Article

Top Article References

cAMP is a key player in the intracellular signaling pathways of hormones, neurotransmitters, odorants, and chemokines. By activating PKA and cyclic nucleotide-gated ion channels, this protean second messenger can change cellular attributes as diverse as the membrane potential and the rate of cell division. The key step in regulating intracellular cAMP is modulation of adenylyl cyclase activity. Adenylyl cyclase, the enzyme that synthesizes cAMP, is subject to coincident regulation by both extracellular and intracellular stimuli. After the original cloning of the mammalian adenylyl cyclase gene by Krupinski et al. (1989), at least nine isoforms of mammalian adenylyl cyclases have been cloned and analyzed, revealing diverse regulation by G proteins, Ca²⁺, and phosphorylation (reviewed in Sunahara et al., 1996, and a monograph edited by Cooper, 1997). Historically, extensive biochemical analysis has been hampered by the inability to express and purify cyclase in its native, membrane-bound form. Fortunately, the catalytic core domains of adenylyl cyclase can be used to construct a soluble enzyme that maintains sensitivity to its physiological regulators and can be expressed in useful levels in Escherichia coli (Tang and Gilman, 1995). This soluble enzyme has a turnover number (ranging from 4 to 100 sec¹) comparable to or greater than the maximal activity of the membrane-bound enzyme (Dessauer and Gilman, 1996; Whisnant et al., 1996; Yan et al., 1996; Scholich et al., 1997; Sunahara et al., 1997). It has emerged as an extraordinarily useful model for studying the biochemical mechanisms of catalysis and regulation of full-length mammalian adenylyl cyclases. In this review, we focus on how adenylyl cyclase catalyzes the formation of cAMP from ATP in response to its regulators. For the first time, this can be addressed in molecular detail using the recently solved structures of adenylyl cyclase associated with or without its stimulatory G protein and product analogs (Zhang et al., 1997b; Tesmer et al., 1997).

Catalytic Core of Adenylyl Cyclase

Extracellular signals (i.e., hormones, neurotransmitters, odorants, and autocrines) require at least three membrane components to modulate intracellular cAMP: a heptahelical receptor, a heterotrimeric G protein, and adenylyl cyclase. The main stimulatory effect of adenylyl cyclase by hormones and neurotransmitters is mediated by G_s, the  subunit of G_s protein that stimulates adenylyl cyclase. Although each of the nine isoforms (types I-IX) of mammalian adenylyl cyclases has its unique and diverse regulation (Table 1), they share a common structure: two cytoplasmic domains (C₁ and C₂), each following a transmembrane stretch that hypothetically has six -helices (M₁ and M₂) (Fig. 1). The C₁ and C₂ domains have ~230 amino acid regions (C_1a and C_2a) that share 50% or high similarity and form the catalytic core (Fig. 2). Systematic mutational analysis has shown that the amino acid residues from both C_1a and C_2a domains contribute to ATP binding and catalysis (Tang et al., 1995; Yan et al., 1997b). Truncation analysis has permitted expression and high yield purification of G_s-regulated, soluble enzymes using only C_1a and C_2a domains, from either the same or different mammalian adenylyl cyclases (Tang et al., 1995; Tang and Gilman, 1995; Dessauer and Gilman, 1996; Whisnant et al., 1996; Yan et al., 1996, 1998; Scholich et al., 1997; Sunahara et al., 1997). Such soluble enzymes also can be activated by the diterpene forskolin, derived from the root of the Indian plant Coleus forskolhii (Seamon and Daly, 1986).

View larger version (43K): [in this window] [in a new window]   Fig. 1.   Models of adenylyl cyclases. Mammalian adenylyl cyclase with indicated sites for catalysis and binding of the regulators (top) and the related adenylyl and guanylyl cyclases (bottom). Adenylyl cyclases (types I-IX): type I (Genbank accession numbers M25579, L05500), type II (M80550), type III (M55075), type IV (M80633), type V (M88649, M83533, Z29371, M96159), type VI (M94968, M96160, L01115, M93422), type VII (U12919, Z49806), type VIII (L26986, Z35309), type IX (Z50190), XlAC: Xenopus laevis adenylyl cyclase (Z46958), DmAC: Drosophila melanogaster adenylyl cyclase (rutabaga: M81887, DAC9:AF005630), DdACA: Dictyostelium discoideum adenylyl cyclase involved in aggregation (M87279), MtAC: Mycobacteria tuberculosis adenylyl cyclase (AF017731), TbESAG: Trypanosoma brucei expression site-associated gene (X52118, X52120, X52121), LdRAC: Leishmania donovani receptor-adenylyl cyclase (U17042, U17043), DdACG: Dictyostelium discoideum adenylyl cyclase that is involved in germination (M87278), yeast AC from Schizosaccharomyces pombe (M24942) and Saccharomyces cerevisiae (M12057); BlAC: Brevibacterium liquefaciens adenylyl cyclase (X57541), RmAC: Rhizobium meliloti adenylyl cyclase (M35096), Tobacco axi AC: adenylyl cyclase from Tobacco axi (AF026389), GC-A-G: mammalian guanylyl cyclases GC-A (J05677), GC-B (M26896), GC-C (M55636), GC-D (L37203), GC-E (L36029), GC-F (L36030), and GC-G (AF024622), DmGC: Drosophila melanogaster guanylyl cyclase (X72800, L35598), SpGC: sea urchin guanylyl cyclase (M22444), CeGC: Caenorhabditis elegans guanylyl cyclase (Yu et al., 1997), sGC-: soluble guanylyl cyclase  and  subunits (M57405, X63282, U27117 for  and M22562, M57507, U27123 for ).

View larger version (76K): [in this window] [in a new window]   Fig. 2.   Alignment of representative adenylyl and guanylyl cyclase sequences. Roman numerals, mammalian adenylyl cyclase types I, II, V, and IX from cow, rat, dog, and mouse, respectively, with accession numbers M25579, M80550, M88649, and U30602. sGC and sGC are bovine soluble guanylyl cyclase 1 and 1 subunits, accession numbers X54014 and Y00770. ACG and rGC1 are the germination-specific adenylyl cyclase of D. discoideum, and the photoreceptor specific membrane guanylyl cyclase-1, accession numbers M87278 and S74247. Secondary structures are shown for VC1 (top) and IIC2 (bottom). Unshaded boxes, residues that form hydrophobic pockets. Dark letters in lightly shaded boxes, residues with C1-like roles. White letters in darkly shaded boxes, residues with C2-like roles.

The tertiary structure of the cyclase domain has been elucidated by the structure of a nearly inactive C_2a homodimer of type II adenylyl cyclase (IIC₂) (Zhang et al., 1997a, 1997b) and the fully active, heterodimeric combination of type V C_1a and type II C_2a (Tesmer et al., 1997). As expected from the high homology between C_1a and C_2a, their tertiary structures are almost identical, consisting of a three-layer / sandwich with the C_1a and C_2a domains arranged in head-to-tail fashion as a wreath (Fig. 3, A and B). The cyclase domain contains a substructure that resembles the palm domains of the polymerase I family of prokaryotic DNA polymerase, including E. coli DNA polymerase I and Thermus aquaticus (Taq) polymerase. The interface of C_1a and C_2a domains can accommodate two potential ATP binding sites (Fig. 3, A and B). Only one of the sites consists of crucial residues for catalysis and binds substrate, whereas the other is the site for the binding of forskolin (Dessauer et al., 1997; Liu et al., 1997; Tesmer et al., 1997; Yan et al., 1997b, 1998).

View larger version (115K): [in this window] [in a new window]   Fig. 3.   Adenylyl cyclase structures. Green and red, C1 and C2 domains, respectively. Atoms are shown by small spheres colored white (carbon), blue (nitrogen), red (oxygen), yellow (sulfur), green (phosphorous), and purple (magnesium). A, Forskolin, but not G protein, activated conformation of the catalytic core of type I adenylyl cyclase. The C1 and C2 domains are modeled based on homology to IIC2 homodimer crystal structure (Zhang et al., 1997b; Liu et al., 1997). ATP has been docked into the active site in the anticonformation observed in the P-site inhibitor complex (Tesmer et al., 1997), and two Mg2+ ions are shown as observed for T7 polymerase (Doublie et al., 1998). B, Crystal structure of VC1/IIC2 heterodimer in the Gs-activated conformation (Tesmer et al., 1997). Gs is omitted for clarity. MP-Forskolin, 7-desacetyl,7-(O-N-methylpiperazino)--butyryl forskolin. The active-site lid region that is ordered in this structure but not in the IIC2 homodimer is marked. C, Crystal structure of Gs complexed with the VC1/IIC2 heterodimer. Curved arrow at top, proposed direction of the Gs-induced rotation of C1 relative to C2. The Gs binding site of adenylyl cyclase as defined by scanning mutagenesis (Yan et al., 1997a) is shown in white. D, Detail of forskolin binding site on the type I heterodimer model. E, Detail of Mg and ATP binding sites on the type I adenylyl cyclase heterodimer model. ?, A Mg2+ ion that has not been observed crystallographically but is modeled by analogy to DNA polymerase.

The wreathlike dimer arrangement is likely to exist in other enzymes that perform the same or similar reaction (Fig. 1; reviewed in Tang et al., 1997). These include many homologous adenylyl cyclases and guanylyl cyclases (enzymes that convert GTP to cGMP). An integral membrane adenylyl cyclase that has six putative transmembrane-helices is found in Mycobacteria (MtAC). One transmembrane-helix form of adenylyl cyclases occurs in parasites (TbESAG and LdRAC) and in slime mold (DdACG). The family of membrane-bound receptor guanylyl cyclase from mammals (GC A-G), fruit fly (DmGC), sea urchin (SpGC), and nematode (CeGC) is growing rapidly (Yu et al., 1997, and references therein). There are peripheral membrane adenylyl cyclases from yeast and soluble adenylyl cyclases from bacteria and plant (BlAC, RmAC, Tobacco axi). Although not determined experimentally, cyclases with the homodimeric cyclase domain are predicted to have two NTP binding sites, consistent with positive cooperativity of membrane-bound guanylyl cyclase with respect to GTP (Wedel et al., 1997; Liu et al., 1997). NO-activated soluble guanylyl cyclases consist of  and  subunits that are homologous to C_1a and C_2a of mammalian adenylyl cyclase (Hobbs, 1997). The C_1a/C_2a structure is unlikely to apply to the energy state-regulated adenylyl cyclases from Enterobacteria and the calmodulin-sensitive adenylyl cyclase/toxins from Bacillus anthracis and Bordetella pertussis because no sequence similarity can be detected among them (Tang et al., 1997, and references therein).

Structure of the G_s and Forskolin Binding Sites of Mammalian Adenylyl Cyclases

Many neurotransmitters and hormones (i.e., epinephrine, dopamine, and luteinizing hormone) activate G_s-coupled receptors, leading to stimulation of mammalian adenylyl cyclase by GTP-G_s at a picomolar concentration. Although G_s is palmitoylated, the high affinity of G_s for adenylyl cyclase is likely to depend on another unidentified lipid modification (Linder et al., 1993; Kleuss and Gilman, 1997). The G_s binding site of adenylyl cyclase has been localized to a small region of C_1a (amino terminus) and a much larger negatively charged and hydrophobic groove on C_2a (2 and 3/4) by mutagenic mapping (Yan et al., 1997a). The G_s binding site is ~30 Å away from the catalytic site (Fig. 3C) (Yan et al., 1997a; Tesmer et al., 1997). The adenylyl cyclase binding site of G_s also has been mapped to switch 2 (2) and the loop regions of 3/5 and 4/6; only switch 2 region changes conformation on the binding of GTP (Berlot and Bourne, 1992; Lambright et al., 1994).

The crystal structure of VC₁/IIC₂/G_s reveals the molecular nature of the G_s/adenylyl cyclase interaction (Tesmer et al., 1997). The interface between adenylyl cyclase and G_s is substantial (1800 Å²), and switch 2 is a major component of the adenylyl cyclase binding site of G_s (Tesmer et al., 1997). The main contact between C₁ and G_s is the hydrophobic interaction between AC1 F293 and G_s W281 (Tesmer et al., 1997; Yan et al., 1997a). The C₂/G_s interaction is both polar (negative in C₂ and positive in G_s) and hydrophobic (Tesmer et al., 1997; Yan et al., 1997a). Kinetic analysis has suggested more than one G_s binding sites based on the biphasic activation of type VI enzyme by G_s, and the low affinity site of type VI enzyme seems to be blocked by a peptide from the C_1b region (Chen et al., 1997). This putative G_s binding site at the C_1b region may play a role in interacting with the region that is not involved in the contact between C_1a/C_2a and G_s in the VC₁/IIC₂/G_s structure (i.e., 4/6, which is 10 Å away from C_1a/C_2a; Tesmer et al., 1997).

Forskolin, a hypotensive diterpene, has been profoundly useful as a pharmacological probe for adenylyl cyclase function in vivo and in vitro (Seamon and Daly, 1986). Forskolin binds to only one site on the C₁/C₂ heterodimer, which is virtually identical to the two nearly symmetric sites on the IIC₂ homodimer (Dessauer et al., 1997a; Tesmer et al., 1997; Zhang et al., 1997b). Interactions between forskolin and adenylyl cyclase are predominantly hydrophobic, but specificity is enhanced by hydrogen bonds between O1 and O11 and C₁ and between the 7-acetyl and a serine at the 2'-3' turn on C₂ (927 and 942 on type I and II, respectively) (Fig. 3D). This is supported by structure-activity studies showing that 1-OH and 9-OH groups of forskolin and interaction between 7-acetyl groups of forskolin and S942 of IIC₂ are required for optimal activation of adenylyl cyclase by forskolin (Robbins et al., 1996; Yan et al., 1998).

One major surprise in the structure of adenylyl cyclase is the existence of a highly conserved hydrophobic pocket at the interface of C_1a/C_2a. Judged by primary sequence, this pocket in mammalian type IX enzyme is different from the other eight types of mammalian adenylyl cyclases, and its homolog in fruit fly is such that it cannot be activated by forskolin (Premont, 1992; Iourgenko et al., 1997). A single mutation (tyrosine to leucine) of mammalian type IX enzyme can confer both binding and activation by forskolin (Yan et al., 1998). The sterically occluded nature of the binding site suggests that bulky molecules, such as proteins, are unlikely to bind to this site. Therefore, if a physiologically relevant activator of adenylyl cyclase exists that is directed to the forskolin binding site, it is likely to be a small molecule rather than a protein. Seamon and coworkers tested candidate molecules, but as yet no such small molecule has been identified (Laurenza et al., 1989). Perhaps newly available genetic tools will facilitate a renewed search (Yan et al., 1998). An analogous compound could modulate the soluble guanylyl cyclase that is predicted to have only one GTP binding site and a deep hydrophobic pocket similar to the forskolin binding site of C_1a/C_2a. There is no report of forskolin activation of soluble guanylyl cyclase, and the key forskolin binding residues are absent in the pocket of soluble guanylyl cyclase.

The C₁/C₂ heterodimer counterpart of the second C₂ homodimer forskolin site seems to be incapable of binding forskolin because of a replacement of the critical serine by an aspartic acid (Fig. 3E, ACI D354). The aspartic acid precludes forskolin binding because it overlaps sterically with the 7-acetyl group. This aspartic acid in the C1 domain is critical for catalysis; therefore, the residue at this position probably is the most important difference between C₁ and C₂ domain sequences. Rotating the C₁-C₂ heterodimer about the pseudo-2-fold axis relating the domains reveals that ATP and forskolin make many analogous interactions (Fig. 3, D and E). The forskolin O1 and adenine N6 both donate hydrogen bonds to an aspartic acid at the same position on 5 or 5', respectively. This suggests that the forskolin binding site evolved from one of two active sites in an ancestral adenylyl cyclase. Although the hypothetical endogenous counterpart of forskolin is, in a physiological sense, unknown, it seems that in a structural and evolutionary sense, its closest counterpart is ATP.

Regulation of Mammalian Adenylyl Cyclases by G Proteins, Ca²⁺ Signals, and Phosphorylation

In addition to their regulation by G_s and forskolin, mammalian adenylyl cyclases are subjected to complex regulation by other G proteins, Ca²⁺ signals, and phosphorylation (Table 1). One of the major advances in understanding cAMP-mediated signal transduction is in redefining the action of the members of the G_i family (G_i, G_o, G_z) that can be activated by diverse hormones and neurotransmitters (i.e., adenosine, epinephrine, and cannabinoid). On activation, the G_i family of proteins releases two potential regulators, G and G; this process can be blocked by treatment with pertussis toxin. Traditionally, the activation by the G_i family is viewed as producing only inhibition of adenylyl cyclase activity. Such inhibition does occur and can be mediated through the  subunit of G_i, G_o, or G_z or through G (Tang and Gilman, 1991; Taussig et al., 1994; Kozasa and Gilman, 1995). For certain subtypes of adenylyl cyclases, such as type V and VI enzymes (predominantly expressed in heart), enzyme activity is inhibited by G_i and not altered by G (Taussig et al., 1994). For type I enzyme, which is predominantly expressed in brain, where G_o makes up to 1% of total membrane proteins, the activity of type I enzyme on stimulation of G_o-coupled receptor is inhibited predominantly by G, and the  subunit of G_o facilitates this inhibition (Tang and Gilman, 1991; Taussig et al., 1994). Characterizing the response from different isoform of adenylyl cyclase has led to the conclusion that the activation of G_i/G_o-coupled receptors also can lead to the activation, rather than inhibition, of adenylyl cyclase (Gao and Gilman, 1991; Tang and Gilman, 1991; Yoshimura et al., 1996). In brain and other tissues that express type II, IV, and VII enzymes, the activation of G_i/G_o releases G, which potentiates the enzymatic activity when the enzymes are simultaneously activated by G_s.

Where and how do G and G_i bind and regulate adenylyl cyclase activity? The G binding site of type II adenylyl cyclase has been mapped to the C₂-3 region (Chen et al., 1995). G is likely to modulate type II enzyme by binding on only one domain (C_2a) to affect its conformation, indirectly promoting optimal alignment at the catalytic site (Zhang et al., 1997b). G must bind a different site of type I enzyme because the 3 region of type I enzyme is not conserved with that of type II enzyme and peptides from this region do not block a G-mediated response (Chen et al., 1995). The action of G_i is not competitive with the binding of G_s (Taussig et al., 1994). Because the major contacts between G_s and adenylyl cyclases (switch 2) also are conserved in G_i-based VC₁/2C₂/G_s structure, other regions, such as 3/5 of G_s, may play an crucial role in how G_s achieves the specificity over G_i (Tesmer et al., 1997; Skiba and Hamm, 1998). The binding site for  subunit of G_i, G_o, or G_z has not been determined, but it is speculated to bind the 2/3 region of C_1a on the opposite site to G_s binding site (Yan et al., 1997a; Tesmer et al., 1997). If this is the case, the binding site is close to the catalytic site and the binding of G proteins could disturb the optimal alignment, perhaps by blocking the "counterclockwise" rotation of C₁ ( Fig. 3C).

Bioactive agonists also can activate G_q-coupled receptors, leading to PKC activation, which in turn modulate adenylyl cyclases in an isotype-specific manner. In cells that have an overexpressed individual isoform of adenylyl cyclase, phorbol esters activate most of the adenylyl cyclases, including type II, III, V, and VII adenylyl cyclases (Yoshimura and Cooper, 1993; Jacobowitz and Iyengar, 1994; Kawabe et al., 1994). Both type II and V adenylyl cyclases are phosphorylated by PKC (Kawabe et al., 1994; Zimmermann and Taussig, 1996). In vivo analysis using chimeric type I/II enzymes and mutational analysis has mapped a region on C_2a (1034-1068) for PKC response, likely Thr1057 (Levin and Reed, 1995; Bol et al., 1997). Thr1057 is close to the carboxyl-terminal lid that is disordered in IIC₂, but it forms a catalytic lid over the active site in the heterodimer. The ability of this C₂ segment to form the catalytic lid could be altered by PKC phosphorylation of adenylyl cyclase (Fig. 3B).

Intracellular Ca²⁺ modulates diverse physiological responses; thus, it is not surprising that a Ca²⁺ signal regulates the enzyme activity of adenylyl cyclases and that cAMP also modulates Ca²⁺ release (Cooper et al., 1995). Activation of calmodulin by Ca²⁺ can activate type I and VIII enzymes directly and inhibit type III and type IX enzymes via calcium-dependent calmodulin kinase II and calcineurin/PP2B, respectively (Tang et al., 1991; Cali et al., 1994; Wei et al., 1996; Antoni et al., 1995). An amphipathic region at the C_1b region of type I enzyme has been demonstrated to be involved in calmodulin binding and activation (Vorherr et al., 1993; Wu et al., 1993). Currently, the precise mechanism of calmodulin activation is unknown. At physiological submicromolar levels, elevated intracellular Ca²⁺ concentration inhibits type V and VI enzyme activity. This inhibition is more profound with entry from extracellular Ca²⁺ than the release of Ca²⁺ from the internal stores (Boyajian et al., 1991; Yoshimura and Cooper, 1992; Cooper et al., 1994).Whether Ca²⁺ modulates type V and VI enzymes directly or via a Ca²⁺ binding protein remains elusive.

cAMP-mediated signaling is subject to desensitization after receptor activation. Substantial advances have been made in understanding desensitization at the receptor level, which involves G protein-coupled receptor kinase/arrestin, PKA, and receptor sequestration (Freedman, 1996, and references therein). Such desensitization also occurs at the level of adenylyl cyclase exemplified by the study of type VI enzyme. Type VI enzyme can be desensitized via direct phosphorylation by PKA or PKC. PKA phosphorylation of type VI enzyme results in a 50% reduction in G_s-stimulated activity (Chen et al., 1997). The phosphorylation site probably is Ser674 of the C_1b region because mutation of this site blocks the sensitivity of type VI enzyme to PKA (Chen et al., 1997). This mechanism could explain the cAMP-dependent desensitization of glucagon stimulation in hepatocytes, but it does not occur in desensitization of adenosine receptor 2a in PC12 cells (Premont et al., 1992; Lai et al., 1997). In PC12 cells, the predominant adenylyl cyclase is Ca²⁺-inhibited type VI enzyme, and adenylyl cyclase activity can be desensitized on the stimulation of A2a receptor, a G_s-coupled receptor (Chern et al., 1995). Interestingly, the desensitization of type VI enzyme in PC12 cells is cAMP independent and can be blocked by treatment with pertussis toxin but not by a selective PKA inhibitor, H89 (Lai et al., 1997). The desensitization in PC12 cells is mediated by a novel PKC (Ca²⁺ independent) that phosphorylates and inhibits type VI enzyme (Lai et al., 1997).

Catalytic Mechanism

Adenylyl cyclase converts ATP to cAMP and pyrophosphate without a covalently enzyme-bound intermediate (Eckstein et al., 1981; Dessauer and Gilman, 1996). The turnover number of purified membrane-bound and soluble adenylyl cyclases ranges from 1 to 100 sec¹ and K_m value for ATP of 30-400 µM. The forward reaction is sequential and bireactant in Mg²⁺-ATP and free Mg²⁺ (Garbers and Johnson, 1975; Somkuti et al., 1982). The mammalian adenylyl cyclase reaction proceeds by the inversion of configuration at the -phosphate, consistent with a direct in-line displacement of pyrophosphate by attack of the 3'-OH on the -phosphate (Eckstein et al., 1981). The same inversion of configuration has been found for the homologous guanylyl cyclases (Koch et al., 1990). Product release is random, with a preference for cAMP to dissociate first (Dessauer et al., 1997).

The three available molecular structures of adenylyl cyclase indicate that cyclase exists in at least three conformational states: (1) catalytically inactive IIC₂, (2) G_s and forskolin bound IIC₂ complexed with VC₁, and (3) G_s and forskolin bound IIC₂ complexed with VC₁, adenosine analog, and pyrophosphate. These represent at least approximately ground state (E), substrate-free activated state (E*), and substrate/product bound state (E**) (Tesmer et al., 1997; Zhang et al., 1997b). The enzyme cycle is likely to proceed as follows (Fig. 4): the catalytic region of adenylyl cyclase (C₁/C₂ interface) undergoes a conformational transition (E to E*) that is promoted by activators [e.g., G_s, forskolin, G protein (type II adenylyl cyclase) and Ca²⁺-calmodulin] or is blocked by inhibitors [e.g., G_i, G protein (type I adenylyl cyclase), free Ca²⁺]. Although with ~10-fold lower affinity than for ATP, adenylyl cyclase does bind GTP, but it does not use GTP as a substrate. Thus, binding of substrate (ATP) must induce a further conformational change (E* to E**), which enables the enzyme to confirm its substrate (proofreading) and proceed through catalysis. After catalysis, the E* state could reform either before or after release of the product. Further structural characterization of the conformational changes on activation and during the enzyme reaction cycle will be critically important.

View larger version (11K): [in this window] [in a new window]   Fig. 4.   Model of the enzyme cycle of adenylyl cyclase. See text for description. A, Adenosine analogs (P-site inhibitors). Less favored pathways for the release of PPi before the release of cAMP of E** to E* exist and are not shown for simplicity.

A class of adenosine analogs known as P-site inhibitors (due to the importance of an intact purine ring) has been used to probe the kinetic mechanism and active site structure of adenylyl cyclase (Johnson et al., 1989). The potency of P-site inhibitors increases additively with 2'- or 5'-deoxy and 3'-phosphoryl (PPP > PP > P) moieties (Desaubry et al., 1996). A point mutation at Lys923 of type I enzyme, a residue that is crucial in discrimination of ATP over GTP, affects the affinity for both ATP and P-site inhibitors (Tang et al., 1995). P-site inhibitors compete with the binding of nonhydrolyzable ATP and cAMP; the reaction product, PPi, enhances the binding of P-site inhibitors (Dessauer et al., 1997a). Thus, P-site inhibitors seem to bind the cAMP site with high affinity. In steady state kinetic analyses, P-site inhibitors behave noncompetitively or noncompetitively with respect to MgATP. This apparent paradox can be resolved if P-site inhibitors and reaction products (cAMP and PPi) bind to a different adenylyl cyclase conformation (E**) than does ATP (E*) (Fig. 4) (Tesmer et al., 1997).

Structure of the Active Site and Implication for Regulation

Catalysis requires at least three steps. ATP must bind in a conformation in which the 3'-hydroxyl closely approaches the -phosphorous. The 3' OH of the ATP ribose must be activated for the attack on the 3' oxygen atom of the -phosphate. The transition state for the displacement of pyrophosphate by the 3' oxygen atom must somehow be stabilized. Although the structure of a complex between adenylyl cyclase and ATP analog is not yet available, the C_1a/C_2a structure with P-site inhibitor/pyrophosphate complex and mutational data have provided an illuminating starting point for understanding the active site (Tang et al., 1995; Liu et al., 1997; Tesmer et al., 1997; Yan et al., 1997b) (Fig. 3D).

The active site of mammalian adenylyl cyclase is located in a deep cleft formed at the domain interface. It is composed primarily of 1, 1, the 2-3 turn of the C_1a domain, and 4' and 5' of the C_2a domain. Adenine binds in a hydrophobic pocket corresponding to part of one of the two forskolin sites on the IIC₂ homodimer (Liu et al., 1997; Tesmer et al., 1997). Phosphate groups interact with an arginine side chain from each side of the cleft (Arg398 and Arg1011, AC1). The invariant Asn1025 (AC2) that is crucial for catalysis forms a water-mediated interaction with the adenine in the structure of VC₁/IIC₂/G_s/P-site inhibitor complex. The carboxyl-terminal region (AC2 1058-1071) forms a lid that is disordered in the IIC₂ homodimer but becomes ordered in VC₁/IIC₂/G_s complex (Tesmer et al., 1997; Zhang et al., 1997b) (Fig. 3, A and B). Lys1067 (AC2) from this carboxyl-terminal "lid" is involved in binding ATP based on photoaffinity labeling and mutagenesis (Droste, 1996; Tang et al., 1995). Hydrogen bonds between the adenine N6 and Asp1018 (AC2 numbering); and N1 and Lys938 (IIC₂) seem to confer specificity for adenine (Liu et al., 1997; Tesmer et al., 1997). In guanylyl cyclases, these residues are replaced by a cysteine and a glutamic acid, respectively. Mutations of these two residues at retinal-specific guanylyl cyclase (ret GC-1) converts this enzyme into a highly active and fully regulated adenylyl cyclase (Tucker et al., 1998).

A single Mg²⁺ ion binds to pair of aspartates on C₁ (Tesmer et al., 1997). The pair of aspartates is present in the overlaid structure of the palm domain of DNA polymerase I. The common Mg²⁺ binding site substantiates a biochemically relevant similarity between the two. There are almost certainly two, not one, Mg²⁺ ions in the ATP complex based on kinetics (Garbers and Johnson, 1975). Recent high-resolution structural analyses of DNA polymerase  and DNA polymerases from T7 phage and Bacillus stearothermophilus reveal that the aspartic acid pair on polymerases is capable of binding two Mg²⁺ ions (Sawaya et al., 1997; Doublie et al., 1998; Kiefer et al., 1998; Steitz, 1998). Polymerase Mg²⁺ ion B corresponds to the site observed in the P-site complex, except that it binds all three phosphate groups in a tridentate arrangement (Sawaya et al., 1997; Doublie et al., 1998; Steitz, 1998). Polymerase metal ion A is thought to be responsible for catalysis because it interacts with the -phosphate and the model-built 3'-OH (Doublie et al., 1998, Kiefer et al., 1998). The site corresponding to metal ion A on adenylyl cyclase is sterically compatible with Mg²⁺ ion occupancy and seems to be the most likely site for the catalytic metal ion.

The four residues most clearly implicated in catalysis by mutational studies are closely juxtaposed in the vicinity of the metal ion and the reactive ribosyl and -phosphate groups. Asn1025 and Arg1029 (Fig. 3D, AC1 1007 and 1011, respectively) are directly involved in catalysis based on mutagenesis and kinetic analysis (Yan et al., 1997b). The Arg1029 guanidino group is poised in both VC_I/IIC₂/G_s/forskolin/P-site inhibitor crystal structure and IC₁/IC₂/ATP models to interact with the -phosphate (Liu et al., 1997; Tesmer et al., 1997). The best model for the ATP/cyclase complex shows the arginine interacting primarily with the bridging rather than nonbridging oxygen atoms, although the details remain to be established (Tesmer et al., 1997). Asn1025 is close to the -phosphate bridging oxygen atoms in current ATP complex models and may assist Arg1029 in stabilizing the transition state or the leaving group.

The identity of the catalytic base in the reaction is unresolved. Base catalysis could involve either substrate-assisted catalysis (Schweins et al., 1995) or a protein side chain. If any protein side chain acts as a catalytic base, it is most likely Asp354 (Liu et al., 1997; Tesmer et al., 1997). Mutation of Asp354 in AC1 (Tang et al., 1995) and its counterparts in guanylyl cyclases (Yuen et al., 1994; Thompson and Garbers, 1995) leads to enzymes that are completely inactive within the detection limits of the assays used, typically ~10³ of wild-type enzyme. The counterpart of this aspartic acid in B. Stearothermophilus DNA polymerase forms a hydrogen bond to the model-built 3'-end of the primer (Kiefer et al., 1998), consistent with a role in deprotonating the 3'-OH. Mutation of the adjacent Asp310 reduces forskolin-stimulated activity by 900-fold (Liu et al., 1997). The mutant phenotypes are consistent with a polymerase-like two-metal mechanism in which one metal activates the 3'-OH as well as coordinating the -phosphate (Sawaya et al., 1997; Doublie et al., 1998; Steitz, 1998).

The finding that the C₁ and C₂ domains juxtapose four critical catalytic residues, two each from C₁ and C₂, provides the first opportunity to understand regulation of adenylyl cyclase in atomic detail. The precise alignment of C₁ and C₂ seems to be crucial to simultaneously enable metal and nucleotide binding and transition state stabilization. The physical association of C₁ and C₂ is an essential precondition for formation of the catalytic site, although it is not sufficient for activated catalysis by itself. In the absence of activators, soluble C₁ and C₂ domains have approximately micromolar affinity for one another. Their affinity for each other is greatly increased by the presence of forskolin, G_s, or both (Whisnant et al., 1996; Yan et al., 1996). Forskolin acts as hydrophobic glue to attach the two domains, stabilizing the dimer by plugging a hydrophobic crevice (Zhang et al., 1997b). G_s also bridges the two domains by binding to both C₁ and C₂, explaining its ability to promote dimerization (Tesmer et al., 1997; Yan et al., 1997a). Because C₁ and C₂ are covalently attached to each other in intact adenylyl cyclase, the dimerization-promoting effects of forskolin and G_s are probably less important in intact adenylyl cyclase than in the soluble system. Comparison of the crystal structure of the active VC₁/IIC₂/G_s with the IIC₂ homodimer suggests a possible structural mechanism that could underlie the allosteric activation of the preassembled dimer (Tesmer et al., 1997; Zhang et al., 1997b). If the position the C₂ domain is fixed, the C₁ domain in the heterodimer is found to rotate ~7° "counterclockwise" (Tesmer et al., 1997). This leads to small structural changes, typically <2 Å, in the positions of the main chain of active site residues. Although these changes may seem small, catalysis is so sensitive to precise stereochemical details that it is not unreasonable to imagine these small structural changes can lead to large changes in catalytic rate. We do not yet know how much of the small conformational change is due to G_s binding and how much is due to intrinsic differences between the heterodimeric and homodimeric AC structures. There still is no structure of a low activity conformation of adenylyl cyclase bound to ATP. The 1-2 loop of C₂ plays a critical role in this mechanism by communicating the structural change in the G_s binding groove outward, yet this loop is poorly conserved among different mammalian adenylyl cyclase isoforms. Mutational studies of this region will be important to dissect the mechanism of allosteric linkage. The activation mechanism must be considered tentative until this information becomes available.

Future Directions

Might intact mammalian adenylyl cyclase exist as a dimer? Target-size analysis suggested that the activated adenylyl cyclase is a dimer, whereas higher order complexes among receptor, G protein, and adenylyl cyclase exist at the resting state (Rodbell et al., 1981). Hydrodynamic analysis of detergent solubilized or purified adenylyl cyclase also shows a possible monomer/dimer transition (Neer et al., 1984; Yeager et al., 1985). A functional oligomer of type I enzyme has been observed using immunoprecipitation of recombinant type I enzyme and its mutant (Tang et al., 1995). Although the dimerization is unlikely to be involved in catalysis directly, based on the mutational and structural analysis, it could be crucial for the regulation or localization of the enzyme (Tang et al., 1995). Each isoform has its distinct pattern of expression, but significant overlap in the tissue distribution of different isotypes has been observed (i.e., in hippocampus and cerebral cortex). If certain heterodimers between two different isotypes of enzyme are allowed, dimerization of adenylyl cyclase would add a new dimension to the already complex regulation of adenylyl cyclase.

The molecular structure of adenylyl cyclase has provided a framework for understanding the mechanisms of catalysis and regulation in atomic detail for the first time. The structure and biochemical analysis also have led to several intriguing questions. For example, what is the physiological role of the hydrophobic pocket of adenylyl cyclase that binds forskolin? How does the binding of the regulators or phosphorylation outside the catalytic core (C_1a/C_2a) (i.e., C_1b region) regulate adenylyl cyclase? What are the roles of transmembrane domains? The latter question has become more puzzling now that we know the soluble regions alone are capable of G_s- and forskolin-regulated catalysis. Membrane localization and coordination of C_1a/C_2a interaction are the clearcut roles for transmembrane domains, but why are there 12 when 1 or 2 would have sufficed? A role in transmembrane transport was suggested by the topological similarity to other transporters (Krupinski et al., 1989), but efforts by several workers have thus far failed to demonstrate such a function. If the accumulation of data concerning adenylyl cyclase continues with the same exponential rise seen over the past 10 years, we can hope that answers to these questions should be available in the near future.