Differential Regulation of Adenylyl Cyclases by Gαs *

Regulation of adenylyl cyclases 1, 2, and 6 by Gαs was studied. All three mammalian adenylyl cyclases were expressed in insect (Sf9 or Hi-5) cells by baculovirus infection. Membranes containing the different adenylyl cyclases were stimulated by varying concentrations of mutant (Q227L) activated Gαs expressed in reticulocyte lysates. Gαsstimulation of AC1 involved a single site and had an apparentK act of 0.9 nm. Gαsstimulation of AC2 was best explained by a non-interactive two site model with a “high affinity” site at 0.9 nm and a “low affinity” site at 15 nm. Occupancy of the high affinity site appears to be sufficient for Gβγ stimulation of AC2. Gαs stimulation of AC6 was also best explained by a two-site model with a high affinity site at 0.6–0.8 nm and a low affinity site at 8–22 nm; however, in contrast to AC2, only a model that assumed interactions between the two sites best fit the AC6 data. With 100 μm forskolin, Gαs stimulation of all three adenylyl cyclases showed very similar profiles. Gαs stimulation in the presence of forskolin involved a single site with apparentK act of 0.1–0.4 nm. These observations indicate a conserved mechanism by which forskolin regulates Gαs coupling to the different adenylyl cyclases. However, there are fundamental differences in the mechanism of Gαs stimulation of the different adenylyl cyclases with AC2 and AC6 having multiple interconvertible sites. These mechanistic differences may provide an explanation for the varied responses by different cells and tissues to hormones that elevate cAMP levels.

Regulation of adenylyl cyclases 1, 2, and 6 by G␣ s was studied. All three mammalian adenylyl cyclases were expressed in insect (Sf9 or Hi-5) cells by baculovirus infection. Membranes containing the different adenylyl cyclases were stimulated by varying concentrations of mutant (Q227L) activated G␣ s expressed in reticulocyte lysates. G␣ s stimulation of AC1 involved a single site and had an apparent K act of 0.9 nM. G␣ s stimulation of AC2 was best explained by a non-interactive two site model with a "high affinity" site at 0.9 nM and a "low affinity" site at 15 nM. Occupancy of the high affinity site appears to be sufficient for G␤␥ stimulation of AC2. G␣ s stimulation of AC6 was also best explained by a two-site model with a high affinity site at 0.6 -0.8 nM and a low affinity site at 8 -22 nM; however, in contrast to AC2, only a model that assumed interactions between the two sites best fit the AC6 data. With 100 M forskolin, G␣ s stimulation of all three adenylyl cyclases showed very similar profiles. G␣ s stimulation in the presence of forskolin involved a single site with apparent K act of 0.1-0.4 nM. These observations indicate a conserved mechanism by which forskolin regulates G␣ s coupling to the different adenylyl cyclases. However, there are fundamental differences in the mechanism of G␣ s stimulation of the different adenylyl cyclases with AC2 and AC6 having multiple interconvertible sites. These mechanistic differences may provide an explanation for the varied responses by different cells and tissues to hormones that elevate cAMP levels.
Nine adenylyl cyclases have been cloned and characterized (1)(2)(3). Though these adenylyl cyclase isoforms display distinct signal receiving capabilities from Ca 2ϩ (4,5), G␤␥ (6), and protein kinase C (7), they all share the common capability to be stimulated by G␣ s . Due to its expected nature, regulation by G␣ s has received somewhat less attention than other modes of regulation. Little is known about the regions of adenylyl cyclases involved in G␣ s interactions. Early studies with the purified olfactory adenylyl cyclases showed that G␣ s could be covalently linked to adenylyl cyclase (8). As a prelude to the mapping of adenylyl cyclase regions involved in interactions with G␣ s , we have studied regulation of recombinant AC6, AC2, and AC1 by mutant (Q227L) activated G␣ s . Much to our surprise, we find G␣ s regulation of the three different adenylyl cyclase appears to be mechanistically different. Stimulation of AC1 is the simplest and appears to involve a single site of interaction. Stim-ulation of AC2 and AC6 appears to be best explained by multiple sites of interaction. This complex regulation is observable only in the absence of forskolin. In the presence of forskolin, stimulation by G␣ s of all three adenylyl cyclases tested appears to involve a single high affinity site.

EXPERIMENTAL PROCEDURES
Materials-In vitro translation kits and Flexi rabbit reticulocyte lysates and reagents for RNA synthesis were from Promega (Madison, WI), and [␣-32 P]ATP was from ICN. Most biochemicals were from Sigma, and cell culture supplies were from Life Technologies, Inc. All other reagents were the highest grade available.
In Vitro Synthesis of G␣ s and Q227L-G␣ s -The wild-type G␣ s (short form without the Ser) cloned from a human liver library (8) and the Q227L-G␣ s (␣ s *) 1 in pAGA were the gift of Dr. Juan Codina (UT Health Sciences Center, Houston). As described previously by Sanford et al. (9) pAGA is derived from pGEM-3Zf(Ϫ) and contains the 5Ј-untranslated sequences of the alfalfa mosaic virus RNA (for sequence, see Ref. 10) upstream of the ␣ s * sequence. Transcription and translation was according to the method of Sanford et al. (9). The vector containing pAGA-␣ s * was linearized with XhoI, and RNA was transcribed using T7 RNA polymerase. The newly transcribed mRNA was purified by phenol/ chloroform extraction, precipitated by ethanol, and stored in RNasefree water until use. The ␣ s * mRNA was translated in rabbit reticulocyte lysates with all 20 amino acids (20 M) and approximately 3 ϫ 10 5 cpm of [ 35 S]Met (ϳ1,200 Ci/mmol). The translation product was washed by 10-fold dilution and concentrated in a Centricon-10 concentrator. After three washes, the proteins in 25 mM Na-Hepes, 1 mM EDTA, and 20 mM-␤-mercaptoethanol were frozen in a dry-ice acetone bath and stored at Ϫ80°C until used. There are seven Met in G␣ s . Assuming that all seven methionines were labeled, specific activity of the translated G␣ s was calculated by measurement of percent incorporation of 35 Slabeled L-Met into trichloroacetic acid precipitable material.
Expression of Adenylyl Cyclases in Insect Cells-AC2 and AC6 were expressed in Sf9 or Hi-5 cells as described previously (11,12). Bovine AC1 was tagged with the FLAG epitope at the N terminus using a strategy similar to that used for AC2 (11). The recombinant F-AC1 containing virus was then constructed by use of the shuttle vector pVL1392. Sf9 cells were grown in serum-free insect medium while Hi-5 cells were grown in Grace's medium supplemented with L-glutamine, yeastolate, and lactalbumin with 10% fetal bovine serum and 10 g/ml gentamycin. Cells in log phase growth were infected with a multiplicity of 5-10 and harvested 48 h postinfection.
Preparation of Adenylyl Cyclase-containing Membranes-Infected cells were collected by centrifugation at 4°C. All further procedures were at 4°C. Cells were washed in lysis buffer containing 20 mM Na-Hepes, pH 8.0, 5 mM EDTA, 1 mM EGTA, 150 mM NaCl, 2 mM dithiothreitol, and a mixture of protease inhibitors including 1 mM phenylmethylsulfonyl fluoride (freshly prepared), 1 mM o-phenanthroline, 3.2 g/ml each leupeptin and soybean trypsin inhibitor, and 2 g/ml aprotinin and then resuspended in 10 volumes of lysis buffer. Cells were lyzed by nitrogen cavitation in a Parr bomb at 600 p.s.i. for 30 min. Lysates were centrifuged at 1000 ϫ g for 10 min (without break) to pellet the cell debris. The 1000 ϫ g supernatant was spun at 100,000 ϫ g for 45 min. The 100,000 ϫ g pellet was resuspended in 20 mM Na-Hepes, 1 mM EDTA, pH 8.0, 200 mM sucrose, 2 mM dithiothreitol, and the protease mixture to obtain a final concentration of 3-5 mg/ml protein. Aliquots were frozen in a dry-ice acetone bath and stored at Ϫ80°C.
Immunoblotting-Membranes containing AC2 and AC6 were blotted * The work was supported in part by National Institutes of Health Grants DK-38761 and GM54508 (to R. I.). 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.
with the AC comm antiserum at 1:2000 dilution. The bands were visualized by horseradish peroxidase-coupled second antibody using the ECL system. Adenylyl Cyclase Assays-Sf9 or Hi-5 cells containing adenylyl cyclase was assayed for activity in the presence of 0.1 mM[␣-32 P]ATP (ϳ1000 cpm/pmol), 10 mM MgCl 2 and other additives as described previously (12). Typically, 3-5 mg of crude membrane protein was assayed in a volume of 30 -50 l for 15 min at 32°C. The concentration of G␣ s used is described in the individual experiments.
Data Analysis-The G␣ s concentration effect curves were analyzed using the program PROPHET (Version 4.3) on a Sun Sparc workstation. Prophet is sponsored by the NIH and distributed by BBN Systems Technology. Initial determinations of the models to be used for fitting were made by visual inspections of the Hofstee plots. AC1 data were fitted to a single-site model using the equation V ϭ V max *A/(K act ϩ A), where V is the activity of adenylyl cyclase at A, the concentration of the activator G␣ s , V max is the maximum reaction rate, and K act is the concentration of the activator at half V max . Initially, the AC2 data were also fitted to a one-site model. However, a significant improvement in the fit was obtained when the data were fit to a noninteractive two-site model using the equation where V is the activity of AC2 at concentration A of G␣ s and the parameters for the two sites are labeled with corresponding subscripts. The AC6 data were first fit to a one-site model and then to a noninteractive two-site model. Neither model fit the data adequately. Hence, the AC6 data were then fitted to an interactive two-site model. At low concentrations of G␣ s , the data are fitted with function, V ϭ V max1 *A/ (K act1 ϩ A), since only one (the high affinity) site would be effective at low concentrations of G␣ s . We then assumed that the low affinity site would become operative only above a certain threshold concentration of G␣ s . The data points above this threshold concentration were then fitted to a modified two-site model described by the equation The threshold concentration was estimated by initial visual inspection of the G␣ s concentration-effect curve followed by an iterative fitting process to obtain a threshold value that generated a curve that best fit the data points. Three parameters were used to evaluate the model that best fit the data. The first was the p values of fitted constants, the second was the residual mean squares (RMS) value of the fit, and the third was the F-test statistics to show the significance of the improved fit. Plots of the data points and the fitted curves were generated by Prophet. The printed plots were exported to the Canvas program in a Mac 8100 by use of an optical scanner. The plots were labeled and assembled within Canvas. The final plots as shown were printed as Canvas files.

RESULTS
The wild type and mutant activated G␣ s expressed in reticulocytes was tested for its ability to stimulate S49 cyc Ϫ membranes. Wild-type G␣ s did not stimulate adenylyl cyclase activity in the presence of GTP. However, extensive stimulation was seen with the nonhydrolyzable GTP analog Gpp(NH)p (data not shown). However, with ␣ s *, extensive stimulation was observed with or without added GTP (Fig. 1). In contrast, control lysate did not stimulate the cyc Ϫ adenylyl cyclase. These results indicated that we had synthesized an activated G␣ s , which we used in further studies. When we studied different adenylyl cyclase isoforms expressed in Sf9 cells, we were cognizant of the activity of the endogenous adenylyl cyclases of Sf9 cell membranes. In the case of AC1 and AC2, the endogenous activities represented 3-5% of the total activity under all conditions and hence were not subtracted. In the case of AC6, however, the endogenous activity was between 20 -30%. Hence for all AC6 experiments, the difference between control (Sf9 cells infected with baculovirus containing thyroid peroxidase) and AC6-containing membranes for all concentration points was taken as AC6 activity.
The effect of varying concentrations of G␣ s * in the presence and absence of forskolin on AC1 expressed in Sf9 cells is shown in Fig. 2. In the absence of forskolin, G␣ s * stimulated AC1 with an apparent K act of 0.9 nM. Stimulation by G␣ s * showed a simple one-site Michaelis-Menten mechanism, and a standard Hofstee transformation yielded a straight line ( Fig. 2A, inset). Addition of 100 M forskolin decreased the apparent K act of G␣ s * 6-fold to 0.16 nM; however, the nature of the curve was unaltered (Fig. 2B).
We studied the effect of varying concentrations of G␣ s * in the presence and absence of forskolin on AC2. In the absence of forskolin, stimulation by G␣ s * did not saturate even at 35 nM (Fig. 3A). Upon Hofstee transformation, a curvilinear profile was observed (Fig. 3A, inset). The data best fit a Michaelis-Menten noninteracting two-site model. The apparent K act for the two sites were 0.9 and 15 nM. To determine if this heterogeneity of interactions arose from multiple forms of AC2 due to differential post-translational modification, we analyzed the AC2-containing membranes by immunoblotting with the ACcomm antiserum. A single labeled band of approximately 120 kDa was observed. This profile indicates that the expressed AC2 is not a mixture of differentially modified proteins. We next determined the profile of the G␣ s response in the presence of forskolin. With 100 M forskolin, G␣ s * stimulation of AC2 yielded a simple profile (Fig. 3B); the Hofstee transformation was linear, with a single apparent K act of 0.39 nM measured (Fig. 3B, inset). In Fig. 3B, the activity due to forskolin alone (313 pmol/min/mg) had been subtracted prior to data analysis and fitting. We next tried to define the conditions of G␣ s interaction with AC2 that was optimal for G␤␥ stimulation. Without G␣ s , no significant stimulation of AC2 by G␤␥ was observable (data not shown). At 2.0 nM, G␣ s stimulation by G␤␥ was seven-fold. In contrast, at 20 nM G␣ s , G␤␥ stimulated only three-fold over G␣ s (Fig. 3C). These results suggest that occupancy of the high affinity site may be sufficient for G␤␥ stimulation.
We next studied G␣ s stimulation of AC6. G␣ s stimulation of AC6 was biphasic (Fig. 4A). Hofstee transformation yielded a curvilinear profile (Fig. 4A, inset II). The G␣ s concentrationresponse data for AC6 best fit to a two-site model. However, the two-site fitting for AC6 was different from that used for AC2. At low concentration of G␣ s , the data showed a single-site interaction profile (Fig. 4A, inset I) The second site was observable only at concentrations of G␣ s above a certain threshold value. For the experiment shown in Fig. 4A, the threshold value was 3 nM. AC6 activities above 3 nM G␣ s are the summation of interactions at both sites. Hence a modified two-site model was used where, for activities above 4 nM ␣ s *, the activity due to the second site is the difference between the observed activity and the maximal activity due to occupancy of the first site. This method of fitting is suggestive of interaction between the sites and may imply that the high affinity site needs to be fully occupied to permit occupancy of the low affinity site. Given the very pronounced biphasic profile, we analyzed the AC6-containing membranes from multiple forms of AC6. However, as shown in the immunoblot in Fig. 4A, a single band was seen with the AC comm antibody. The observed band was at 135 kDa as expected for AC6. We then compared the effect of varying concentrations of G␣ s on AC6 activity in the presence and absence of forskolin in the same experiment. In the absence of forskolin (Fig. 4B), the profile was very similar to that in Fig. 4A. In the presence of forskolin, stimulation of AC6 by G␣ s was simple (Fig. 4C). Hofstee transformation yielded a straight line (Fig. 4C, inset), indicating that the data could be fit to a one-site model. G␣ s activated AC6 with an apparent K act of 0.2 nM in the presence of forskolin. Furthermore, even in the presence of saturating G␣ s , forskolin potentiated AC6 activity by about 30%. To ascertain if the unusual profile of G␣ s stimulation seen with AC6 was possibly due to aggregation in the insect cell membranes, we solubilized the AC6-containing membranes with dodecyl maltoside and then assayed the solubilized AC6 in the presence of varying concentrations of G␣ s . The solubilized enzyme had lower activity, and the G␣ s response was right shifted. However the biphasic nature of the curve was not affected by solubilization (Fig. 4D). These observations indicate that the biphasic nature of the AC6 response is not due to nonspecific aggregation. DISCUSSION Stimulation through G␣ s is the major mechanism by which mammalian adenylyl cyclases are activated and cellular cAMP levels elevated. Given the ubiquitous nature of this effect, one might expect the existence of a conserved mechanism for G␣ s activation of the different mammalian cyclases. The data shown here indicate otherwise. G␣ s stimulates AC1 in a mono-FIG. 3. Effect of varying concentrations of ␣ s * on AC2 activity in the absence (A) and presence (B) of 100 M forskolin. 5 g of Sf9 cell membranes containing AC2 was assayed in the presence of indicated concentrations of ␣ s *. Immunoblot of the AC2 containing membranes with the AC comm antiserum is shown to the right of the curve in panel A. 5 g of membrane protein was used for immunoblotting. The Prophet program was used to analyze the data and obtain the fitted curves. Inset shows Hofstee transformations of the data for panels A and B. The apparent K act and V max obtained from the fitting are shown in the boxes. For the data in Fig. 3A, the p values for the constants using the two-site model were p(V max1 ) Ͻ 0.006, p(K act1 ) Ͻ 0.001, p(V max2 ) Ͻ 0.001, and p(K act2 ) Ͻ 0.004. The RMS values of the one-versus the two-site fits were 436 versus 206. The F statistic for the one-site model was 1298 and for the two-site model was 1380. For the data in panel B, the p values for constants were less than 0.001. C, effect of varying concentrations of ␣ s * on stimulation of AC2 by 100 nM G␤␥ subunits purified from bovine brain. Values are means of triplicate determinations. Coefficient of variance was less than 5%. These experiments are representative of five other experiments except for panel C, which was repeated twice. For other details, see "Experimental Procedures." tonic fashion and is the simplest of the three isoforms studied here. Forskolin increases the affinity of G␣ s for AC1 without altering the profile of the activity curve. It is noteworthy that the apparent k act for G␣ s is similar to the K i for G␤␥ (13), thus for all practical purposes G s -coupled receptors would be unable to activate AC1, as has been recently demonstrated by Storm and co-workers (14). Stimulation of AC2 and AC6 by G␣ s is more complex. In the absence of forskolin, G␣ s stimulation of AC2 and AC6 is best explained by a two-site model. However, forskolin not only increases the affinity of G␣ s but also renders stimulation monotonic, indicating that both the low and high affinity sites can be converted to a higher single affinity site in the presence of forskolin. The single G␣ s site in the presence of forskolin indicates that the two sites displayed by AC2 and AC6 are not due to two populations of enzymes when expressed in Sf9 cells.
For AC2, occupancy of the "high affinity" G␣ s site appears to be sufficient for G␤␥ stimulation, since maximal-fold stimulation by G␤␥ is observed at 2 nM G␣ s . This may indicate that some of the determinants for G␣ s binding are in the cytoplasmic tail. Interaction of G␣ s with these determinants could open up one of the regions involved in G␤␥ binding. Such a model would provide a molecular explanation for why G␤␥ only stimulates AC2 activated by G␣ s .
Though G␣ s stimulation of both AC2 and AC6 best fit a two site model, our kinetic analysis indicates that there may be crucial mechanistic differences in G␣ s stimulation of the two adenylyl cyclases. The AC2 data can be fit to Michaelis-Menten type of equations, suggesting two noninteracting sites. In contrast, the model used to fit the AC6 data required an iterative two-site fitting procedure where the estimated V max of the high affinity site had to be subtracted from the observed activity at higher G␣ s concentrations to fit the second site so as to obtain an overall best fit. Such a fitting protocol may indicate interactions between the two sites such that the high affinity site needs to be occupied to obtain occupancy of the low affinity site.
The kinetic descriptions of G␣ s interactions with the various adenylyl cyclases provided here should not be construed as an indication that each molecule of AC2 or AC6 contains two G␣ s binding sites. This is one possible explanation. In such a case, AC1 would either have only one G␣ s site or two sites with very similar affinities. Studies by Neer and co-workers on the hydrodynamic properties of the G␣ s -adenylyl cyclase from bovine caudate which yield molecular weights for G␣ s -adenylyl cyclase complex in the presence of forskolin of 197-225 kDa (15) may support the stoichiometery of 2 G␣ s /adenylyl cyclase. This pos-V max obtained from the fitting are shown in the boxes. A, profile of stimulation by varying concentrations of G␣ s . Inset I is a magnified plot of the data at lower concentrations of G␣ s . Immunoblot of the AC6containing membranes with the AC comm antiserum is shown to the right of the curve in panel A. 5 g of membrane protein was used for immunoblotting. The p values for the constants using the interactive two-site model were V max1 Ͻ 0.001, K act1 Ͻ 0.001, V max2 Ͻ 0.001, and K act2 Ͻ 0.001. The RMS value of the one-site fit was 208, the non-interactive two-site fit was 17.8, and the interactive two-site model was 5. The F statistic for the one-site model was 38, for the non-interactive two-site model was 275, and for the interactive two-site model was 1377. B, G␣ s stimulation was measured in the absence of forskolin. The p values for the constants using the interactive two-site model were less than Ͻ 0.001. C, G␣ s stimulation was measured in the presence of 100 M forskolin. These data and those in Fig. 4B were part of one experiment. The p value for constants was less than 0.01. D, stimulation of solubilized AC6 by different concentrations of AC6. AC6 was solubilized in 0.8% dodecyl maltoside as described previously (11). The 60,000 ϫ g supernatant was used for the assay. Detergent concentration in the assay 0.03%. 5 g of solubilized membrane protein was used for the assay. The experiments in panels A-C are representative of five other experiments The experiment in panel D was repeated twice. For other details, see "Experimental Procedures." FIG. 4. Effect of varying concentrations of ␣ s * on AC6 activity. 5 g of Hi-5 cell membranes containing AC6 was assayed in the presence of indicated concentrations of ␣ s *. The Prophet program was used to analyze the data and obtain the fitted curves. Insets (except I in panel A) show Hofstee transformations of the data. The apparent K act and sibility needs to be explored further in detail for the different adenylyl cyclases. There also exist data that do not support 2 G␣ s sites/adenylyl cyclase model. Cross-linking of G␣ s to the olfactory adenylyl cyclases suggest a 1:1 stoichiometery (8). If each adenylyl cyclase molecule contains one G␣ s site, our data may suggest that adenylyl cyclases have a propensity to dimerize in the membrane environment. If different adenylyl cyclases have differing capabilities to form homodimers, then our data would suggest that AC1 had a very low capability to form homodimers while AC2 and AC6 are more likely to dimerize. Future studies analyzing cross-linking G␣ s to adenylyl cyclase as well as accurate molecular size determinations of the G␣ sadenylyl cyclase complex within the membrane are required to determine the stoichiometery of G␣ s -adenylyl cyclase interactions. Irrespective of the details that emerge from these studies, it has become abundantly clear that each adenylyl cyclase can respond differently to G␣ s and that such differential responses could be crucial factors in determining why different tissues and organs show specialized capability to regulate cAMP elevation.