Three Discrete Regions of Mammalian Adenylyl Cyclase Form a Site for Gsα Activation*

The interaction between the α subunit of G protein Gs (Gsα) and the two cytoplasmic domains of adenylyl cyclase (C1 and C2) is a key step in the stimulation of cAMP synthesis by hormones. Mutational analysis reveals that three discrete regions in the primary sequence of adenylyl cyclase affect the EC50values for Gsα activation and thus are the affinity determinants of Gsα. Based on the three-dimensional structure of C2·forskolin dimer, these three regions (C2 α2, C2 α3/β4, and C1β1) are close together and form a negatively charged and hydrophobic groove the width of an α helix that can accommodate the positively charged adenylyl cyclase binding region of Gsα. Two mutations in the C2 α3/β4 region decrease theV max values of Gsα activation without an increase in the EC50 values. Since these three regions are distal to the catalytic site, the likely mechanism for Gsα activation is to modulate the structure of the active site by controlling the orientation of the C2 α2 and α3/β4 structures.

The interaction between the ␣ subunit of G protein G s (G s␣ ) and the two cytoplasmic domains of adenylyl cyclase (C 1 and C 2 ) is a key step in the stimulation of cAMP synthesis by hormones. Mutational analysis reveals that three discrete regions in the primary sequence of adenylyl cyclase affect the EC 50 values for G s␣ activation and thus are the affinity determinants of G s␣ . Based on the three-dimensional structure of C 2 ⅐forskolin dimer, these three regions (C 2 ␣2, C 2 ␣3/␤4, and C 1 ␤1) are close together and form a negatively charged and hydrophobic groove the width of an ␣ helix that can accommodate the positively charged adenylyl cyclase binding region of G s␣ . Two mutations in the C 2 ␣3/␤4 region decrease the V max values of G s␣ activation without an increase in the EC 50 values. Since these three regions are distal to the catalytic site, the likely mechanism for G s␣ activation is to modulate the structure of the active site by controlling the orientation of the C 2 ␣2 and ␣3/␤4 structures.
Mammalian adenylyl cyclase is the enzyme responsible for integrating multiple extracellular and intracellular signals to generate cAMP and thus activate cAMP-dependent protein kinase and cyclic nucleotide-gated ion channels (1,2). All nine cloned mammalian and Drosophila rutabaga adenylyl cyclases are stimulated directly by the ␣ subunit of G s (G s␣ ), 1 and all but type IX are activated by forskolin. G s␣ and forskolin bind and activate adenylyl cyclases separately or synergistically when presented together (3,4). Mammalian adenylyl cyclases are integral membrane proteins consisting of two homologous cytoplasmic domains (C 1 and C 2 ), each following a membrane domain (M 1 and M 2 ) (1, 2). The C 1 and C 2 domains form the catalytic core and can be engineered as a G s␣ -and forskolinsensitive soluble adenylyl cyclase, i.e. by mixing of IC 1 protein (C 1 domain of type I adenylyl cyclase) and IIC 2 protein (C 2 domain of type II adenylyl cyclase) in vitro (5)(6)(7)(8). In this paper, we describe mutations at three discrete regions of the soluble adenylyl cyclase, one in the IC 1 protein and two in the IIC 2 protein, that significantly affect G s␣ activation with little change in forskolin activation.

Construction, Expression, and Purification of Wild Type and Mutant
Forms of IC 1 and IIC 2 Proteins-Plasmids used to express mutant forms of IC 1 and IIC 2 were constructed by site-directed mutagenesis using pProExHAH6-IC 1 or -IIC 2 as the phagemid (9). Oligonucleotides used for mutagenesis contained 10 -12 complementary nucleotides flanking each side of the target codon(s) that was replaced with the appropriate codon. Mutations were confirmed by dideoxy nucleotide sequencing of phagemid DNA.
To express wild type and mutant forms of hexohistidine-tagged IC 1 and IIC 2 , the plasmids that encoded wild type or mutant forms of IC 1 or IIC 2 were transformed into Escherichia coli BL21(DE3) cells. E. coli cells that harbored the desired plasmid were cultured in T7 medium containing 50 mg/ml ampicillin at 30°C (10). When A 600 reached 0.4, isopropyl-1-thio-␤-D-galactopyranoside (100 M) was added. After 3-4 h, the induced cells were then collected and lysed; IIC 2 proteins were purified using the nickel nitrilotriacetic acid column and fast protein liquid chromatography Q-Sepharose column as described (5). The Coomassie Blue staining of SDS-polyacrylamide gel electrophoresis was used to determine the protein peak in the fractions from Q-Sepharose column. The concentration of proteins was determined using Bradford reagent and bovine serum albumin as standard (11). The construction of plasmid H 6 -pQE60-G s␣ and the expression and purification of hexohistidine-tagged G s␣ were performed as described (10). G s␣ was activated by 30 M AlCl 3 and 10 mM NaF, and adenylyl cyclase assays were performed at 30°C for 20 min (5,12).
Molecular Modeling of the Interaction between G s␣ and Mammalian Adenylyl Cyclase-The G s␣ structure was modeled using the sequence alignment and homology-modeling program LOOK version 2.0 (Molecular Applications Group) based on its sequence homology to GTP␥Sbound forms of bovine G protein transducin ␣ (13). The same protocol was tested by modeling the structure of G i␣ , which resulted in a model closely agreeing with GTP␥S-bound G i␣ structure (the root mean square deviation of the C␣ atoms was found to be 1.17 Å) (14). A C 1 C 2 heterodimer was modeled based on the structure of (IIC 2 ) 2 ⅐forskolin 2 (15). G s␣ was docked onto the C 1 C 2 heterodimer using program O (16) and data from the mutational analysis of G s␣ and C 1 C 2 soluble adenylyl cyclase (Ref. 17 and this paper).

Amino Acids in the IIC 2 -␣2 Region of IIC 2 Important in G s␣
Activation-We use the sequence comparison to guide the mutagenic mapping of the G s␣ binding site (Fig. 1). The IIC 2 , but not the IC 1 , protein has weak G s␣ -and forskolin-stimulated activity (ϳ1000-fold less than mixed IC 1 and IIC 2 proteins) (18). Thus, the C 2 domain must include amino acid residues that contribute to binding and partial activation by G s␣ and forskolin. Some of these residues are expected to be conserved among the C 2 domain of mammalian and fly adenylyl cyclases but might not be conserved among the C 1 domain of mammalian and fly adenylyl cyclases and the cyclase domains of membrane-bound guanylyl cyclases. Fourteen IIC 2 mutants (to either alanine or leucine) at the 13 residues that fit this criterion were constructed, and all of them had relatively normal expression based on immunoblot (Figs. 1 and 2 (mutants IIC 2 C911A, R913A, I919A, and D921A and 10 other mutants not shown). We then tested for G s␣ and forskolin activation using E. coli lysates containing the IIC 2 mutant proteins and wild type IC 1 * This work was supported by National Institutes of Health Grant GM53459. 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.
FIG. 1. Sequence alignment and secondary structure of the C 2 domain of types II and IX adenylyl cyclases (ACII C2 and ACIX C2), C 1 domain of type I adenylyl cyclase, the consensus sequences of C 1 and C 2 domains of mammalian and fly adenylyl cyclases (AC C1 or C2 core), and cyclase domains of membrane-bound guanylyl cyclases (Memb GC core) (21). Numbers above and below each row are amino acid residues of type II and type I adenylyl cyclases, respectively. Sequences underlined are absolutely conserved within the family of the indicated groups of cyclases, and sequences in black boxes are conserved among types I-VIII and rutabaga adenylyl cyclases but differ in type IX adenylyl cyclase. Secondary structure is based on the three-dimensional structure of the (IIC 2 ) 2 ⅐forskolin 2 model (15). The sequences that are mutated in this study are marked by asterisks. Protein sequences include adenylyl cyclases from mammalian and  (Table I). Due to the semiquantitative nature of using E. coli lysates, we graded the enzyme activity of the lysates containing IIC 2 mutants relative to that containing wild type IIC 2 as follows: near normal (ϩϩϩ, Ͼ50% of the control), moderately reduced (ϩϩ, 25-50% of the control), significantly reduced (ϩ, 5-25% of the control), and little or no activation (Ϯ, Ͻ5% of the control). We expected that mutations at the G s␣ binding site in the IIC 2 protein would cause a significant reduction in G s␣ activation but have little or no effect in forskolin activation. Only two of these IIC 2 mutants, R913A and D921A, fit these criteria. 2 To confirm that IIC 2 mutants R913A and D921A had reduced G s␣ activation and to further characterize these mutants, we purified both IIC 2 R913A and D921A to homogeneity and tested for their G s␣ -and forskolin-activated activity when mixed with purified IC 1 protein in vitro (Figs. 2 and 3; Table I). Both IIC 2 R913A and D921A had near normal enzyme activity when stimulated by forskolin, whereas they had about a 15-fold reduction in G s␣ -stimulated activity (Fig.  3, A and B; Table I). In the presence of 10 M forskolin, both IIC 2 R913A and D921A had relatively normal V max values but had significantly increased EC 50 values for G s␣ activation ( Fig. 3C and Table I).
While this research was in progress, the three-dimensional structure of the IIC 2 ⅐forskolin complex was solved, and the structure revealed that Arg-913 and Asp-921 were located on the amphipathic ␣2 helix (Fig. 1) (15). To test the effect of mutations at the conserved residues located at the hydrophilic surface of ␣2, IIC 2 mutants E910A, L914A, N916A, E917A, and D924A were constructed and tested for their activity in response to G s␣ and forskolin activation (Table I; Figs. 2 and 3). Similar to IIC 2 mutants R913A and D921A, the lysates containing IIC 2 E910A, L914A, N916A, and E917A had significantly reduced or little G s␣ activation but only moderate reduction in forskolin activation (Table I). The D924A mutation did not affect G s␣ and forskolin activation (data not shown). IIC 2 E910A, L914A, N916A, and E917A were purified to homogene-2 Five IIC 2 mutants (F898L/Y899L, I919A, F994A, Q1040A, R1059A, and K1065A) and four mutants (Y899L, C911A, T943A, and Y1054L) had no (Ͻ5% that of the control) or significantly reduced (5-25% that of the control) activations by G s␣ and by forskolin, respectively. IIC 2 F898L and S891A had near normal and moderately reduced activation by G s␣ and by forskolin, respectively.  ity and tested for their activation by G s␣ and forskolin (Figs. 2 and 3; Table I). All four mutants had about a 10-fold reduction in G s␣ activation and less than a 2-fold reduction in forskolin activation. Similar to IIC 2 R913A and D921A, all four mutants had significant increases in EC 50 values for G s␣ activation. These data indicate that six amino acid residues, Glu-910, Arg-913, Leu-914, Asn-916, Glu-917, and Asp-921, of IIC 2 are involved in G s␣ activation.
Amino Acids in the IIC 2 ␣3/␤4 Region of IIC 2 Important in G s␣ Activation-In contrast to the sensitivity of other mammalian and fly adenylyl cyclases to both G s␣ and forskolin, type IX enzyme is activated by G s␣ but not by forskolin (19). We hypothesize that the crucial residue(s) for forskolin binding is missing in the C 2 domain of type IX enzyme. Sequence comparison among the C 2 domains reveals that eight amino acid residues are absolutely conserved among type I-VIII and rutabaga adenylyl cyclases but differ in type IX enzyme (Fig.  1). Five of them (Gln-880, Ser-881, Ser-942, Ser-990, and Asn-992) have been mutated to alanine and tested for their activation by G s␣ and forskolin. 3 Fortuitously, another region that affects G s␣ activation was revealed. Lysates containing mutant IIC 2 N992A had near normal forskolin activation but a significantly reduced G s␣ activation. Lysates containing mutant IIC 2 S990A had near normal G s␣ and forskolin acti-vation; however, a consistent 2-fold higher relative percent of G s␣ activation (119 Ϯ 13%) than of forskolin activation (57 Ϯ 10%) was observed. When we tested a lysate containing the IIC 2 double mutant S990A/N992A, the G s␣ -and forskolinactivated activity was near normal, and the percent of G s␣ activation (76 Ϯ 9%) was less than that of forskolin activation (139 Ϯ 28%).
To further characterize IIC 2 S990A, N992A, and S990A/ N992A, the three mutant proteins were purified to homogeneity and tested for G s␣ and forskolin activation (Figs. 2 and 4; Table I). IIC 2 S990A was normal in forskolin activation, whereas IIC 2 N992A and S990A/N992A had only a slight reduction in forskolin activation (Table I and Fig. 4A). Interestingly, IIC 2 S990A had about 3-fold-enhanced G s␣ activation, whereas IIC 2 N992A had 4-fold-reduced G s␣ stimulation (Table  I and Fig. 4B). The G s␣ activation of double mutant IIC 2 S990A/ N992A was nearly normal, presumably due to compensation by the two mutations (Table I and Fig. 4B). When simultaneously stimulated by G s␣ and forskolin, IIC 2 N992A had a lower V max value but relatively normal EC 50 value (Table I and Fig. 4C). In contrast, IIC 2 S990A had a decrease in both EC 50 and V max values; the decrease in EC 50 could explain the apparent higher G s␣ activation when assayed only with G s␣ (Table I and Fig.  4C). Double mutant IIC 2 S990A/N992A had a near normal V max value and a slightly increased EC 50 value. The threedimensional structure of IIC 2 ⅐forskolin reveals that Ser-990 is the only residue that joins the ␣3 and ␤4 regions of IIC 2 ; thus, it might play a pivotal role in controlling the relative orientation between ␣3 and ␤4 of IIC 2 (Fig. 1). How the change from Ser-990 to Ala alters both EC 50 and V max values for G s␣ activation remains elusive. To further examine the region containing Ser-990 and Asn-992, we constructed and tested six more alanine-scanning IIC 2 point mutants in the Asn-987-Lys-995 region. Two more amino acid residues, His-989 and Phe-991 were shown to be involved in G s␣ activation. 4 Lysates containing IIC 2 H989A and F991A had a significantly reduced G s␣ activation but had near normal or moderately reduced forskolin stimulation, respectively (Table I). Similar results were observed when the purified mutant proteins were used (Table I and Fig. 4). When the mutants were stimulated by G s␣ and forskolin simultaneously, IIC 2 H989A exhibited a lower V max value but relatively normal EC 50 value. When the same assay was applied to IIC 2 F991A, a significant increase in EC 50 value was observed; due to low enzyme activity, the V max value of this mutant could not be determined. It is worth noting that two IIC 2 mutants in this region (at the ␣3/␤4 region, IIC 2 S990A and N992A) had reductions in V max values but had little increases in EC 50 values (Fig. 4C); this is in contrast to the IIC 2 ␣2 mutants that all have increased EC 50 values.
F293 at the N Terminus of IC 1 Is Important in G s␣ Activation-The C 1 and C 2 domains of mammalian adenylyl cyclase have ϳ25-50% identity, and there is a high degree of sequence conservation between dimer interface residues in C 1 and C 2 based on the interaction of the IIC 2 dimer in the (IIC 2 ) 2 ⅐forskolin 2 crystal structure (15). Thus, the interaction between C 1 and C 2 domains might be similar to that of the IIC 2 dimer in (IIC 2 ) 2 ⅐forskolin 2 crystal structure. Since the ␣2 region of the IIC 2 protein is close to the interface of the IIC 2 dimer, we asked whether G s␣ could interact with the amino acid residue(s) located at the C 1 domain near the ␣2 helix of the IIC 2 protein in order to facilitate the interaction between the C 1 and C 2 domains. The contact of the IIC 2 dimer in IIC 2 ⅐forskolin model predicts that the sequences at the proposed N terminus of IC 1 are likely candidates (Fig. 6A). Truncation analysis revealed that the IC 1 mutant, ⌬271-292, a deletion of amino acid residues 271-292, had normal G s␣ or forskolin activation (Table I). We then constructed and tested the G s␣ -and forskolinstimulated activity of four IC 1 mutants, F293A, H294A, S305A, and L307A, that have a mutation in the N-terminal region of IC 1 . Only one IC 1 mutant, IC 1 F293A, exhibited little G s␣ activation but retained a near normal forskolin stimulation when either E. coli lysate containing IC 1 F293A or purified IC 1 mutant protein (Figs. 2 and 5; Table I) were used. 5 When stimulated by G s␣ and forskolin, a significant increase in the EC 50 value of mutant IC 1 F293A was also observed (Fig. 3). We also tested the conserved amino acid residues in the putative ␤4/␤5 region, which is adjacent to the putative N terminus of IC 1 , and found that none of the mutants exhibited a preferen- 4 When a lysate containing mutant IIC 2 N987A was assayed, significant reduction in both G s␣ and forskolin activation was observed. However, purified IIC 2 N987A exhibited near normal forskolin activation but somewhat reduced G s␣ activation. Such a discrepancy is likely due to the lower expression of IIC 2 N987A (though not shown by immunoblot in Fig. 2). Lysates containing IIC 2 K998A and K995A had near normal G s␣ and forskolin activation, and lysate containing IIC 2 D993A had moderately reduced G s␣ and forskolin activation. Mutant IIC 2 F994A is described in Footnote 2. 5 Based on immunoblot, IC 1 F293A, H294A, and S305A had normal expression. IC 1 H294A and S305A had significantly reduced and little G s␣ -and forskolin-stimulated activity, respectively. IC 1 L307A had more than a 10-fold reduction on expression based on immunoblot and also had no detectable enzyme activity even when G s␣ and forskolin was used.
FIG. 6. A, schematic of C 1 (green)C 2 (white) model based on (IIC 2 ) 2 ⅐forskolin 2 crystal structure illustrating the binding sites for G s␣ (red) and G␤␥ (yellow) (24). Forskolin is shown with white bonds where it binds at either end of the active site cleft. B, space-filling model of G s␣ ⅐adenylyl cyclase complex. The upper panel shows the docked view, and the lower panel shows two molecules rotated by 90 degrees. Residues of both G s␣ (17) and adenylyl cyclase (this paper) implicated in binding are highlighted in red. The structure of G s␣ was modeled based on the crystal structure of GTP␥S-bound G protein transducin ␣ (13). The membrane surface is based on the crystal structure of the complexes of both G i/t␣ chimera⅐G␤␥ and G i␣ ⅐G␤␥, and it is at least 28 Å away from plasma membrane (25,26). The structure of the C 1 C 2 heterodimer was modeled on the basis of the crystal structure of (IIC 2 ) 2 ⅐forskolin 2 (15). N-C 1 , N terminus of C 1 . C, surface representation of adenylyl cyclase and G s␣ . The coloring is according to electrostatic potential using GRASP (27) and contoured in the range from Ϫ5kT (red) to ϩ5kT (blue). The figure shows the complementary nature of the interaction surfaces, negative on the adenylyl cyclase and positive on G s␣ . The orientation matches the lower panel of B. tial reduction in G s␣ activation. 6 These data indicate that the conserved Phe-293 at the C 1 domain is crucial in G s␣ activation. In addition, the data provide support for the idea that the structure of the C 2 dimer is valid in examining the structure of the C 1 C 2 heterodimer. DISCUSSION The mutagenesis based on the sequence comparison of adenylyl and guanylyl cyclases and the molecular structure of IIC 2 has revealed that 10 amino acid residues (Glu-910, Arg-913, Leu-914, Asn-916, Glu-917, Asp-921, His-989, Ser-990, Phe-991, and Asn-992) within two regions (␣2 and ␣3/␤4) of IIC 2 are essential for G s␣ activation. Although these two regions of IIC 2 proteins are 68 amino acids apart in the primary sequence, they are in close proximity in the structure of the IIC 2 ⅐forskolin dimer (Fig. 6, A and B). The G s␣ binding site is separate from but close to the proposed G protein ␤␥ binding site (N terminus of ␣3) of type II adenylyl cyclase, which is consistent with ability of the G␤␥ to synergize the G s␣ activation of type II enzyme (Fig. 6A) (20 -22). The putative G s␣ binding site forms a negatively charged and hydrophobic groove 10 ϫ 10 ϫ 15 Å, capacious enough to bind an ␣ helix (Fig. 6, B and C). 7 This negatively charged groove could attract the positively charged surface formed by the putative adenylyl cyclase binding region of G s␣ (␣2/␤4 (switch 2), ␣3/␤5, and ␣4/␤6) (13,14,17) (Fig. 6, B and C). It is also worth noting that the sequences at the C 1 ␣2 region are reasonably conserved among the G i␣ -sensitive adenylyl cyclase (types I, V, and VI) but not other isoforms; thus, it may be the determinant for G i␣ binding, a site independent of that for G s␣ (23).
The complex of C 1 and C 2 domains are necessary for potent activation by G s␣ . Although how the C 1 domain interacts with the C 2 domain remain elusive, we hypothesize that their interaction is similar to the contact of the IIC 2 dimer based on the following observations: the relative high degree homology of C 1 and C 2 domains, the sequence conservation at the dimer interface of C 1 and C 2 domains based on the structure of IIC 2 dimer, and the success of the C 1 C 2 model to predict the importance of the N terminus of C 1 domain for G s␣ activation. The C 1 C 2 model was constructed using the homology modeling based on the (IIC 2 ⅐forskolin) 2 structure (Fig. 6). The model shows that the putative adenylyl cyclase binding regions of G s␣ can dock well to the negatively charged groove that is its presumed site in adenylyl cyclase (Fig. 6B); the validity of this model remains to be tested experimentally.
How does G s␣ activate adenylyl cyclase? Based on mutational analysis, we hypothesize the following events leading to the activation of adenylyl cyclase by G s␣ . The greatest effects on EC 50 values for G s␣ activation map to the C 2 ␣2 helix, suggesting that in the first step, G s␣ binds to adenylyl cyclase with an energetic driving force provided primarily by the ␣2 helix of the C 2 region. The sensitivity to mutation at Phe-293 demonstrates a potential role for G s␣ in bridging the C 1 and C 2 domains and promoting their juxtaposition in a catalytically productive manner. The observation that mutation of the ␣3/␤4 of C 2 can alter the V max for G s␣ -stimulated catalysis suggests that this region is an allosteric linker between the G s␣ binding site and the active site. Indeed, both ␣2 and ␤4 directly participate in forming the ventral cleft-containing active site. The G s␣ binding regions occur on portions of the two longest ␣ helices in the cyclase structure. The ␣2 and ␣3 helices provide 30-and 37-Ålong lever arms, respectively, such that a modest change in their mutual orientation at the G s␣ binding site could be converted scissorswise into a large change in the structure of the active site, leading to catalytic activation. The determination of the molecular structure of C 1 C 2 and G s␣ C 1 C 2 is in progress, and the solution will yield valuable insight into how G s␣ activates adenylyl cyclase. 6 Eight mutants with mutations at the putative ␤4/␤5 region of IC 1 , T403A, V406A, V410A, L411A, K415A, W416A, Q417A, and Y418A, were constructed and tested for their activation by G s␣ and forskolin. All the mutants expressed normally, based on immunoblot. Six mutants, V406A, V410A, K415A, W416A, Q417A, and Y418A, were near normal in G s␣ and forskolin activation. Mutants IC 1 L411A and T403A had significantly reduced or no stimulation by G s␣ and forskolin, respectively. 7 The C-terminal part of the ␣3 helix of a symmetry-related adenylyl cyclase molecule completely fills this groove in the crystal. The Phe-991 side chain of the symmetry-related adenylyl cyclase binds to a prominent hydrophobic patch in the center of the groove floor formed by the Leu-914 side chain and by methylene groups on Arg-913, Glu-917, and Ser-990.