl cyclase appears to be immune to inhibition by these proteins. Examination of the effects of native and mutant Gi, proteins, as well as analysis of competition for binding of G,, to adenylyl cyclases, indicate that at least certain adenylyl cyclases have independent sites for interaction with G,, (site 1, stimulatory) and Gi, (site 2, inhibitory). High concentrations of Gi, can interact with site 1 on types I and I1 adenylyl cyclase and activate the enzymes. Types I and I1 adenylyl cyclase also appear to have independent sites for interaction with G protein fly sub- units. The type I enzyme is strongly inhibited, while type I1 adenylyl cyclase is activated if G,, is also present. EDTA, and 0.1 mM GDP. Sequence analysis indicated that both GTPyS- and AMFG-activated NC-G., were homogeneous preparations that lacked the first 34 amino acid residues of the native protein. GTPyS- and AMFG-acti-vated NC-G,, displayed identical capacities to activate adenylyl cyclase. GDP-bound NC-G., was generated by incubating the AMFG-activated protein with excess EDTA for 1 h at 22 "C, and AMFG and EDTA were removed by gel filtration. The affinity of NC-G., for P-y was assessed by examining the capacity of Py to inhibit the steady-state GTPase activity of the protein. GTPase assays were performed as described (34), except the concentration of Mg2+ was 0.2 mM.

and Go heterotrimers results from an identical process of guanine nucleotide exchange and subunit dissociation, it has been difficult to decide whether GTP-a, Py, or both mediate inhibition of adenylyl cyclase.
During the past 4 years cDNAs encoding six distinct isoforms of adenylyl cyclase have been cloned and expressed (6-13). These discoveries and those of additional partial clones (14 -16) indicate that the family of adenylyl cyclases is unexpectedly large and diverse. The proteins share the same basic topology and have extensive regions of sequence homology, but they differ in their tissue distribution, relative abundance, and, most interestingly, regulatory properties (17, 18). The latter fact has compounded the difficulties in assessing mechanisms of inhibition of adenylyl cyclase activity.
Although all membrane-bound forms of mammalian adenylyl cyclase are activated by G,,, they differ dramatically in their responses to other regulatory molecules. For example, types V and VI adenylyl cyclase are inhibited by low micromolar concentrations of Ca2+, while types I and I11 are activated by Ca2+/calmodulin; types I1 and IV adenylyl cyclase are insensitive to physiological concentrations of Ca2+ (13, [18][19][20]. The G protein Py subunit complex inhibits type I adenylyl cyclase, but it greatly potentiates GB,-mediated activation of the type I1 and IV enzymes; the other forms of adenylyl cyclase are relatively insensitive to P y (9, 19, 21). Observations of this type have heightened appreciation of the necessity for systematic identification of the particular isoform(s) of adenylyl cyclase present in individual cells and thorough characterization of the regulatory properties of each. Toward this end we have expressed each isoform of adenylyl cyclase in Sf9 cells using the recombinant baculovirus system. We have used membranes derived from these cells to assess the regulatory properties of individual adenylyl cyclases with respect to interactions with forskolin, calmodulin, G,,, and G protein Py subunits (9, 19, 21), and we have purified types I and I1 adenylyl cyclase from these membranes to demonstrate their direct interactions with P y (22).
Recently, we also demonstrated that recombinant (Escherichia coli-derived) myristoylated Giul could inhibit adenylyl cyclase activity in membranes derived from Sf9 cells expressing either the type I or the type V enzyme (23). To characterize such responses more thoroughly, we have now examined the effects of myristoylated G~,I, Gi,z, Gio3, and Go, on adenylyl cyclases types I, 11, V, and VI, and we have initiated studies designed to determine the mechanism of inhibition of adenylyl cyclase activity by these proteins. for the culture of Sf9 cells and the production, cloning, and amplification of recombinant baculoviruses have been described by Summers and Smith (24). Baculoviruses encoding types I, 11, and V adenylyl cyclase have been described previously (19,21,23). AcDNA that encodes canine type VI adenylyl cyclase (11) was excised from pcDNAamp274 with EcoRI and Ssp1 and was cloned into pVL1392 that had been digested with 6093 This is an Open Access article under the CC BY license.

EXPERIMENTAL PROCEDURES Sf9 Cell Culture and Recombinant Baculoviruses-Procedures
subunits (Gim1, Gim2, Gia3, and Go,), proteins were coexpressed with yeast protein N-myristoyltransferase (26,27). Purification of recombinant a subunits was achieved by modifications of the methods of Linder et a1. (281, as described by Lee et al. (26). Protein concentrations were estimated by staining with Amido Black (29).
Adenylyl Cyclase Assays-Adenylyl cyclase activity was measured a s described by Smigel (32). All assays were performed for 5-7 min at 30 "C i n a final volume of 100 pl. The concentration of MgC1, was 4 m~.
Membranes and G protein subunits were incubated for 3 min a t 30 "C in a total volume of 40 pl prior to initiation of the assay; GDPPS (25 y) was included during this incubation, as was recombinant G,, (30)  , and variable amounts of the indicated GTPyS-activated G protein a subunit in 20 pl of buffer containing 20 m~ NaHEPES (pH 8.0), 4 mM MgCI,, 1 mM dithiothreitol, 100 p~ forskolin, and 100 p~ GDPPS. The reaction mixtures were incubated at 30 "C for 10 min (2-3 min are required to reach equilibrium). [3sSlGTPyS-G., that was not associated with membranes was removed by filtration through 0.22-pm Millipore duropore membranes, followed by washing with 6 ml of 20 mM Na-HEPES (pH 8.0), 4 ~l l~ MgCl,, 1 m~ dithiothreitol, and 10 p~ forskolin. The amount of labeled G,, retained on the filters was determined by scintillation counting. Specific binding was calculated by subtracting binding to Sf9 cell membranes prepared from cells expressing 0-galactosidase from that observed to membranes from cells expressing adenylyl cyclase. Further details of this assay will be presented elsewhere. 2 Generation and Purification of Duncated G,,-Methods for the expression in E. coli and purification of G, subunits containing hexahistidine tags at the amino terminus have been described by Lee et al. (26). Hexa-histidine G., (16 mg) was purified to near homogeneity from a 5-liter culture of E. coli by Ni2+-NTA affinity chromatography (Qiagen). G,, lacking the amino-terminal 34 amino acid residues (and the hexa-histidine tag, designated NC-G.,) was synthesized by limited tryptic digestion. Purified hexa-histidine G. , (in 50 mM Tris-HC1 (pH 8.0), 1 mM EDTA, and 1 mM dithiothreitol) was activated a t 20 "C for 30 min with either 10 p~ GTPyS and 10 mM MgSO, or with 30 p~ AlCl,, 10 mM MgCl,, 10 mM NaF, and 10 VM GDP (AMFG) and then diluted to a final concentration of 1.7 mg/ml in the same buffer. Trypsin was added (0.02 mg/ml final concentration) and the sample was incubated at 4 "C for 30 min. Digestion was terminated by addition of soy bean trypsin inhibitor (0.08 mg/ml final concentration). More than 95% of the G,, was truncated and existed as a single species with an apparent molecular mass of 40 kDa. Immunoblotting with site-specific antibodies indicated that the amino terminus was missing and that the carboxyl terminus was intact.
The sample was diluted &fold and loaded directly to a 10-ml MonoQ fluoride) containing the appropriate activator (either GTPyS or AMFG). Protein was eluted from the column with a 60-ml linear gradient of NaCl(0-300 mM) in the same buffer. Fractions were analyzed by electrophoresis. To remove residual undigested hexa-histidine GB,, the sample was loaded to a 2-ml Ni2+-NTA column equilibrated with buffer A containing the appropriate activator, and the flow-through and washes were analyzed for NC-G,, by electrophoresis. Fractions containing NC-G., were stored in 50 mM NaHEPES (pH 8.0), 1 m~ EDTA, and 0.1 mM GDP. Sequence analysis indicated that both GTPyS-and AMFGactivated NC-G., were homogeneous preparations that lacked the first 34 amino acid residues of the native protein. GTPyS-and AMFG-activated NC-G,, displayed identical capacities to activate adenylyl cyclase. GDP-bound NC-G., was generated by incubating the AMFG-activated protein with excess EDTA for 1 h a t 22 "C, and AMFG and EDTA were removed by gel filtration.
The affinity of NC-G., for P-y was assessed by examining the capacity of Py to inhibit the steady-state GTPase activity of the protein. GTPase assays were performed as described (34), except the concentration of Mg2+ was 0.2 mM.

RESULTS
Inhibition of Isoforms of Adenylyl Cyclase by G,, Subunits: Types V a n d VI-Our initial observation that purified myristoylated Gial could inhibit type V adenylyl cyclase (23) prompted us to examine the generality of this response with additional isoforms of the enzyme. Results with type VI adenylyl cyclase, which shares many regulatory properties with the type V enzyme, are shown in Fig. 1. The activity of type VI adenylyl cyclase (activated with GTPyS-G,,) is also inhibited by recombinant myristoylated GTPyS-Gial. The apparent affinity of type VI adenylyl cyclase for the inhibitory protein is roughly 50 nM, and more than 80% of enzymatic activity can be inhibited (at a 50 I " concentration of activated r G s a ) . Boiled protein is without effect, ruling out inhibition due to components of the buffer or to unbound GTPyS. The significance of the concentrations of Gi, proteins required to inhibit adenylyl cyclase activity has been discussed previously (23).
As anticipated, substantially higher concentrations of the GDP-bound form of myristoylated Gial are required to inhibit type VI adenylyl cyclase than are needed for the GTPyS-bound form of the protein, presumably reflecting the relative affinities of the two forms of the (Y subunit for the enzyme. (Some of the activity observed with the GDP-bound form of myristoylated Gia1 may be due to exchange of GDP for an activating nucleotide (e.g. GTPpS).) In addition, as described for type V adenylyl cyclase, myristoylation of Gial at the amino terminus is required to observe inhibition; micromolar concentrations of the unmodified protein are without effect. Myristoylated, activated Go, is also ineffective.
Similar patterns of effects were observed when activated, myristoylated Gia1, GiaP, and Gias were compared for their capacities to inhibit types V and VI adenylyl cyclase in Sf9 cell membranes (Fig. 2). Neither adenylyl cyclase distinguishes among the three related Gi, proteins. Inhibition of type VI adenylyl cyclase appeared to occur at slightly lower concentrations of the G,, proteins, and the extent of inhibition is somewhat greater than with type V adenylyl cyclase. In addition, the Gi, proteins are somewhat more efficacious inhibitors of G,,-activated adenylyl cyclase activity than of forskolin-stimulated activity with both the types V and VI enzyme. The effects of the activated G,, proteins on G,,-stimulated adenylyl cyclase activity in S49 cell cyc-membranes are very similar to those seen with the type VI enzyme in Sf9 cell membranes, consistent with the observation that S49 cells express predominantly type VI (35).
ripe ZZ-We were next interested in determining if activated Gi, could inhibit forms of adenylyl cyclase (e.g. type I and type 11) that have distinctly different regulatory properties. Fig. 3 shows results obtained with membranes from Sf9 cells expressing the type I1 enzyme. In the presence of GTPyS-G,,, myristoylated GTPyS-Gi,I inhibits enzymatic activity only weakly, while micromoloar concentrations of GTPyS-Go, are without effect. The effects of GTPyS-activated Gias and GiuS are similar to those of Giml (not shown). Since G protein py subunits activate type I1 adenylyl cyclase (conditionally, in the presence of G,,), we suspected that GTPyS-Gi,I does not itself inhibit type I1 adenylyl cyclase but rather the observed inhibition might be due to interaction between small amounts of GDP-Gi,, in the preparation and endogenous Py in the Sf9 cell membranes, thereby preventing Py from activating adenylyl cyclase. Consistent with this hypothesis is the finding that the GDP-bound forms of both Gi,, and Go, are more potent inhibitors of type I1 adenylyl cyclase than is myristoylated GTPyS-Gjml. Also consistent with the hypothesis is the fact that GTPyS-activated Giul could not inhibit forskolin-activated type I1 adenylyl cy- clase (Fig. 3B ); p y does not activate forskolin-stimulated type I1 adenylyl cyclase. Of interest, not only did GTPyS-Gi,, fail to inhibit forskolinactivated type I1 adenylyl cyclase, it actually activated the enzyme at concentrations in the p~ range (Fig. 3B); Giaz and Gim3 behaved similarly (not shown). In this case the GTPySbound form of Gial was more active than the GDP-bound form of the protein, and GTPyS-Go, was nearly devoid of activity.
Because forskolin and G,, activate type I1 adenylyl cyclase synergistically (71, these data suggest that high concentrations of activated Gi, may be mimicking G,, in these experiments.
Additional data that substantiate this notion are presented below.
Type I-We have shown previously that activated myristoylated Giu, can inhibit type I adenylyl cyclase partially when the enzyme is stimulated by Ca2+/calmodulin and less extensively in the presence of forskolin; inhibition is very modest when G,, is the activator. Very similar patterns are observed when Gia1, Gia2, and Giu3 are compared (Fig. 4). Maximal inhibition is achieved at 100-300 nM concentrations of the proteins; the extent of inhibition is 50% with Ca2+/calmodulin and only 20% with forskolin.
The inhibitory effect of the G protein f l y subunit complex on type I adenylyl cyclase, described previously (19), is both more prominent and is exerted at somewhat lower concentrations than the effect obtained with Gi,. The effects of Py and myristoylated Giml are compared in Fig. 5 for both the Gs,-and the Ca2+/calmodulin-activated type I enzyme. The inhibitory effects of Gi, and py are additive, at least at certain concentrations (Fig. 6 A ) ; the percentage inhibition that can be obtained with Gi, is larger in the presence of /3y (Fig. 6B). The experiments in Fig. 6 were performed in the presence of forskolin. A similar pattern was seen with Ca2+/calmodulin-activated type I adenylyl cyclase. Surprisingly, myristoylated GTPyS-Go, is also capable of inhibiting Ca2+/calmodulinor forskolin-activated type I adenylyl cyclase (Fig. 7). Although the extent of inhibition of enzymatic activity by Gi, and Go, is comparable, Go, is approximately 10 times less potent. Nevertheless, this effect of Go, may be physiologically relevant, since the concentrations of Go, in brain are very high and exceed those of Gi, by a considerable amount (31). Activated G,, (300 nM) did not inhibit type I adenylyl cyclase (not shown). We observed a consistent capacity of activated Gin proteins to cause a paradoxical increase in type I adenylyl cyclase activity at concentrations higher than those necessary to observe maximal inhibition. This effect was obvious in the presence of Ca2+/ calmodulin (Figs. 4,5, and 7) or forskolin (Figs. 4,6, and 71, but not G,, (Fig. 5). While less dramatic, this behavior is reminiscent of the effects of Gi, on forskolin-stimulated type I1 adenylyl cyclase described above. The fact that this phenomenon is not observed in the presence of G. , again suggests that it is due to interaction of high concentrations of Gi, with a binding site for GB, on adenylyl cyclase (see below).
Mechanism of Inhibition of Adenylyl cyclase by Gin-The observation that Gi, can inhibit adenylyl cyclase activity in the absence of G,, (i.e. in the presence of forskolin or calmodulin) indicates that inhibition need not be due to competition between the two homologous G protein CY subunits for a common binding site on adenylyl cyclase. Although kinetic analysis of the effects of Gi, on G,,-activated type V adenylyl cyclase indicates a largely competitive relationship between the two proteins over a relatively narrow range of concentrations of G,, (2-20 m; not shown), this is not true when a broader range of concentrations is examined (Fig. 8). As the concentration of G,, is raised, the maximal extent of inhibition that can be obtained with Gi, is clearly reduced.
These data are thus consistent with a model in which G,, and Gi, bind to distinct sites on adenylyl cyclase. Activation results from binding of G,, to site 1, while binding of Gi, to site 2 causes inhibition. We propose that Gi, may also have a modest affinity for site 1 and make contacts appropriate to activate adenylyl cyclase under unusual circumstances. Thus, in the presence of forskolin or Ca2+/calmodulin, Gi, first inhibits type I adenylyl cyclase (site 2) but then activates at high concentrations (site 1). Binding of Gin to site 1 of type I1 adenylyl cyclase leads to substantial activation of the enzyme in the presence of forskolin, presumably because site 2 is absent or inapparent To test this model, we expressed and purified the myristoylated form of a mutant of Giul predicted to have a n increased affinity for site 1. Mutational analysis has identified several regions of G,, that are necessary for activation of adenylyl cyclase (34, 36, 37). For example, replacement of residues between positions 263 and 269 of G. , with the corresponding residues of Gi, (254-260) results in G,, mutants with little affinity for adenylyl cyclase. We reasoned that reciprocal mutations in Gi, (replacement with G,, sequence) should increase the affinity of the mutant Gin for site 1. Analysis of the effects of a mutant designated myristoylated Gialci-,, (residues 258-261 (Phe-Trp-Asp-Trp) changed to Leukg-Tyr-Ile) is shown in Fig. 9. In the absence of other activators, Gi,l(i-s, has a weak stimulatory effect on types I and I1 adenylyl cyclase and little effect on the type VI enzyme (Fig. 9, A-C). However, in the presence of forskolin (which activates adenylyl cyclase synergistically with G,, (particularly types I1 and VI)), Gi,l(i+sl is a more potent activator of type I or type I1 adenylyl cyclase than is the wild-type Gial protein; it is obviously not as potent as G,, (Fig. 9, D E ) . A more dramatic effect of the mutation is observed on forskolin-activated type VI adenylyl cyclase (Fig. 9F). The mutant Gi, is not an inhibitor of the enzyme; it is an activator, presumably because of preferential affinity for site 1 under these conditions.
Binding of G, , -A binding assay was developed to detect interactions of activated G,, with membranes from Sf9 cells expressing different isoforms of adenylyl cyclase.2 Binding of [35SlGTPyS-G,, to membranes from cells expressing either type I or type I1 adenylyl cyclase is 3-6-fold higher than is binding of the labeled protein to membranes from cells infected with a baculovirus encoding P-galactosidase. The amount of binding observed is consistent with the amount of adenylyl cyclase present in these membranes, based on their specific enzymatic activities. Although the extent of binding of G,, to membranes containing type I or type I1 adenylyl cyclase is similar in the presence or absence of forskolin, it is necessary to include forskolin to obtain an adequate signal when using membranes containing type V or type VI adenylyl cyclase (data not shown). Because the signal-to-noise ratio in these assays is not high, we interpret the results semiquantitatively. As expected, addition of unlabeled GTPyS-G,, inhibits binding of the labeled complex by competition (Fig. 10). Half-maximal competition is observed at roughly 10-30 m concentrations of GTPyS-G,,, in reasonable agreement with concentrations of GTPyS-G,, necessary to activate the adenylyl cyclases (Fig. 9). Activated myristoylated Gi,i(i+sl is also a n effective competitor for G,,-binding sites with all three types of adenylyl cyclase. Required concentrations ( p~) are close to those required to observe activation of adenylyl cyclase by this mutant protein. High concentrations of myristoylated GTPyS-Giml compete for G,, binding sites on type I and type I1 adenylyl cyclases (compare curves with those for nonmyristoylated Gia1, which we take as a n ineffective control); these are the two adenylyl cyclases where activation is seen at high concentrations of myristoylated Gia1. There is no significant competition by myristoylated Giml for G.,-binding sites on type VI adenylyl cyclase. Under the conditions utilized, this protein inhibits and does not activate the enzyme. These data are entirely consistent with the model discussed above, wherein G,, and Gi, interact with distinct sites on adenylyl cyclase to activate and inhibit the enzyme, respectively.

Interactions Between G,, and Py on Types Z and ZZ Adenylyl
Cyclase-As discussed above, Py appears to be the most potent and efficacious inhibitor of type I adenylyl cyclase, and the protein is also a n effective conditional activator of the type I1 enzyme, stimulating activity dramatically in the presence of G,, (but not forskolin). Despite the fact that Py inhibits both Gs,-and Ca2+/calmodulin-activated type I adenylyl cyclase, the question has arisen about the nature of the interaction between G,,, Py, and types I and I1 adenylyl cyclase. Do G,, and Py interact independently with these enzymes or might they interact as the heterotrimer (despite the low affinity of GTPyS-G,, for PY)?
In the absence of a workable binding assay for Py, we have addressed this issue by examining the effects of Py on types I and I1 adenylyl cyclase in conjunction with an altered G,, that has been truncated at the amino terminus by digestion of the GTPyS-activated protein with trypsin (Fig. 1lA). This treatment, which removes the amino-terminal 34 amino acid resi-dues, leaves the guanine nucleotide binding properties of the protein intact, as well as its interactions with adenylyl cyclase. However, it greatly reduces the affinity of the protein for Py (38,39). The reduced affinity of truncated G,, for Py is documented in Fig. llB, wherein interactions between G,, and Py were assessed by examination of the capacity of Py to inhibit the steady-state GTPase activity of Gse. The basis of this effect lies in Py-induced slowing of the rate of dissociation of GDP from Gsa. Interaction between Py and G,, could not be detected at the highest concentrations of Py tested (100 m). Truncated G,, activates types I and I1 adenylyl cyclase normally (Fig. 12, A and C). Of interest, there was no change in the capacity of Py to inhibit G,,-activated type I adenylyl cyclase or to activate G,,-stimulated type I1 adenylyl cyclase when these assays were performed in the presence of wild-type G,, or truncated G,, (Fig. 12, B and D ) . These experiments support the hypothesis that these adenylyl cyclases have independent binding sites for Gs, and PY.

DISCUSSION
The discoveries of several isoforms of membrane-bound adenylyl cyclases in mammals and the elucidation of pathways for type-specific regulation of their activities permit synthesis of this information in the form of distinct schemes for regulation of intracellular concentrations of cyclic AMP (Fig. 13). Three patterns have emerged to date: those represented by adenylyl cyclases types V and VI, by types I1 and Iv (although information is incomplete about type W, it is poorly expressed in Sf9 cells), and by type I. (Type I11 is similar to type I in that it is activated by Ca2+/calmodulin, but other information is incomplete.) The complexity of these schemes is remarkable, as is the plasticity among them. They will certainly become more complex and interesting as additional isoforms of adenylyl cyclase are studied and as additional layers of regulation are incorporated (e.g. covalent modifications and allosteric regulation by small molecules).
There are a few constant features. All types of adenylyl cyclase studied to date are activated by GSe. All types studied to date are regulated, directly or indirectly, by members of three of the four major classes of G proteins (Gs, Gi, and G,; the functions of GluI3 are unknown). Of less obvious regulatory significance are the facts that all types are activated by forskolin and inhibited by P-site analogs." Beyond this, evolution has endowed the different isoforms with impressive diversity, often using the same or related molecular players to accomplish different ends. There is type-specific and bidirectional regulation by fly, Gi,, Go,, Ca2+/calmodulin, and, indirectly, G,. G, and G, or G,, Gi, and G, can apparently collaborate a s activators. Gi and G, can collaborate as inhibitors. The effects of G, are indirect in all cases, and three different Ca2+-related mechanisms are involved: direct inhibition of types V and VI by Ca2+ (13,18,20), activation of type I1 by protein kinase C (mechanism unknown) (40)(41)(42), and activation of type I by Ca2+/ calmodulin (19). Given the symmetrical structures of the adenylyl cyclases, we speculate that duplication of the gene provided the opportunity to acquire independent binding sites for the two homologous G protein CY subunits, G,, and Gi,.
Types V and VI adenylyl cyclase represent the simplest pattern of regulation. They are activated by G,, and they are inhibited strongly by Gi,. Under the relatively simple conditions explored to date, the inhibitory effects of Ca2+ appear to be more modest than are those of Gi,. I t can be anticipated that additional forms of adenylyl cyclase will be discovered that are Approximately 2.5 pg each of purified wild-type G. , ( WT-as), truncated G,, (NC-as), or bovine brain py was treated with N-ethylmaleimide and then resolved on a n 11% polyacrylamide gel. Proteins were visualized by staining with Coomassie Blue. B, GTP hydrolysis by Grim.  or NC-as (0) (5 nM each) was incubated in 50 pl of buffer containing 0.5 pm [y32P1GTP (16,500 counts/min/pmol), 1 mM EDTA, and 0.2 mM MgSO, in the presence of increasing amounts of bovine brain py for 10 min a t 30 "C. Release of R2Pi was determined and expressed as percent of Pi released in the absence of py (100%. 0.088 pmol and 0.070 pmol for WT-as and NC-as, respectively). Pi release associated with By alone was subtracted from each value. Results are the mean of duplicate determinations and are representative of three similar experiments. simply activated by G,, and inhibited by Gi, without being subject to regulation by Ca2+; this is the classical picture. Type I1 is noteworthy for strong, conditional activation by py. The apparent function of this adenylyl cyclase as a detector for coincidental activation of G,-and Gi,"-or G,-linked receptors has been discussed previously (17,40). Interaction between G,and Gi,,-controlled pathways appears to be based on differential affinity of the cyclase for G,, and py. Thus, the concentra- tions of Py required to coactivate the enzyme are hypothesized to be generated only by activation of Gi or Go, not by activation of G., which is present in much lower abundance (17,21). We are unable to detect substantial inhibition of type I1 adenylyl cyclase by Gi, or Go,. This makes sense in view of the stimulatory effect of fly. It would be difficult to rationalize release of both a n activator and an inhibitor as a result of dissociation of the subunits of the Gi or Go oligomer. However, it is conceivable that Gi, might inhibit type I1 adenylyl cyclase under some condition not tested in the present experiments.
%e I adenylyl cyclase appears to have independent sites for regulation by the following four types of proteins: G,,, GiJGo,, py, and Ca2+/calmodulin. G,, and Ca2+/calmodulin activate the enzyme to similar extents, and they can interact synergistically in doing so (19). Under the conditions studied, f l y is the more efficacious inhibitor of the enzyme. Gi, and Go, can also inhibit, but this effect is largely limited to the Ca2+/calmodulin stimulated activity. This is the only form of adenylyl cyclase studied to date that can be inhibited by Go,. Although the apparent affinity of Go, for type I adenylyl cyclase is lower than that of Gi,, Go, is present in higher concentrations. We assume that Go, also acts at site 2 on type I adenylyl cyclase. Go, has no obvious effect on types V or VI adenylyl cyclase, which are inhibited strongly by Gi,. This result appears to maintain harmony among the multitude of regulatory interactions, since types V and VI are inhibited by Ca2+ and the inhibitory effect of Go on Ca2+ influx would oppose this mechanism. By contrast, Go, can apparently inhibit type I adenylyl cyclase both directly (at site 2) and by opposing G,-mediated increments in intracellular Ca2+ concentrations.
We detected no differences in the capacity of Giul, Gia2, and GiaS to inhibit types V, VI, and I adenylyl cyclase. The same lack of specificity has been observed with other effectors that are controlled by G protein subunits. Different isoforms of phospholipase Cp fail to discriminate among Gq,, Gll,, and G16, (43, 44), and all three Gi,s activate cardiac K+ channels with similar potencies and efficacies (45). Although information is less complete, the same is largely true for different isoforms of G protein fly subunit complexes (with the exception of the retinal complex, ply1). Several different species of j3y have indistinguishable interactions with adenylyl cyclases (46) and phospholipases. The significance of heterogeneity within subgroups of G protein subunits may lie with interactions between G protein oligomers and their receptors (48-50) or with patterns of cellular and subcellular distribution.