Mammalian Membrane-bound Adenylyl Cyclases (*)
- From the Department of Pharmacology, University of Texas Southwestern Medical Center, Dallas, Texas 75235
-
↵§ New address: Dept. of Biological Chemistry, Univ. of Michigan Medical School, Ann Arbor, MI 48109.
INTRODUCTION
Demonstration that the principal control of cellular cyclic AMP concentrations lies at the level of its synthesis has focused
attention on adenylyl cyclase, the enzyme that catalyzes the conversion of intracellular ATP to cyclic AMP. The prototypical
hormone-sensitive adenylyl cyclase system is comprised of three types of plasma membrane-associated components: heptahelical,
G protein-coupled receptors for a variety of hormones, neurotransmitters, and autacoids; stimulatory and inhibitory heterotrimeric
G proteins; and the catalytic entity itself. The G proteins regulate the activity of the enzyme in response to interaction
of ligands with appropriate receptors. Each G protein contains a guanine nucleotide-binding α subunit and a complex of tightly
associated β and 
subunits. Upon activation of a G protein by an agonist-bound receptor, GDP is released from α in exchange for GTP. Binding
of GTP causes conformational changes that result in dissociation of GTP-α from β
, liberating two macromolecular complexes capable of regulation of downstream effectors.
Mammalian adenylyl cyclases are activated by the diterpene forskolin, and the development of a forskolin affinity matrix by Pfeuffer and Metzger (1) greatly facilitated purification of the enzyme. The scarcity of the protein and its considerable lability in detergent-containing solutions had impeded progress for years. However, it was apparent that there were at least two distinct forms of adenylyl cyclase that differed in their capacity to be stimulated by calmodulin (2, 3, 4, 5, 6). Using a forskolin affinity resin, Krupinski and co-workers (7) were able to purify a sufficient amount of a calmodulin-sensitive adenylyl cyclase from bovine brain to obtain partial amino acid sequence and thereby isolate cDNA clones encoding the full-length protein; this was termed the type I isoform. Application of low stringency hybridization and polymerase chain reaction techniques has now permitted isolation of six additional full-length clones (types II-VI and VIII)(8, 9, 10, 11, 12, 13, 14). However, definition of the extent of molecular diversity in this family is not yet complete. Both alternatively spliced transcripts of adenylyl cyclases (15) and partial sequences of novel isoforms (e.g. type VII) (16) are in evidence.
Adenylyl cyclases have molecular weights of roughly 120,000 (range of 1064-1248 amino acid residues) and a complex (but as
yet only deduced) topology within the membrane (Fig. 1). A short cytoplasmic amino terminus is followed by six transmembrane spans (designated M1) and a large (roughly 40 kDa) cytoplasmic domain (C1). The motif is then repeated: a second set of six transmembrane spans (M2) is followed by a second cytoplasmic domain (C2). This structure is (to date) unique for a “simple” enzyme. It is, however, highly reminiscent of the structures of certain
channels and ATP-dependent transporters, particularly the P glycoprotein and the cystic fibrosis transmembrane conductance
regulator. This relationship prompted speculation that the adenylyl cyclases might also serve as channels or transporters,
but there is as yet no evidence to support this conjecture. Of interest, an adenylyl cyclase from Paramecium does appear to be a K
channel(17); however, its structure is not yet known.
Structure of adenylyl cyclase. The predicted topology of membrane-bound adenylyl cyclases is shown (see text for discussion).
Cylinders represent membrane-spanning regions, while boldfacelines indicate regions of high amino acid similarity among all members of the family. Nomenclature is as follows: N, amino-terminal domain: M
, first set of membrane-spanning regions; C
and C
, the first large intracellular cytoplasmic domain; M
, second set of transmembrane spanning regions; and C
and C
, second large intracellular domain.
The overall amino acid sequence similarity among the different isoforms of adenylyl cyclase is roughly 50%. However, two regions
highlighted in Fig. 1(designated C
and C
) are more highly conserved (up to 93% sequence identity), and this strong relationship extends to corresponding domains of
topographically similar adenylyl cyclases from Drosophila(18) and Dictyostelium(19). Of importance, the C
and C
domains are also highly homologous to each other as well as to the catalytic domains of membrane-bound guanylyl cyclases
and domains that are found in each of the subunits of cytosolic heterodimeric guanylyl cyclases. Based on these relationships,
it is predicted that one or both of these domains of mammalian adenylyl cyclases is the site for catalysis of cyclic AMP synthesis.
Analysis of sequence relationships among the adenylyl cyclases permits some grouping. The type II and IV isoforms are clearly more related to each other than to the others; a similar relationship exists between type V and type VI. The regulatory properties of these isoforms appear to reflect their evolutionary relationship (see below). Although type I, III, and VIII adenylyl cyclases all are activated by calmodulin, they do not form a closely related group.
Thorough documentation of the cellular and subcellular localization of the different isoforms of adenylyl cyclase in mammalian tissues has been hampered by low levels of expression (generally 0.01-0.001% of membrane protein) and limited availability of high affinity, isoform-specific antibodies. Most information thus comes from analysis of patterns of mRNA expression. All isoforms of adenylyl cyclase appear to be expressed in the brain, apparently in region-specific patterns. The type I enzyme is largely restricted to the nervous system, while type III is found predominantly in olfactory neuroepithelium. There are substantial differences in patterns of expression in peripheral tissues.
Structural Correlates of Activity
Despite the similarities discussed above that lead to the conclusion that the C
and C
domains of the adenylyl cyclases represent catalytic domains, it has not been possible to detect significant enzymatic activity
after expression of either of these domains as discrete proteins. The same is true of an entire C1 or C2 domain or of M1C1 or M2C2 (halves of the molecule). However, concurrent expression of constructs encoding M1C1 and M2C2 in Sf9 cells using recombinant baculoviruses permits detection of the encoded proteins by immunoblotting and a substantial
amount of enzymatic activity that can be regulated in characteristic type-specific fashions (see below) by G protein subunits
or calmodulin(20). It is thus assumed that interaction between the C1 and C2 domains is essential for catalysis. This is consistent with the fact that neither subunit of the heterodimeric cytosolic
guanylyl cyclases is sufficient for catalysis (21) and that the membrane-bound guanylyl cyclases are oligomers(22, 23, 24). It is also notable that point mutations in either the C
or C
domains of the adenylyl cyclases can impair enzymatic activity severely (with retention of G
binding activity) and that mutations in either domain can elevate the K
for substrate. 1It is thus speculated that both domains can bind ATP. It is unknown if both domains can catalyze cyclic AMP synthesis or if
one is the dominant catalyst while the other is regulatory.
The front half of one isoform of adenylyl cyclase can be coexpressed with the back half of another, creating so-called noncovalent
chimeras. Concurrent expression of a “front half” construct of type I adenylyl cyclase, truncated to remove domain C
, with the back half of type II adenylyl cyclase, which largely lacks C
, permits assembly of a functional adenylyl cyclase that responds very well to both forskolin and activated G
. The variable C
and C
domains are thus not necessary for responses to these regulators.1 We assume that G
interacts with C
and/or C
and perhaps regulates interactions between these domains. A peptide corresponding to residues 495-522 of the C
domain of type I adenylyl cyclase binds calmodulin and inhibits calmodulin-stimulated enzymatic activity(26). Point mutations in this region also interfere with activation of adenylyl cyclase by calmodulin(27). The C
domain is thus likely to be a (or the) site of interaction of calmodulin with adenylyl cyclases that are sensitive to the
protein.
Regulation of Enzymatic Activity
Studies of the regulation of the isoforms of mammalian adenylyl cyclases reveal a wealth of common and disparate features.
All isoforms are activated by both forskolin and the GTP-bound α subunit of the stimulatory G protein Gs. All are inhibited by certain adenosine analogs termed P-site inhibitors; 2′-deoxy-3′-AMP is particularly potent. However,
all of the isoforms of adenylyl cyclase are further regulated in type-specific patterns by other inputs, particularly including
those that are dependent on Ca
or that arise from other (non-G
) G protein subunits (Fig. 2).
Patterns of regulation of adenylyl cyclase activity. PKC, protein kinase C; CAM, calmodulin; AC, adenylyl cyclase. See text for discussion.
Regulation by G Protein Subunits
Activation of heterotrimeric G proteins results in the dissociation of two regulatory moieties, the GTP-bound α subunit and
a dimer of the β and 
subunits. Regulation of adenylyl cyclases by G
has been appreciated for some time and was the basis for discovery of this G protein. However, knowledge of direct interactions
of other G protein subunits with specific isoforms of adenylyl cyclase is more recent.
Smigel (4) and Katada et al.(28) first noted inhibition of type I adenylyl cyclase activity by the G protein β
subunit complex, but in neither case was it clear that the effect was exerted directly. In fact, Katada et al.(28) attributed the inhibition to sequestration of calmodulin by β
. Interest in this phenomenon was rekindled when Tang et al.(20) noted prominent inhibition by β
of type I adenylyl cyclase expressed in Sf9 cells, and purification of the expressed protein permitted demonstration of its
direct interaction with this subunit complex(29).
Inhibition of adenylyl cyclases by β
is confined for now to the type I enzyme. Notably, however, when the effects of β
on other isoforms were tested, surprising stimulatory effects were observed with type II (25) and type IV (10) adenylyl cyclases. Of particular interest, stimulation of these enzymes by β
is highly conditional. Effects of β
alone are barely detectable, but the magnitude of stimulation of activity is substantial (5-10-fold) in the presence of G
. Stimulation of type II adenylyl cyclase by β
requires significantly higher concentrations of β
than of G
, and the source of the β
is presumed to be the Gi or Go oligomers, which are present in high concentrations in brain. The potential significance of this synergistic interaction
is discussed below.
The G proteins designated as Gis (G
, G
, and G
) were discovered as substrates for pertussis toxin and as the G protein oligomers responsible for inhibitory regulation of
adenylyl cyclase activity. However, when the effects of the resolved subunits of Gi were tested, it was difficult to observe substantial inhibition of the enzyme by the α subunits of these proteins. Nevertheless,
Wong and co-workers (30) showed that transfection of cells with cDNAs encoding constitutively activated G
subunits lowered intracellular concentrations of cyclic AMP, suggesting that these proteins could inhibit adenylyl cyclase
activity. We know, retrospectively, that the failures to reconstitute the response in vitro were due to the use of inadequate concentrations of G
purified from natural sources (or to the counteracting stimulatory effects of contaminating G
when higher concentrations of G
were used), to the use of nonmyristoylated G
expressed in Escherichia coli, and/or to tests performed with isoforms of adenylyl cyclase not susceptible to inhibition by G
. When adequate concentrations of E. coli-derived myristoylated G
were tested on type V and type VI adenylyl cyclase, prominent inhibition of G
- and forskolin-stimulated activity was observed(31, 32). The three isoforms of G
are equally potent and efficacious. The three G
proteins can also inhibit type I adenylyl cyclase, but the effect is not as prominent as that observed with β
. Furthermore, inhibition of type I adenylyl cyclase activity by G
is largely absent when the G
-stimulated activity is examined; inhibition is largely confined to activity observed in the presence of calmodulin or forskolin.
Type I adenylyl cyclase can also be inhibited by G
, while this subunit has no effect on the type V enzyme. Both type I and V adenylyl cyclases are also inhibited by G
, 2the only member of the Gi subfamily of α subunits that is not a substrate for pertussis toxin.
The subject of inhibitory regulation of type II adenylyl cyclase is problematic. We have demonstrated that this isoform (when
expressed in Sf9 cell membranes) is not inhibited by G
(either native or myristoylated recombinant protein). However, Chen and Iyengar (33) noted that expression of a constitutively activated mutant of G
inhibited type II adenylyl cyclase activity when coexpressed in COS-7 cells. In view of the observation of Tang and Gilman(25), discussed above, that β
activates type II adenylyl cyclase, we view it as unlikely that activation of heterotrimeric Gi would release two regulators (G
and β
) with opposing effects on the same enzyme. In support of this hypothesis, Federman et al.(34) noted that activation of Gi resulted in conditional activation (and not inhibition) of type II adenylyl cyclase expressed in HEK-293 cells.
Regulation by Ca
Changes in intracellular Ca
can have profound effects on cellular concentrations of cyclic AMP if appropriate isoforms of adenylyl cyclase are present.
Types I and VIII adenylyl cyclase (and type III to a lesser extent) are markedly stimulated by nanomolar concentrations of
Ca
/calmodulin, and intracellular cyclic AMP concentrations rise dramatically when transfected cells expressing these isoforms
are exposed to agonists that elevate intracellular Ca
. The other isoforms of adenylyl cyclase (II, IV, V, and VI) are insensitive to calmodulin.
All adenylyl cyclases are inhibited by high (100-1000 μM) concentrations of Ca
as a result of competition for Mg
, which is required for catalysis. By contrast, types V and VI adenylyl cyclase are inhibited by low micromolar concentrations
of Ca
(11, 13, 35). This effect is independent of calmodulin and is presumably mediated directly. Inhibition of cyclic AMP accumulation in
some intact cells has been shown to follow elevation of Ca
concentrations(36, 37, 38). In some cases this has been correlated with expression of either type V or VI adenylyl cyclase(11, 13, 35, 39).
Regulation by Phosphorylation
The cellular cyclic AMP concentrations achieved in response to exogenous regulators are highly dependent on the state of phosphorylation of components of hormone-sensitive adenylyl cyclase systems. This is particularly obvious in the case of the receptors for stimulatory and inhibitory ligands, which are desensitized and down-regulated following phosphorylation by various kinases; these especially include cyclic AMP-dependent protein kinases, protein kinase C, and a variety of receptor-specific kinases that view the agonist-bound forms of receptors as preferential substrates(40). Although there are some reports of phosphorylation of G protein subunits(41, 42), they are surprisingly few. Reagents are now becoming available to investigate phosphorylation of adenylyl cyclases.
The possibility of feedback inhibition of adenylyl cyclase activity in response to phosphorylation by cyclic AMP-dependent protein kinase is obvious, but evidence for this mechanism remains sparse. Its existence is suggested by studies of chick hepatocytes and variants of the S49 lymphoma cell line(43). These cells have several isoforms of adenylyl cyclase but share the type VI enzyme. It contains two consensus sites for phosphorylation by cyclic AMP-dependent protein kinase, one of which is conserved in the closely related type V isoform.
Potential regulation of adenylyl cyclase by protein kinase C has been explored more extensively, prompted by a wealth of confusing literature demonstrating almost every imaginable effect of phorbol esters on cellular cyclic AMP concentrations. Three reports indicate that the activity of type II adenylyl cyclase, expressed by transfection, can be augmented substantially by stimulation of protein kinase C(44, 45, 46). Nevertheless, we (and presumably others) have failed to detect phosphorylation of type II adenylyl cyclase (expressed in Sf9 cell membranes, for example) by addition of activated protein kinase C. The effect may be indirect or mediated by a specific isoform of the kinase.
Kawabe and associates (47) have phosphorylated type V adenylyl cyclase in vitro using protein kinase C and demonstrated a marked increase in enzymatic activity. The effect was specific for the α and 
isoforms of protein kinase C, suggesting cross-talk between this adenylyl cyclase and both Gq-mediated pathways (for protein kinase C-α) and growth factor tyrosine kinase pathways (for protein kinase C-
). Despite the robust nature of the response observed in vitro, only modest effects of activation of protein kinase C in vivo have been observed with type V adenylyl cyclase(44).
There is less consensus with regard to effects of protein kinase C on type I adenylyl cyclase. Again, the Ca
connection would seem to make this isoform (and type VIII) logical candidates for feedback regulation by a Ca
-activated kinase. Although some have observed enhanced forskolin- (48) or calmodulin- (44) stimulated type I adenylyl cyclase activity in response to activation of protein kinase C, others have not(45).
Patterns of Regulation and Functional Consequences
Three distinct patterns of regulation of mammalian adenylyl cyclases are evident for the type I, II, V, and VI isoforms (Fig. 2). Types III, IV, and VIII have been omitted from Fig. 2 because they have been studied less extensively. However, type IV appears to resemble type II closely, and the features of type VIII noted to date resemble those of type I.
As noted above, the only constant feature (with regard to G protein-mediated regulation) is that all membrane-bound mammalian
adenylyl cyclases discovered to date are activated by G
. One can make reasonable hypotheses for regulation of each of the well studied isoforms of adenylyl cyclase (directly or
indirectly) by both of the other two major subclasses of G proteins whose signaling mechanisms are (at least partially) understood:
the Gi subfamily and the Gq subfamily. Accepting this premise, there are then four possibilities: Gi and Gq both stimulate adenylyl cyclase activity; Gi and Gq both inhibit; Gi inhibits/Gq stimulates; Gi stimulates/Gq inhibits. Remarkably, three of these four patterns appear to be represented by the first four adenylyl cyclases to be studied
in detail. Thus, even at what will ultimately be considered to be an extremely superficial level of analysis, the system has
surely evolved to permit extensive integration and cross-talk, and adenylyl cyclases can be considered as focal points or
final common paths for convergence of the activities of a very large number of regulatory elements (particularly if receptors
are included in the counting game).
Five years ago the prevailing view was of two basic types of adenylyl cyclase: calmodulin stimulated or not, both of these
likely inhibited in a Gi-mediated fashion. These basic patterns are represented by type I (VIII) on the one hand and by types V and VI on the other.
However, each scenario has offered surprises. For type I, Ca
-mediated stimulatory signals may arise from either calmodulin or protein kinase C. Inhibition (at least in vitro) may be mediated more by β
than by G
's. This brain-specific isoform can also be inhibited by G
, which is expressed predominantly in brain. We speculate that the β
-mediated inhibitory effects arise predominantly from activation of Gi or Go because these G proteins are expressed at the highest concentrations (particularly in brain), and the effects of β
require higher protein concentrations than do those of G
or G
.
Studies of learning and memory in a number of animal models demonstrate a likely role for calmodulin-activated adenylyl cyclases.
A Ca
-stimulated adenylyl cyclase is postulated to be the locus of integration of inputs underlying the development of classical
conditioning in the marine mollusk Aplysia(49). Impaired learning is evident in the Drosophila mutant rutabaga. These flies lack calmodulin-stimulated adenylyl cyclase activity(50), and the defect has been mapped to a gene encoding a calmodulin-activated adenylyl cyclase that is homologous to the mammalian
type I and VIII isoforms(18). Activation of glutamate-gated Ca
channels within the CA1 field of the mammalian hippocampus activates a calmodulin-stimulated adenylyl cyclase(51). This has been proposed to trigger cyclic AMP-dependent protein kinases and transcription factors that contribute to the
development of long term potentiation(52). Types I and VIII adenylyl cyclase are both present in this region of the hippocampus.
For type V and type VI adenylyl cyclase, inhibition can be mediated directly by Ca
, which can arise from Gq-regulated pathways and, of course, others. The observation that Ca
-dependent isoforms of protein kinase C can phosphorylate and activate type V adenylyl cyclase in vitro suggests that Ca
may under some circumstances have a stimulatory effect on this enzyme; however, this situation has not been observed in vivo. Regulation of types V and VI adenylyl cyclase by β
has not been detected. Although these isoforms are also found in brain, types V and VI may be the dominant forms of adenylyl
cyclase in peripheral tissues, where concentrations of β
are low compared with values in brain. Regulation of types V and VI by G
has also dropped from view.
Type II adenylyl cyclase has provided the biggest surprise. This enzyme is activated by G
, β
, and (indirectly?) by protein kinase C. The effects of β
are largely dependent on concurrent activation of the enzyme by G
. This adenylyl cyclase is thus designed to detect coincidental activation of regulatory inputs. The capacity of β
to serve as an appropriate regulator of such a coincidence detector is dependent on its affinity for the enzyme (rather than
release of distinct isoforms of β
following activation of different pathways). G
activates adenylyl cyclase at picomolar concentrations, while β
is effective at nanomolar concentrations. Liberation of both G
and β
by activation of the Gs oligomer is presumed to provide insufficient β
to stimulate adenylyl cyclase. Activation by β
is thus likely dependent on liberation of the subunit from a G protein present at significantly higher concentration, such
as Gi or Go. The biochemical properties of type II adenylyl cyclase provide a gratifying explanation for a phenomenon described extensively
in the 1970s, highly synergistic stimulation of cyclic AMP accumulation in brain slices following stimulation by pairs of
neurotransmitters and/or neuromodulators, now known to work through Gs- and Gi-regulated pathways. These responses have often been observed in regions of the brain where expression of type II adenylyl
cyclase is particularly abundant (e.g. cerebellum and hippocampus)(53). Electrophysiological studies of hippocampal pyramidal cells appear to provide a specific physiological example of the significance
of conditional regulation of type II adenylyl cyclase activity(54).
This discussion of patterns of regulation of adenylyl cyclase activity and a few of their physiological consequences is necessarily
simplistic. We have ignored a number of other regulatory phenomena, either because they have been incompletely defined and/or
because there is little meaningful to say, other than to note their existence. Included among these are synergistic interactions
between a variety of regulators (e.g. G
and calmodulin, G
and forskolin), complex kinetic patterns of inhibition, activator-dependent patterns of inhibition, and the likelihood that
regulation by mechanisms such as phosphorylation will also be highly dependent on the context of concurrently acting activators
or inhibitors. Also ignored have been the effects of agents such as forskolin or P-site inhibitors, whose physiological correlates
are unknown. Compilation of the catalogue of adenylyl cyclase phenomenology is a necessary task for the immediate future.
This will require more labor than intellect and may at times be frankly boring. However, appreciation of the chemical and
structural basis of this catalogue and its interpretation, particularly with regard to the physiology of the central nervous
system, will be major challenges and accomplishments.
Footnotes
-
↵* This minireview will be reprinted in the 1995 Minireview Compendium, which will be available in December, 1995.
-
↵1 W.-J. Tang, M. Stanzel, and A. G. Gilman, submitted for publication.
-
↵2 T. Kozasa and A. G. Gilman, submitted for publication.
- © 1995 by The American Society for Biochemistry and Molecular Biology, Inc.
