Physiological roles of mammalian transmembrane adenylyl cyclase isoforms

Abstract

Adenylyl cyclases (ACs) catalyze the conversion of ATP to the ubiquitous second messenger cAMP. Mammals possess nine isoforms of transmembrane ACs, dubbed AC1–9, that serve as major effector enzymes of G protein-coupled receptors (GPCRs). The transmembrane ACs display varying expression patterns across tissues, giving the potential for them to have a wide array of physiological roles. Cells express multiple AC isoforms, implying that ACs have redundant functions. Furthermore, all transmembrane ACs are activated by Gα_s, so it was long assumed that all ACs are activated by Gα_s-coupled GPCRs. AC isoforms partition to different microdomains of the plasma membrane and form prearranged signaling complexes with specific GPCRs that contribute to cAMP signaling compartments. This compartmentation allows for a diversity of cellular and physiological responses by enabling unique signaling events to be triggered by different pools of cAMP. Isoform-specific pharmacological activators or inhibitors are lacking for most ACs, making knockdown and overexpression the primary tools for examining the physiological roles of a given isoform. Much progress has been made in understanding the physiological effects mediated through individual transmembrane ACs. GPCR-AC-cAMP signaling pathways play significant roles in regulating functions of every cell and tissue, so understanding each AC isoform’s role holds potential for uncovering new approaches for treating a vast array of pathophysiological conditions.

CLINICAL HIGHLIGHTS

The potential physiological roles of adenylyl cyclases (ACs) are enormous, making these enzymes a potential therapeutic target for diseases of every organ and system. Understanding the physiological functions regulated by each AC isoform is therefore essential for drug development. Therapeutic strategies directed toward AC will lack specificity unless a specific AC isoform is targeted.

1. INTRODUCTION

Adenylyl cyclases (ACs) are a family of enzymes that catalyze the conversion of adenosine triphosphate (ATP) into cyclic adenosine monophosphate (cAMP), a ubiquitous second messenger across the kingdoms of life (1–4). The enzymes are expressed throughout most organisms and play key roles in the diversity and complexity of cAMP signaling pathways (5). This review focuses on the mammalian transmembrane ACs, excluding the soluble mammalian adenylyl cyclase (often termed sAC or AC10). sAC is regulated by bicarbonate and calcium, not G proteins, and plays key roles in pH sensing and many specific (patho)physiological functions (6–9). Mammals express nine isoforms of transmembrane ACs, termed AC1–9. Cloning of these isoforms reveals highly conserved catalytic domains but significant sequence diversity outside these domains (3). The major domains of ACs include an intracellular NH₂ terminus, a coiled-coil helical domain that separates two clusters of six transmembrane-spanning helixes (for a total of 12 transmembrane-spanning domains) from the large intracellular C1 and C2 cytoplasmic loops (10). The C1 and C2 loops are divided into C1a and C2a subdomains that dimerize to form the catalytic site and C1b and C2b regulatory subdomains. The C1a and C2a subdomains are highly conserved across ACs (1). Divergence outside of these conserved subdomains plays roles in the signaling and regulatory diversity across ACs. AC structure has been recently reviewed (11) and is summarized in FIGURE 1.

FIGURE 1.

General structure of mammalian transmembrane adenylyl cyclases (ACs). TM1 and TM2 are transmembrane domains each consisting of 6 transmembrane α-helices. The second extracellular loop of TM2 contains N-linked glycosylation sites that target some isoforms to lipid rafts (12). C1 and C2 are intracellular cytoplasmic loops, with C1a and C2a forming the catalytic domain and forskolin binding site (in the case of AC1–8). The C1a domain is the site of Gα_i binding, whereas the C2a domain interacts with Gα_s. C1b and C2b are regulatory subdomains, and the NH₂-terminal domain is involved in many protein-protein interactions. The C1 and C2 domains are separated from each transmembrane domain by 2 coiled-coil helices (shown in yellow). These 2 helices and the C1b domain of AC5 contain the indicated mutations associated with dyskinesias (13–16). This structural model is based largely on the recent cryo-EM structure of AC9 (10).

1.1. cAMP Signal Transduction

Whereas ACs produce cAMP from ATP, G protein-coupled receptors (GPCRs), protein kinase A (PKA), A kinase anchoring proteins (AKAPs), and phosphodiesterases (PDEs) all contribute to the initiation, propagation, termination, or organization of cAMP signaling (FIGURE 2). The first step of signal generation is the activation of the G protein by an agonist-occupied GPCR. The stimulatory G protein alpha subunit, Gα_s, then activates AC enzymatic activity leading to an increase in cAMP synthesis (17). Other GPCRs preferentially activate the inhibitory G protein alpha subunit, Gα_i, which can inhibit AC activity and reduce cAMP synthesis (18).

FIGURE 2.

Schematic diagram of signaling via transmembrane adenylyl cyclases (ACs). Once activated by an agonist, a G protein-coupled receptor (GPCR) will activate coupled G proteins. Gα_s activates AC activity, which catalyzes the conversion of ATP to cAMP, whereas Gα_i inhibits AC activity. AC isoforms are also regulated by various other signals, as described in TABLE 1. Phosphodiesterases (PDEs) terminate the signal by decyclizing cAMP to 5′-AMP. cAMP activates several effector proteins, including PKA, Epac, Popeye Domain Containing (POPDC), and cyclic nucleotide-gated channels. PKA is anchored by A kinase anchoring proteins (AKAPs) to specific phosphorylation targets. PKA, Epac, cyclic nucleotide-gated channels, and POPDCs elicit various downstream physiological responses depending on the cell type.

The most ubiquitous and well-studied effector molecule of cAMP is PKA, which exists as a holoenzyme of two regulatory (cAMP binding) subunits and two catalytic (kinase) subunits (19–21). PKA catalytic subunits phosphorylate consensus sequences in downstream effector proteins to elicit cellular responses (22). Although in theory thousands of proteins can be phosphorylated by PKA, in practice most PKA action is targeted to specific effector proteins via tethering to AKAPs. AKAPs bind the regulatory subunits of PKA as well as key targets of PKA phosphorylation to enable rapid, high-fidelity coupling of the kinase to downstream effectors (23–25). The AKAP family is large and diverse and so can play the roles of cAMP signaling complex scaffolds and organizers that contribute to the unique characteristics of differentiated cells. Several AKAPs are known to bind to specific AC isoforms (26–28).

PDEs terminate the cAMP signal by decyclizing the cAMP to AMP (29). Eleven gene families and 21 individual genes for PDEs exist, and they differ in terms of substrate specificities (30). PDE4, PDE7, and PDE8 hydrolyze only cAMP, whereas PDE1, PDE2, PDE3, PDE10, and PDE11 hydrolyze both cAMP and cyclic guanosine monophosphate (cGMP). PDE5, PDE6, and PDE9 only hydrolyze cGMP. Certain PDE isoforms specifically associate with particular GPCR-AC or AC-AKAP complexes (31, 32). AKAPs bind specifically with AC isoforms as well as a whole range of other signaling proteins and possess a PKA binding motif to coordinate PKA activation of downstream effectors (28). Some AKAPs also possess a PDE binding domain, which recruits PDEs to AC/GPCR complex sites to regulate cAMP levels and limit diffusion of the cAMP signal (33).

Another important downstream effector of cAMP signaling is exchange protein activated by cAMP, or EPAC (34). EPACs are guanine-nucleotide exchange factor proteins that are directly activated by cAMP (35). EPACs were more recently discovered and are thus less studied than PKA, but EPACs are important effectors in mediating various physiological responses of cAMP signaling (36). Cyclic nucleotide-gated channels, which nonselectively gate cations in response to direct binding by cAMP or cGMP, have specific roles in regulating membrane potential in sensory neurons of the olfactory and retinal systems and are involved in the regulation of heart rate (37–41). Additional cAMP effectors, Popeye Domain Containing (POPDC) proteins, have been described more recently (42, 43). POPDC proteins have diverse and complex functions, but early studies indicate they are widely expressed and play roles in various physiological effects of cAMP signaling. POPDC proteins are known to regulate heart rate and rhythm, epithelial cell tight junction formation, trabecular meshwork contraction and intraocular pressure, cell adhesion, and vesicular fusion (44–47).

1.2. Distinguishing the Functions of AC Isoforms

Complex signaling pathways activated by cAMP and the conservation of the catalytic domain of ACs have made it difficult to understand the specific roles of individual AC isoforms. Limited by a general lack of isoform-specific inhibitors (11), studies to tease out differences among isoforms have largely relied on genetic approaches. The general expression patterns of the various AC isoforms in cells and tissues have been elucidated (48–51). Most cell types express multiple isoforms, but abundant expression of certain isoforms in particular tissues is notable (FIGURE 3). AC1 and AC8 consistently show high expression in areas of the brain such as the hippocampus, cerebellum, cortex, thalamus, and hypothalamus. AC1 also is notably expressed in the heart, kidney, skeletal muscle, pancreas, and reproductive organs (56). AC2 expression is widespread, but it shows particularly high expression in lung and brain tissue. AC3 is most concentrated in olfactory epithelia and the pancreas (1). AC4–7 show high expression in regions throughout the heart and vascular system (57). AC5 expression is high in the basal ganglia and the heart (56, 58). AC6 expression is widespread but notably high in the cerebellum, choroid plexus, and the heart (56, 58). AC9 is widespread but most concentrated in skeletal muscles, the pituitary gland, and cerebral cortex (1, 59).

FIGURE 3.

Tissue distribution of transmembrane adenylyl cyclases (ACs): the predominant AC isoforms expressed by major organs and systems in the human body. AC isoforms are color coded by group (group I in blue, group II in green, group III in red, group IV in brown). The digestive tract expresses all AC isoforms except AC2 (52). The vasculature expresses primarily AC3, AC5, and AC6 (53–55).

The widespread expression of ACs precludes classification of the isoforms by tissue type or other anatomical distribution. Instead, it is more useful to classify the isoforms by their regulatory properties (FIGURE 4). The major direct regulation of ACs is by G protein subunits and Ca²⁺ via calmodulin (CaM) (1). The Ca²⁺/CaM-stimulated isoforms are AC1 and AC8, with weak stimulation of AC3 (60, 61). Although inhibition of AC activity by high concentrations (mid micromolar) of free Ca²⁺ is seen in all isoforms, high-affinity inhibition by Ca²⁺ is observed in AC5 and AC6 (62, 63). Gβγ, Gα_i, and Gα_s subunits have stimulatory and/or inhibitory effects on specific AC isoforms. Gα_s stimulates all AC isoforms (64). Gα_i inhibits AC1, AC5, and AC6 (65, 66) and may have minor influence on the activity of other isoforms that is counteracted by the stimulatory effects of coincident activation of Gβγ. Gβγ inhibits AC1, AC3, and AC8 but conditionally stimulates AC2, AC4, AC5, AC6, and AC7 in the presence of another activator (60, 67–69). The diterpene forskolin directly activates all isoforms of mammalian transmembrane AC (70, 71) with the exception of AC9 (72, 73). An additional important isoform-specific regulatory mechanism is via AKAP-mediated targeting of protein kinase C (PKC) (74). PKC stimulates AC1, AC2, AC3, AC5, and AC7 but inhibits AC4 and AC6, with the impact on AC8 unknown (75–78). Overall, this pattern of regulation categorizes the isoforms into four main groups. Group I consists of AC1, AC3, and AC8, which are all Ca²⁺ stimulated. Group II consists of Gβγ-stimulated AC2, AC4, and AC7. Group III is Gα_i-inhibited and Ca²⁺-inhibited AC5 and AC6. Group IV contains AC9 (TABLE 1) (50). This type-specific regulation of AC isoforms, as well as the tissue distributions of the isoforms, is well studied and has been reviewed in depth previously (1, 3, 11, 51, 96). AC1, AC8, and AC3 are largely concentrated in the central nervous system, so stimulation by Ca²⁺/CaM is mainly observed in these tissues (93).

FIGURE 4.

Regulation of mammalian transmembrane adenylyl cyclases (ACs). Numerous signals directly regulate enzymatic activity of specific AC isoforms. AC isoforms are classified into 4 groups (and color coded) based on these regulatory properties. Green arrows denote stimulatory effects, red arrows denote inhibitory effects, and dashed arrows show conditional effects that are dependent upon other, coincident stimuli. NO, nitric oxide.

Table 1.

Physiological regulators of transmembrane adenylyl cyclase isoforms

AC Isoform	Stimulators	Inhibitors	References
AC1	GαsCa2+/CaMPKC	GαiGβγ	(18, 64–66, 79–81)
AC2	Gαs(Gβγ)PKCRaf kinase		(60, 64, 82)
AC3	GαsCa2+/CaMPKC	GβγRGS proteins	(64, 80–83)
AC4	Gαs(Gβγ)	PKC	(60, 64, 76, 84)
AC5	Gαs(Gβγ)PKCRaf kinase	GαiCa2+PKANitric oxideRGS proteinsAnnexin A4	(64–66, 69, 79, 80, 85–89)
AC6	Gαs(Gβγ)Raf kinase	GαiCa2+PKCPKANitric oxide	(64–66, 68, 69, 79, 80, 86–88, 90, 91)
AC7	GαsPKC(Gβγ)		(60, 64, 92)
AC8	GαsCa2+/CaM	GβγPKA	(64, 93–95)
AC9	Gαs		(64)

Parentheses indicate conditional regulators.

1.3. cAMP Signaling Complexes and Compartmentation

To achieve rapid and specific signaling via the diffusible second messenger cAMP, the various components of the signaling pathway arrange in preformed signaling complexes within cells (97, 98). How these complexes form and regulate various cellular responses is not well understood, but direct interactions between certain GPCRs, ACs, AKAPs, and PDEs have been defined (28, 99, 100). The idea that a single AC isoform might have a distinct physiological role has traditionally been difficult to conceptualize. The ACs were initially viewed as largely redundant isoforms because of cells expressing multiple isoforms that have a singular enzymatic output. Appreciation of the distinct regulatory properties caused a shift in viewing these isozymes as central to tailoring the regulation of cAMP by coincident signals in a specific cell type. However, a framework for how these different isoforms might have distinct effects on cell function remained lacking.

Buxton and Brunton’s observations of cAMP signaling compartmentation (101) followed by the appreciation of ACs being stratified across distinct plasma membrane microdomains (54, 102, 103), facilitated the conceptualization of how individual ACs could have distinct physiological roles. AC isoforms are stratified by their localization in either lipid raft or nonraft microdomains (FIGURE 5). AC1, AC3, AC5, AC6, and AC8 localize to lipid raft microdomains, whereas AC2, AC4, AC7, and AC9 localize to nonraft domains (99, 104). The structural determinants of lipid raft localization of certain isoforms appear to depend on N-linked glycosylation of an extracellular loop in TM2 (FIGURE 1) and perhaps also the intracellular C1 and C2 domains, but more work is needed in this area (12, 105, 106). ACs appear stable in their localization to these microdomains, so they can anchor the formation of larger signaling complexes (99, 107). Of the many Gα_s-coupled GPCRs in a given cell, certain receptors will preferentially couple to a specific AC isoform based on colocalization or preassembly (108–113). A given AC exists in complex with many other organizing and regulatory proteins, such as AKAPs, PKA, and PDEs, but may also form specific interactions with other proteins upon activation (11, 102, 114, 115). Microdomains other than lipid raft and nonraft plasma membranes likely exist, including cAMP signaling in endosomes (116).

FIGURE 5.

Representation of cAMP signaling compartmentation. Membrane microdomains and prearranged complexes form in most cells that couple specific G protein-coupled receptors (GPCRs) to unique subsets of physiological responses. Transmembrane adenylyl cyclase (AC) isoforms characteristically localize in either lipid raft (AC1, 3, 5, 6, and 8) or nonraft (AC2, 4, 7, and 9) domains, where they form complexes with unique signaling partners [GPCRs, A kinase anchoring proteins (AKAPs), phosphodiesterases (PDEs), and others] to create distinct cAMP signaling compartments. There are 43 AKAP isoforms and 24 PDE genes (plus many splice variants), making the combinatorial possibilities for cAMP signaling complex formation enormous. Diffusion of cAMP is limited by PKA buffering, colocalized PDE isoforms, and cellular barriers to diffusion (represented as a brown matrix). Effector proteins downstream of PKA can also be localized in specific domains such that each compartment can give rise to a unique pattern of cellular and physiological responses. AC isoforms are color coded by group (group I in blue, group II in green, group II in red, group IV in brown).

Various ACs can be used to examine the role of cAMP compartmentation. Specifically, in contrast to transmembrane AC isoforms, sAC gives rise to a subcellular cAMP distribution different from membrane-bound adenylyl cyclase (mAC), with distinct functional differences (117). Intriguingly, sAC has a bacterial equivalent, i.e., the type III secretion protein ExoY from Pseudomonas aeruginosa (118). However, both sAC and ExoY have a very broad substrate specificity, giving rise to cGMP, cCMP, and cUMP in addition to cAMP (119–121), so that sAC and ExoY have only limited value to specifically examine cAMP compartmentation. Nonetheless, sAC does display localized cAMP signaling in intracellular cytosolic compartments (122).

An important consequence of the localization of cAMP signaling complexes is that it allows cAMP to be produced locally within a cell and have specific signaling effects, as opposed to globally diffusing throughout the cytosol and triggering all possible downstream signals (99, 123, 124). Anchoring of cAMP response elements to the cAMP generating system can increase signal specificity dramatically. For example, regulation of the transient receptor potential vanilloid 1 (TRPV1) channel by forskolin-stimulated cAMP/PKA signaling became 100-fold more potent when AC5 and AKAP79 were coexpressed in HEK-293 cells (125). In mouse dorsal root ganglion neurons, disrupting the interaction between AKAP79 and AC5 reduced PGE2 regulation of TRPV1 by more than half (125). Thus, anchoring of the GPCR-AC signaling pathway and compartmentation of cAMP signaling play key roles in allowing for specific cell responses despite the ubiquitous nature of this second messenger and the expression of multiple AC isoforms in a particular cell.

2. PHYSIOLOGICAL ROLES OF AC ISOFORMS

As discussed above, the widespread distribution and ubiquitous nature of their product, cAMP, has made it difficult to tease out the unique physiological roles of the various AC isoforms. However, this research is an emerging field for the development of clinically useful pharmacological targets. Both pharmacological activators and inhibitors of ACs are available, but these pharmacological tools are not sufficiently selective for any given AC isoform to allow reasonable dissection of the function of an individual AC in a (patho)physiological context. This problem is highlighted by the AC activator forskolin and three AC inhibitors in TABLE 2. Shown are the data for representatives of the three transmembrane AC families, AC1, AC2, and AC5. Forskolin is more potent at activating AC1 than AC2 and AC5, but the difference is far too small to allow for discrimination of AC isoforms in a cellular or tissue context (126). The same problem applies to structurally distinct AC inhibitors. In general, AC2 is less sensitive to inhibition than AC1 and AC5, but the differences are too small to be useful in (patho)physiological contexts (127–129).

Table 2.

Primary pharmacological tools for studying adenylyl cyclase isoforms

Pharmacological Tool	AC1	AC2	AC5	References
Forskolin (EC50, µM)	3.3	38	18	(126)
MANT-ITP (pKi)	8.55	7.85	8.92	(127, 128)
TNP-GTP (pKi)	7.64	6.66	7.57	(128, 129)
SQ22,536 (pKi)	4.27	3.71	5.46	(128)

There are additional problems with the available pharmacological tools. First, forskolin has effects on several pharmacological targets including glucose transporters (130) and nuclear receptors (131), rendering data interpretation difficult. Second, 2′, 3′-(O)-(N-methyl)anthraniloyl (MANT) nucleotides and trinitrophenyl (TNP) nucleotides are far too hydrophilic to cross the plasma membrane (127). Third, SQ22,536 and related compounds interact with adenosine receptors (132). Since adenosine receptors are ubiquitously expressed, interpretation with SQ22,539 in (patho)physiological settings is problematic. Finally, pharmacological data at each individual AC isoform are available for only a few compounds, e.g., SQ22,536 (128). In most cases, only data for representative AC isoforms are available (126, 127, 129). Although not ideal, it is true that different members from a given AC family are pharmacologically similar (128, 133).

Because of a general lack of isoform-specific inhibitors, there has been an emphasis on knockout and overexpression studies to tease out the specific roles of the isoforms in various systems and pathologies (50). These approaches provide the most direct evidence of how a particular AC isoform mediates a specific physiological function. Several studies also take a genetic approach focusing on the association of AC genes, ADCY1–9, with disease conditions. This review focuses on the major studies on the physiological roles of AC isoforms since the last comprehensive review of ACs (11). Research has overwhelmingly focused on AC1, AC8, AC5, and AC6, presumably because of their recognized roles in the heart and brain. Recent progress has been made in defining the functions of AC9 in the heart. Studies examining the physiological roles of AC2 and AC4 are notably lacking.

3. GROUP I (AC1, AC8, AC3)

3.1. Adenylyl Cyclase 1

3.1.1. Learning and behavior.

AC1 studies have largely focused on its role in the nervous system and regulation of neuronal long-term potentiation (LTP), a molecular surrogate for learning. Early studies with the Drosophila Ca²⁺-sensitive AC1 homolog rutabaga (134, 135), followed by AC1 knockout (AC1^−/−) mice, demonstrated that AC1 is imperative for some forms of LTP in several brain regions and that deletion leads to molecular and behavioral impairment of learning (136–138). However, AC1 knockout does not significantly reduce expression of LTP in the CA1 region of the hippocampus, only the rising phase, but the behavioral learning impairment still occurs (136, 139), although a later study demonstrated a deficit in the expression of LTP after one high-frequency stimulation but not two (140). AC1^−/− mice are still able to find a platform in the Morris water maze, but unlike wild-type mice they fail to favor the region in which the platform is placed. Thus, they appear to lack a spatial memory component but still display a short-term ability to complete the task.

A reciprocal increase in LTP and certain learning behaviors can be induced by overexpressing AC1 in the forebrain (141). Although there were striking increases in LTP, AC1-overexpressing mice did not have pronounced behavioral changes in all learning paradigms. AC1 overexpression resulted in increases in novel object recognition but not contextual or cue-conditioned learning behaviors or spatial memory. It is critical to note that AC1 appears important for the longer-term storage of spatial and recognition memory; however, it seems dispensable for acquisition and for other conditioned learning paradigms (139, 140). Several studies have expanded on elucidating the differential regulation of AC1 on various learning paradigms. AC1 overexpression in mice failed to alter fear learning but did increase social recognition and object recognition (142). Although the object recognition data were similar to previous studies (141), the work was in contrast to a decrease in social interaction shown in a later study (143). Of significance, spatial memory was significantly decreased in aged AC1-overexpressing mice, and they displayed a behavioral phenotype similar to AC1^−/−/AC8^−/− double-knockout mice (143). The current data are conflicting, as a separate paper demonstrated increased spatial memory in AC1-overexpressing mice (144).

Two follow-up studies utilizing these AC1 overexpression mice have provided novel insight into other behaviors with a potential AC1 regulatory component. Overexpression of AC1 in the forebrain resulted in mice that were highly hyperactive, lacked behavioral inhibition, and were less social in multiple social behavior paradigms (143). Nighttime home cage activity increased but not daytime activity, similar to a study observing voluntary home cage wheel running in the daytime (145). Although AC1 overexpression led to risky behavior, a lack of behavioral inhibition, and hyperactivity, it seems to be protective during stress situations and can provide molecular and behavioral resilience to stressors (145). In both studies described above, AC1-overexpressing mice had resilience to acute stress paradigms (143, 145). However, AC1^−/− mice did not display an increased susceptibility to acute stress (140).

These studies demonstrated an interesting role for AC1 in stress-related behaviors, although more rigorous exploration is needed to decipher the contrasts in the data. Overexpression studies do not necessarily provide insight into normal physiological regulation but provide insight into potential pathophysiology associated with increased AC1 expression or behavioral consequences of therapeutically stimulating AC1. Stress is linked to depression and anxiety phenotypes, and a recent genetic analysis reports that intronic single-nucleotide polymorphisms (SNPs) in ADCY1 are associated with anxiety and depression scores in a small human study (TABLE 3) (222).

Table 3.

Disease-linked polymorphisms in adenylyl cyclase isoforms

AC Gene	Disease-Linked Polymorphisms	Nonsynonymous Variants	References
Adcy1	Deafness	R1038X	(146, 147)
Adcy2	Bipolar disorders, autism spectrum disorder, pulmonary diseases, congenital heart disease, Tourette syndrome, developmental disorders	A438T, L498S	(148–155)
Adcy3	Obesity, depression, autism spectrum disorder, schizophrenia, bowel disease	N64I, S107P, G423X, V577M, C948Y, E1008A, I1106X, G1110R, F1118X, L1119S	(156–167)
Adcy4	Unknown
Adcy5	Altered glucose metabolism, diabetes and obesity, autism spectrum disorder, movement disorders (chorea, dyskinesia, familial dyskinesia with facial myokymia, dystonia, hypotonia, spastic paraparesis, alternating hemiplegia of childhood), congenital heart defect, possible link to Parkinson’s disease	Y233H, P399L, R418Q/W/G, R438P, A441V, I460F, N467S, I475M, A534T, D588N, R603C, E634D, K691E, R695W, G697V, L720P, A726T, R727K, E908K, R1013C, D1015E, E1025V, M1029K/R, D1060X, R1192X, M1209V	(13, 14, 16, 155, 168–191)
Adcy6	Adhesion of sickle red cells, arthrogryposis multiplex congenital with axoglial defects, autism, cardiac hypertrophy	A674S, R739W, Y992C, E1003K, R1116C	(192–198)
Adcy7	Alcoholism, juvenile idiopathic arthritis, autism spectrum disorder, schizophrenia, ulcerative colitis	V272I, K349M, D439E, S1020G	(199–206)
Adcy8	Autism spectrum disorder, bipolar disorder, dissociative disorder, schizophrenia, depression and alcoholism, myocardial infarction	L327V, V602I, A728P, R968Q	(152, 164, 168, 207–210)
Adcy9	Asthma treatment, pancreatic cancer, body mass index, malaria susceptibility, schizophrenia, autism spectrum disorder, efficacy of dalcetrapib to treat cardiovascular events	K386X, I772M, S782G, N1154S	(154, 211–220)

Missense and nonsense nonsynonymous variant sites are listed. The Human Gene Mutation Database was used to compile some of the information in this table (221).

3.1.2. Neural development.

AC1^−/− mice display deficits in correct neural wiring during development, notably in the primary somatosensory cortex and visual systems (223, 224). Region-specific knockouts later demonstrated that thalamic AC1 deletion had more severe sensorimotor and social deficits than cortical AC1 deletions, suggesting a key role for thalamic AC1 in proper neuronal development (225). The AC1 knockouts, or region-specific knockouts in this study, were devoid of alterations in locomotor activity despite the observation that overexpression of AC1 in the forebrain resulted in 20% increased locomotor activity (143).

As the AC1 overexpression phenotype displayed some of the same behavioral characteristics as in the autism-like mouse Fmr1 knockout model, it was recently hypothesized that knockout of fragile X mental retardation protein (FMRP), protein product of Fmr1, may impact AC1 expression or function (226). Indeed, it was elegantly determined that loss of FMRP leads to increased AC1 mRNA and protein levels (227). Genetic and pharmacological inhibition of AC1 in this mouse model not only reduced the molecular pathology but reversed several of the autism-like behaviors in these mice (227). It should be noted that in a separate study exploring the interaction between FMRP and AC1 in pain, FMRP knockouts did not display increased AC1 levels and were actually protective against increased AC1 expression due to pain (228). However, the pain study was examining the anterior cingulate cortex (ACC), and the autism study explored AC1 expression in the hippocampus.

3.1.3. Pain.

There are significant studies exploring AC1 in pain, and foundational studies utilizing AC1^−/− mice demonstrated an essential role for AC1 in chronic pain behaviors (229). AC1^−/− mice exhibit reduced allodynia in the complete Freund’s adjuvant and formalin models of inflammatory pain but maintained acute nociceptive responses. Furthermore, recent studies have demonstrated the involvement of AC1 in bone cancer (230), models of diabetic neuropathy pain (231, 232), migraine (233), and visceral pain (234, 235). These studies demonstrated that although AC1 does not regulate acute nociceptive functions, it is critical to longer-term pain syndromes and provides a possible novel therapeutic opportunity for the treatment of chronic pain (FIGURE 6).

FIGURE 6.

The role of adenylyl cyclase (AC)1 in chronic pain-induced allodynia in vivo. In wild-type mice subjected to a chronic pain model, AC1 inhibitors can block allodynia. In wild-type mice that have recovered from a chronic pain model, AC1 inhibitors can block naltrexone (NTX)-induced allodynia (latent sensitization model). It is likely that stress-induced allodynia in chronic pain models will also be blocked by AC1 inhibitors. In AC1 knockout mice, chronic pain models do not produce allodynia.

AC1 plays a critical role in synaptic plasticity, and this function drives its physiological role in chronic pain. As a Ca²⁺/CaM-sensitive cyclase, increases in intracellular calcium promote increased AC1 activity. During chronic pain states, persistent neuronal signaling through voltage-gated calcium channels and N-methyl-d-aspartate receptors (NMDA-Rs) leads to increased AC1 activity, and these plasticity alterations in the ACC appear to maintain allodynic behaviors (236, 237). Furthermore, these alterations have been proposed to act in a positive feedback manner, as increased AC1 activity can reciprocally drive increases in the GluN2B subunit of NMDA-Rs (234) as well as increase the phosphorylation and expression of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors (235, 238). The cycle of NMDA-R-driven AC1 activity and upregulation of NMDA-Rs and AMPA receptors leads to persistent activity that can drive long-term potentiation and result in “sensitized” neurons (FIGURE 7). Although many studies have investigated AC1 in the ACC for pain, AC1-induced long-term potentiation in the insular cortex is involved in the development of certain chronic pain models (233, 234, 238, 239) and spinal AC1 contributes to the transition and maintenance of long-term pain (240, 241).

FIGURE 7.

Cellular adaptations to chronic nociceptive input promote allodynia through adenylyl cyclase (AC)1. Top: sufficient acute nociceptive input results in activation of voltage-gated calcium (CaV) channels, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors (AMPARs), and N-methyl-d-aspartate (NMDA) receptors (NMDARs). This activation at varying levels of the pain pathways (i.e., spinal cord, anterior cingulate cortex, insular cortex) can produce an acute behavioral response in rodents, typically a response to a noxious mechanical/thermal stimulus. Although calcium influx through these channels will activate AC1, AC1 does not modulate acute nociceptive responses. Bottom: chronic nociceptive input produces sustained calcium influx and AC1 activation. This promotes a positive feedback loop whereby AC1 activation results in phosphorylation of AMPARs and increased expression of GluN2B-containing NMDARs, both of which further promote increased AC1 signaling. This ultimately results in a sensitized system where nonnoxious stimuli produce a behavioral response (allodynia). Both panels are generalized schemes and likely occur at multiple levels of the pain pathways. Mu opioid receptor (MOR) agonists can inhibit AC1, among other functions. AC1 regulation by MORs in the acute nociceptive setting (top) is likely different from that in the chronic setting (bottom). In the chronic setting there is an interplay between NMDARs, MORs, and AC1 that are involved in latent sensitization (see text).

There is also significant evidence for the contributions of AC1 in a phenomenon known as latent sensitization (240, 242). In one study, animals were subjected to inflammatory pain and then allowed to recover from the pain model over several weeks (240). After recovery, injection of naloxone caused the hyperalgesia and allodynia to return, an unmasking of latent pain. This phenomenon is mediated through constitutive activity of the mu opioid receptor inhibiting AC1 activity. It was also demonstrated that this naloxone-precipitated pain could be effectively blocked by inhibiting AC1 activity or the NMDA receptor. This mechanism further emphasizes the intricate interactions between NMDA receptors and AC1 and how the communication between the two signaling molecules regulates the transition to chronic pain in various areas along the pain neuroaxis through synaptic plasticity (FIGURE 7).

The significant contribution of AC1 to the development of chronic pain, but not acute protective nociceptive functions, provides opportunity for novel nonopioid pain therapeutics. Indeed, drug discovery efforts for inhibitors of AC1 have generated two novel compounds (NB001 and ST034307) that can effectively alleviate chronic pain by inhibiting AC1 in the spinal cord and/or the ACC (243, 244). However, it should be noted that the mechanism of action of NB001 appears to be indirect to AC1 and potentially through CaM interactions (128, 244). Redox reactions can alter CaM activity to fine-tune the regulation of AC1, providing an opportunity for the development of small-molecule inhibitors that leverage CaM-AC1 to treat chronic pain (245). Although there is substantial evidence that AC1 plays a critical role in the development of various chronic pain syndromes, all of it is derived from knockout animal studies or pharmacological manipulations. It is critical that new and novel tools be developed to rigorously validate AC1 regulation and function in chronic pain.

3.1.4. Other physiological functions.

Besides the more established literature on synaptic plasticity, affective behaviors, and pain, AC1 has also been linked to drugs of abuse. AC1^−/− mice had reduced opioid withdrawal symptoms including wet dog shakes, paw tremor, diarrhea, ptosis, and weight loss compared with wild-type mice (246). More recently, it was revealed that AC1^−/− mice have a reduction to ethanol-induced locomotor sensitization and reduced GluN2B-NMDA phosphorylation following repeated alcohol administration (247). Other functions of AC1 include being an effector component of a contrast sensitivity circadian rhythm mechanism in retinal ganglion cells (248, 249). The two studies from this group determined that AC1 levels in a subset of retinal ganglion cells are modulated on a circadian rhythm through dopamine D4 receptor (DRD4) activating clock gene Npas1. Knockouts of DRD4 or Npas1 did not have rhythmic AC1 levels, and AC1^−/− mice had reductions in sensitivity contrast. A single study has also demonstrated a conserved role for AC1 in hearing, where a mutation in the carboxyl tail caused a mild hearing impairment in humans, and this mutation-induced phenotype can be recapitulated in zebrafish (146).

Beyond the brain, AC1 has been linked to the cholesterol efflux pathway in macrophage foam cells (250). Macrophage foam cells are the main cells found in atherosclerotic lesions. AC1 protein and mRNA levels were consistently upregulated during cholesterol loading and apoA-1 activation. Knockdown with short hairpin RNA (shRNA) of AC1 reduced apoA-1-induced cAMP production and decreased cholesterol efflux while intracellular cholesterol remained high. It was suggested that AC1 is the main effector for apoA-1-mediated cholesterol efflux (transport and exocytosis) in macrophage foam cells. This is significant because cholesterol efflux is important for reducing the risk of atherosclerosis. AC1 has also been implicated in chloride secretion in airway epithelial cells, an important function in the pulmonary system (251). This study provided evidence of the colocalization of GPCRs, cystic fibrosis transmembrane conductance regulator (CFTR), TMEM16A, AC1, and EPAC1 in microdomains. AC1 and EPAC1 were responsible for mediating the cross talk between Ca²⁺ and cAMP signaling, and this cross talk activated CFTR and TMEM16A to amplify electrolyte secretion in airway epithelial cells. AC1 regulation by CaM appears to play roles in the physiological effects of both Bordetella pertussis (whooping cough) and Bacillus anthracis (anthrax) infections, raising the prospect that inhibitors of CaM stimulation of AC1 could be a viable therapeutic approach for treating these infections (252, 253).

3.2. Adenylyl Cyclase 8

3.2.1. Central nervous system functions.

AC8 is also largely studied for its role in the brain. The synaptic regulation of AC1 and AC8 has been extensively reviewed (254). Briefly, AC8 knockout (AC8^−/−) mice, like AC1^−/− mice, had slight deficits in mossy fiber LTP in the hippocampus (255). They displayed normal acute synaptic function and normal perforant path LTP. However, AC8^−/− mice also displayed deficits in long-term depression (256) and short-term plasticity measures, whereas AC1^−/− mice did not (255). They also demonstrated normal contextual and passive avoidance memory and locomotor activity (257). AC8^−/− mice had impaired object recognition memory, but their spatial recognition memory appeared intact (257). However, when AC8^−/− mice were tested in a more rigorous and rapid acquisition spatial memory task, animals took longer to find the relocated platform in the Morris water maze. This suggests that AC8 has an important role in rapid novel spatial recognition memory and working/episodic-like memory (257).

Early studies with these AC8^−/− mice suggested that they display normal baseline anxiety behavior and were resistant to repeated stress-induced anxiety (256, 258, 259). One of these studies confirmed previous findings that AC8^−/− mice have reduced stress-induced and anxiety-like behaviors, although the data for social defeat were not significant (259). They also noted an interesting increase in spontaneous activity in the home cage of mice during dark phase (wake hours), a significant decrease in leptin content, reduced stress-induced weight loss and feeding, and a significant enlargement of the adrenal glands through an unknown mechanism. A long-term anxiety-like paradigm also demonstrated a role for AC8 in anxiety but not for AC1 (258). AC8 has a demonstrated role in stress-induced behaviors and anxiety in mouse models, consistent with human studies that have identified SNPs within the AC8 gene that are associated with bipolar disorder (207, 208), posttraumatic stress disorder (209), and alcoholism comorbid with major depressive disorder (168).

3.2.2. Cardiac functions.

Recent studies of AC8 have focused on its role in the sinoatrial node (SAN) of the heart. Via cardiac-specific overexpression of AC8 in mice, it was demonstrated that these mice displayed increased heart rate and decreased heart rate variability (260). They revealed that heart rate and rhythm can be intrinsically driven by AC8 when overexpressed in SAN cells, independent of autonomic drive. It was suggested that the mechanisms behind this cardiac physiology may be due to the modified catecholamine synthesizing and degradation enzyme levels, plasma catecholamine levels, desensitization of β-adrenergic signaling, or blunted vagal input. These results have physiological significance in maintaining heart health and survival, and the authors suggested that further studies may provide evidence for pharmacological intervention to treat arrhythmias and heart failure. It was noted that these modifications due to AC8 overexpression resemble the aging heart, and longer studies are needed to determine whether AC8 overexpression may lead to early heart failure. The age of the mice was not noted in this study. Another study using transgenic AC8 overexpression mice investigated age-related cardiac function (261). AC8-overexpressing mice have increased contractility, reduced strain rate, but accelerated age-related heart dysfunction, early cardiomyocyte hypertrophy, and interstitial fibrosis (261). The early effects of aging were associated with subcellular remodeling of cAMP/PKA pathway. Both studies implicate chronic AC8 overexpression as detrimental to heart function. However, studies using AC8 knockdown are needed to confirm a role for this isoform in cardiac function.

3.2.3. Calcium regulation.

AC8 regulates calcium levels in cells, specifically store-operated calcium (SOC) entry, through interactions with a channel called Orai1 (262, 263). Orai1 functions as an SOC channel, allowing calcium entry into the cell after internal endoplasmic reticulum stores are depleted. AC8 regulated Orai1 through a direct interaction between the NH₂ terminus of AC8 and the NH₂ terminus of Orai1 that modified the phosphorylation state of Orai1 (263). It was later determined that the interaction of AC8 and Orai1 positions AC8 to be readily activated, activating downstream PKA, which is then positioned to phosphorylate Orai1 to inactivate it (264). A recent study further investigated AC8’s involvement in this pathway in cancer cell lines. This study (265) suggested an interesting countermechanism of AC8 and Orai1 regulation where overexpression of AC8 and Orai1 variants results in AC8-mediated enhancement of SOC, not reduction as previously suggested (264).

3.3. Adenylyl Cyclase 1 and 8 Redundancies

AC1 and AC8 are often studied in tandem via double-knockout studies, which have revealed many overlapping functions between the two isoforms. Both AC1 and AC8 are Ca²⁺/CaM sensitive and exhibit a similar expression pattern with overlap in cortical layers, the hippocampus, amygdala, thalamus, striatum, and brain stem regions (266). Thus, one isoform can often take over the functions of the other, where the single knockout displayed no phenotype but the double-knockout mice displayed significant alterations. This was clearly demonstrated when observing the outcome of AC1^−/−, AC8^−/−, or double knockout (DKO) on late-phase long-term potentiation (L-LTP) in the CA1 region of the hippocampus (139). Whereas single-knockout mice alone had no change in L-LTP, the double-knockout mice lost all L-LTP. These deficits were paired with deficits in long-term memory. Double-knockout mice had reduced passive avoidance learning and reduced contextual learning but not cued learning and memory. A further characterization of AC1 and AC8 double-knockout mice revealed no change in anxiety-like behavior in the elevated plus maze, in the light/dark preference test, or in open field (267). Surprisingly, the male double-knockout mice had significantly reduced immobility in the forced swim test (antidepressant-like behavior) whereas females did not, and both males and females interacted more in a social behavior test. Both female and male double-knockout mice had a reduced sucrose preference and reduced body weight. However, it was noted that double-knockout mice were often compared to non-littermate control animals because of the large numbers of mice required, making it difficult to account for differences in maternal behavior.

Double-knockout studies have also revealed roles of AC1 and AC8 in the effects of a multitude of drugs of abuse including ethanol consumption and locomotor responses (247, 268–271), cocaine sensitization (272), opioid tolerance and withdrawal (246, 273), and methamphetamine activity (274). A recent double-knockout study further investigated the role of AC1 and AC8 in the reward areas in the brain (269). Manganese-enhanced magnetic resonance imaging (MEMRI), a novel brain imaging technique in live animals that maps the entry of manganese into active neurons through calcium channels, revealed that AC1/AC8 double-knockout animals had reduced activity in reward areas. These areas included the ACC, nucleus accumbens, medial prefrontal cortex, anterior caudate putamen, medial thalamus, and ventral lateral thalamus. This hypoactivity was linked to a significant decrease in ethanol consumption and preference over time and a range of ethanol concentrations. These results provided an interesting whole brain imaging approach to link previously determined changes in ethanol drinking behavior in double-knockout mice to various brain regions in live animals.

AC1 and AC8 are thought to be present in SAN cells and appear to be involved in SAN pacemaker activity (275–277). The double-knockout mice have altered Ca²⁺ homeostasis and sensitivity in the SAN (276). These changes altered automaticity and pacemaker current, suggesting the Ca²⁺-stimulated AC1 and AC8 isoforms are important for coupling calcium signaling to pacemaker current. More specific means of pharmacological inhibition, more specific antibodies, and utility of single AC knockouts are needed to provide more informative measures of AC1 and AC8 regulation in the SAN. The unclear RNA expression profile of AC1 and AC8 in the SAN emphasizes this point. In guinea pigs AC1 is expressed in the SAN, but there is very low expression of AC8 (277). In one mouse strain, C57BL/6, there is no detected AC8, but AC1 is not examined (278). A separate RNA sequencing (RNA seq) study of SAN cells in transgenic mice in the CD1(ICR) background detected AC1 but not AC8 RNA; however, AC1 was not enriched in the SAN compared with the right atrial myocardium (279). As mentioned above, a recent overexpression study determined that overexpression of AC8 in SAN cells can increase heart rate and decrease heart rate variability (260). These results suggest that AC1, and potentially AC8, may be central to maintaining basal pacemaker current, but the overexpression study suggests an important role for either Ca²⁺/CaM-stimulated cyclase activity or cAMP in general for maintaining heart rate and rhythm and potentially long-term cardiac dysfunction (261).

3.4. Adenylyl Cyclase 3

3.4.1. Olfaction.

AC3 was originally thought to be an olfactory-specific adenylyl cyclase (280) but was later identified in a variety of other cell types (281). It is now accepted that AC3 is a predominantly cilium-specific adenylyl cyclase and can be found on primary cilia throughout the brain (282). AC3 is integral for proper olfaction (283), and this role has led to a number of studies on AC3 in terms of olfaction mechanisms and development (284–287), pheromone detection (288), maternal behavior (289), and olfactory-dependent learning (290, 291).

3.4.2. Depression.

It has been reported that AC3 is the primary adenylyl cyclase in human blood platelets (292), and studies have suggested that reduced AC3 activity in these platelets is a potential biomarker for major depression (293, 294). Although not significant, SNPs in AC3 itself were suggested to be associated with major depressive disorder (MDD) (156), as were lower blood transcript levels of AC3 (295). Recent studies have taken these human studies and determined whether this also translates in animal models of depression-like behavior.

Recently, a study tested AC3^−/− mice against a battery of depression-like paradigms (296). AC3 knockouts had increased immobility in both the tail suspension and forced swim tests, decreased sociability, reduced novelty-suppressed feeding and drinking, a reduced coat grooming score, and decreased nesting scores, all indicative of depression-like behaviors. They also exhibited increased rapid eye movement (REM) sleep, altered non-REM sleep, hippocampal atrophy, and reduced signaling, impaired spatial navigation, and depressed L-LTP, all markers of depression-like features in humans. This study also examined an inducible AC3^−/− mouse model and a conditional knockout model only in the forebrain. Many of the same depression-like symptoms were recapitulated in these models, as in the germline knockouts. This study further validates that reductions in AC3 may result in depression as the human studies suggested.

A very recent study determined that deletion of AC3 from somatostatin (SST)-positive cortical interneurons was sufficient to induce depression-like and anxiety-like behaviors in mice (297). Deletion of AC3 in parvalbumin (PV)-positive interneurons failed to result in a similar phenotype. Confirming previous findings, AC3^−/− mice display a plethora of depression-like symptoms (298). This was paired with a reduction in hippocampal tyrosine hydroxylase (TH), dopamine D1 receptor (DRD1), and dopamine D3 receptor (DRD3) protein levels. Dopamine D2 receptor (DRD2) levels were reduced, but not significantly. Interestingly, when this group conditionally deleted AC3 from the main olfactory epithelium (MOE), mice displayed a depression-like phenotype somewhat similar to the global knockout, although food intake was unaffected. The global AC3 knockout effects on hippocampal protein levels of TH and DRD3, and the trending reduction on DRD2, were recaptured in the MOE AC3 knockout. MOE AC3 knockouts also had reduced GluN2B protein levels. Future studies are needed to rigorously determine the mechanisms of AC3 in depression because multiple AC3 knockout paradigms display a similar phenotype likely affecting different neuronal functions.

3.4.3. Metabolic regulation.

An early study found that AC3^−/− mice become significantly obese, with males increasing in mass by 40% and females increasing by 70% (299). These AC3^−/− mice had significantly more fat versus lean mass, larger adipocytes, and increases in triglycerides, leptin, and insulin. At young ages, AC3^−/− mice were less active and consumed more food, and at old ages they continued to be less active. These physiological changes were hypothesized to be mediated through loss of AC3 expression in primary cilia in the hypothalamus (299), a critical region for food intake regulation (300). This group later verified this hypothesis by deleting AC3 from the hypothalamus and demonstrated an increase in food consumption and body weight (301). A specific ventromedial hypothalamus deletion resulted in the same effect (301), although a different subregion of the hypothalamus, the paraventricular nucleus, has also been implicated in the body weight regulation of AC3 (302).

Recent studies have continued to build on these previous studies demonstrating AC3 regulation of metabolism. Haploinsufficiency of the ADCY3 gene in AC3^−/+ mice produced a phenotype similar to AC3^−/− but was dependent on the type of diet mice were fed (303). When fed normal mouse chow, heterozygous AC3 mice do not gain weight. When fed a high-fat diet, these mice gained significant weight, while they consumed the same amount of food as wild-type mice. Furthermore, when fed a high-fat diet, AC3^−/+ mice have increased triglycerides, leptin, and other hepatic biochemical parameters. Additionally, they display impaired glucose tolerance and insulin resistance. It is intriguing that heterozygous mice display a metabolic profile similar to AC3 knockout mice, but only when fed a high-fat diet. Conversely, mice with a gain-of-function mutant in AC3 display resistance to diet-induced increases in body weight, leptin content, and insulin content (304).

These mouse studies have substantial translational bearing in humans. Variants of the ADCY3 gene have been linked to increased body mass index (BMI, corrected for height) in children (157); increased type 2 diabetes, BMI, fat percentage, fasting plasma glucose, and acute plasma glucose in a Greenlandic cohort (158); and increased BMI in obese adult subjects (159). The human genetic data, coupled with the animal models and AC3 location, provide compelling evidence for AC3 as a unique therapeutic target for obesity disorders. Interestingly, potential activators of AC3 may confer benefit for both obesity and depression.

4. GROUP II (AC2, AC4, AC7)

4.1. Adenylyl Cyclase 2

Reports of AC2 knockout or overexpression are scarce. One previous study in human airway smooth muscle cells suggests a role for AC2 in mediating airway smooth muscle tone in response to prostaglandins (305), but research to directly verify AC2’s role in physiological functions of airways with knockdown or genetic deletion has not been reported. AC2 activation in the brain via Gβγ and Gα_s may be associated with chronic opioid use and tolerance (306), but the specific role of AC2 is inferred by the Gβγ and Gα_s dependence, so AC4 may also be involved in these responses. One major hurdle to understanding AC2’s physiological role is the expression pattern variation between mice and humans. AC2 expression has been observed in human airway smooth muscle cells, but the same expression pattern is not found in mouse airway smooth muscle (305, 307). Studies of ADCY2 polymorphisms suggest roles for AC2 in tobacco-exposed lung function, chronic obstructive pulmonary disease, bipolar disorder, and schizophrenia (TABLE 3) (148–150, 308), but, as with other cases, specific knockdown studies have not been performed to directly establish mechanistic links between this AC isoform and these diseases. AC2-selective inhibitors have been reported and should aid in studies of the physiological roles of AC2 (126, 309, 310). A study using an AC2-selective inhibitor in a cell model of Lesch–Nyhan disease suggests that AC2 plays a role in the pathogenesis of this neuropsychiatric disorder (311). Inducing a deficiency in hypoxanthine phosphoribosyl transferase in cells, which mimics Lesch–Nyhan disease, leads to an overall reduction in AC activity and a specific loss of AC2 expression (311, 312).

One recent study examined the gene expression profiles of young and adult mice in a schizophrenia model (313). Differential expression of several genes involved in calcium homeostasis and signaling along with ADCY2 were identified. ADCY2 expression was reduced in both juvenile and adult schizophrenic mice compared with control animals. These results imply a role for AC2 in schizophrenia pathology. These investigators predict specifically that ADCY2 plays a role in early psychotic symptoms relating to learning and memory. Specific manipulations of AC2 expression are needed to confirm these mechanistic linkages, and an examination of ADCY2 expression in humans with schizophrenia is required before AC2 can be declared a therapeutic target for the disease.

4.2. Adenylyl Cyclase 4

Reports of knockout and overexpression studies of AC4 are also limited in the literature, as are ADCY4 polymorphism studies. One previous tissue-specific AC4 knockout has been reported in kidney collecting duct principal cells, but the results indicate that AC4 does not play a role in the collecting duct’s handling of water or sodium (314). Interestingly, AC4 deletion left vasopressin-stimulated cAMP production unaltered in this region of the kidney, an indication that vasopressin receptors only couple to one of the other AC isoforms (AC3 and/or AC6) expressed in the collecting duct. AC4 appears to have a role in dopaminergic regulation of prefrontal cortical neurons (315). Dopamine-stimulated cAMP production was potentiated by simultaneous activation of M₁ muscarinic acetylcholine receptors in neuronal primary cultures. This potentiation depended upon concomitant stimulation of Gβγ and Gα_s, a classical feature of group II ACs. Knockdown of AC4 abrogated this effect, whereas knockdown of AC2 was inconsequential. Although these studies imply that AC4 is involved in attention, cognition, and emotion, there are as yet no direct studies linking AC4 to these functions. AC4 expression is high throughout the heart and vascular system (57), yet no studies of its physiological roles in these systems have been reported.

Another recent study implicates a potential role of AC4 in regulating lipopolysaccharide- or Escherichia coli-mediated sepsis (316). ADCY4 is expressed in macrophages and participates in the α_2b-adrenoceptor-PDE8A-PKA signaling pathway that underlies the inhibition of caspase-11 inflammasome activation by epinephrine. AC4 knockdown reduced epinephrine-stimulated cAMP synthesis and inhibition of lipopolysaccharide (LPS)-stimulated inflammasome activation, whereas knockdown of the highly homologous isoform, AC2, lacked these effects. Although AC4 was not directly linked to sepsis in mice, pharmacological inhibition of PDE8A did protect mice from caspase-11-mediated sepsis and septic death, implying a central role of the entire α_2b-adrenoceptor-AC4-PDE8A-PKA signaling pathway. Thus, AC4 may be a relevant therapeutic target in sepsis.

A recent epigenetic study may implicate AC4 in breast cancer pathology (317). ADCY4 was significantly downregulated in breast cancer, and high expression of the gene in breast cancer patients correlated with better survival. This suggests that AC4 may act as a tumor suppressor or restrain metastasis. The expression levels of AC4 correlated negatively with methylation of the ADCY4 promoter, with hypermethylation observed in breast cancer patients. This study indicates that AC4 expression could be used as a biomarker for breast cancer, and it warrants further research into AC4 as a possible therapeutic target.

4.3. Adenylyl Cyclase 7

Previous studies have demonstrated roles for AC7 in several different physiological processes and suggest that alterations in the activity of this isoform may lead to diseases. Particular attention has been devoted to studying alterations in the immune system and neuropsychiatric disorders such as alcoholism and depression (318). More recent studies have also indicated that AC7 plays a role in certain cellular processes associated with cancer (319, 320).

The regulatory properties of AC7 are similar to those of other group II adenylyl cyclases. AC7 possesses the Gln-X-X-Glu-Arg motif that should allow for Gβγ interaction and conditional activation of the cyclase by the dimer, but this has not been directly tested in vitro (92, 321–323). Moreover, AC7 is insensitive to Gα_i/o, activated by PKC, conditionally activated by Gα_12/13, and, interestingly, potentiated by ethanol (92, 318, 322, 324–326). The relationship between AC7 function and ethanol has been the focus of a number of studies. At the molecular level, ethanol’s conditional activation of AC7 is insensitive to pertussis toxin but suppressed by PKC inhibition (327, 328). Ethanol appears to directly interact with AC7 but may require additional proteins for its stimulatory effect in certain cellular environments (329, 330). PKCδ phosphorylates AC7, and this step is required for the potentiation of AC7 activity by ethanol in HEK and HEL cells (327). Transmembrane regions of AC7 are not required, as ethanol can still enhance the activity of mutants of AC7 that were created lacking the transmembrane domains (330, 331).

Even though AC7 knockout is largely lethal to mice (324), several AC7 expression studies have implicated a role for AC7 in alcoholism. For instance, ethanol-induced increases of blood adrenocorticotropin (ACTH) and corticosterone are decreased in heterozygous ADCY7^+/− female mice and increased in female mice overexpressing AC7 compared with wild-type animals (332). In addition, ethanol preference is increased in ADCY7^+/− female mice compared with wild-type mice (333). In humans, the presence of a SNP in ADCY7 has been correlated with a reduction in the risk of developing alcohol dependence for females (333). However, in contrast to the mouse data, this SNP is linked with lower levels of ADCY7 transcripts in blood and adipose tissue compared with its major allele (333). It would be interesting to measure the levels of AC7 protein expression in the brain of subjects with this SNP to determine whether the same pattern would be observed. AC activity in platelets from men with alcoholism is significantly decreased compared with matched control subjects (334, 335). Although multiple AC isoforms are likely to be expressed in platelets, AC7 is the main isoform in the platelet-derived HEL cells (199). As this study was done with platelets, it presents a remarkable contrast with the AC7 blood transcript levels from the SNP study. However, a constant feature of AC7 studies has been the presence of sex-specific effects. As noted above, the study with human platelets was done in male subjects, whereas in the mouse studies and the human SNP study significant results were only obtained with females (332–335).

The role of AC7 in depressive disorders has recently been reviewed (318). In summary, the available studies indicate that higher AC7 activity is linked to depression. One study used ADCY7^+/− mice and mice overexpressing AC7 in the central nervous system (CNS) to demonstrate that in female mice lower AC7 expression correlates with a reduction in behaviors that are linked to depression in the forced swim and tail suspension tests (200). However, no significant effects were observed in male mice. The amygdala is a central component of the limbic system that has an important role for emotion and motivation control (336). Notably, the levels of AC7 mRNA are increased in the amygdala of male humans with familial major depression and in a mouse model of depression that lacks the serotonin transporter compared with their respective control subjects (201). A genetic polymorphism in ADCY7 has also been found to correlate with an increased risk of familial depression in women (TABLE 3) (200). Human carriers of that polymorphism display increased threat-related amygdala reactivity in comparison to control subjects (201).

cAMP plays an important role in the regulation of immune responses. For instance, the activity of T and B cells, neutrophil chemotaxis, and cytokine production have all been related to changes in cAMP levels and, therefore, with AC activity (337–342). Genetic deletion of AC7 in the hematopoietic system results in hypersensitivity to lipopolysaccharide (LPS)-induced endotoxic shock, increased production of the inflammatory cytokine TNF-α by bone marrow-derived macrophages (BMDMs) in response to LPS, and deficient antibody responses to antigens (337). The study also highlighted the importance of AC7 for cAMP responses in B and T cells, particularly in T cell-dependent antibody responses, which are crucial for adaptive immunity (337). As further discussed below, it should be noted that there are other AC isoforms expressed in those cells and AC9 activity has also been implicated in T cell function (338).

Several other studies have established a role for AC7 in the production of TNF-α. When an inflammatory response was triggered, TNF-α production in AC7-deficient BMDMs was markedly increased compared with wild-type BMDMs (343). In other studies, knockdown of AC7 in a monocytic leukemia cell line caused an increase in TNF-α production in response to LPS (344). In humans, this correlated with an LPS-induced increase in plasma TNF-α concentrations and a decrease in AC7 transcripts in primary monocytes from male subjects (344). It is noteworthy that sex differences were also observed in this study, with LPS-induced production of TNF-α being more pronounced in young males than in young females. Furthermore, the drop in AC7 transcripts was not observed in blood samples from females (344). Additional evidence for the role AC7 plays in the human immune system comes from a genomewide association study identifying a single-nucleotide polymorphism (SNP) in one of the intron regions of the ADCY7 gene that is associated with two pediatric-age-of-onset autoimmune diseases, psoriasis and Crohn’s disease (TABLE 3) (202).

The conditional activation of AC7 by Gα_12/13 has been implicated in the enzyme’s function in immune responses. Specifically, activation of Gα_12/13-coupled receptors potentiates Gα_s-stimulated AC7 activity in BMDMs (324, 325). Upon treatment with zymosan, a cell wall extract from Saccharomyces cerevisiae that induces inflammatory responses, BMDMs display a robust increase in cAMP production. Notably, that response largely depends on the expression of AC7 (343). Knockout of Gα₁₂ and Gα₁₃ in those cells causes a profound reduction in PKA activity in response to zymosan. Pertussis toxin treatment also decreases PKA activity in that context, but the decrease is less pronounced (343).

The activity of AC7 has also been linked to certain processes related to cancer. For instance, in solid tumors, inadequate oxygen supply results in activation of the hypoxic pathway. Hypoxia is generally linked to tumor resistance to therapy, robust malignancy, and, thus, poor prognosis (345, 346). Carbonic anhydrase IX (CA IX) is an important mediator of cell survival and malignancy in hypoxic conditions (347). Notably, the activity of PKA is essential for CA IX activation, indicating a possible role for specific AC isoforms in the process (348). In four different carcinoma cell lines, AC6 and AC7 mRNA expression was increased in hypoxic versus normoxic conditions (319). Moreover, the upregulation of AC6 and AC7 is dependent on hypoxia-inducible factor-1 (HIF-1), which is also necessary for the expression of CA IX in hypoxic conditions (347). Knockdown of AC6 or AC7 under hypoxic conditions causes a significant reduction in cAMP accumulation and PKA activation and results in a decrease in migration capacity of the cells (319). These data suggest a role for AC6 and AC7 in the malignancy capacity of the cell lines tested.

AC7 has a role in acute promyelocytic leukemia (APL) cell differentiation (320). Knockdown of AC7 inhibits retinoic acid-induced differentiation of the human acute promyelocytic leukemic cell line NB4. Moreover, a microRNA (miRNA) that has been reported to participate in the progression of acute myeloid leukemia (miR-192) diminishes AC7 expression and lowers cAMP accumulation in NB4 cells (320). The study also showed that knockdown of miR-192 causes upregulation of retinoic acid-induced differentiation of NB4 cells and that upregulation can be disrupted by knocking AC7 down. Notably, in patients with relapsed APL the transcript levels of AC7 in bone marrow cells are significantly lower than the levels in patients with newly diagnosed APL (320). These results correlate with an increase in miR-192 levels in bone marrow cells of patients with relapsed APL versus patients with newly diagnosed APL (320).

It is interesting to contrast the meaning of the data in the manuscripts from Simko et al. (319) and He et al. (320). Simko et al. (319) found that the activity of AC7 is important for tumor cell migration in hypoxia, whereas He et al. (320) indicated that a drop in AC7 expression is linked to APL relapse. These two studies present opposite functions for AC7 in two different processes related to cancer. Whereas one associates increased AC7 activity with metastasis, the other links decreased AC7 activity with cancer relapse (319, 320). These data highlight the complexity of tumorigenic processes and also the selective nature of AC isoform functions within different processes related to the same category of diseases.

5. GROUP III (AC5, AC6)

5.1. Adenylyl Cyclase 5

AC5 is very broadly expressed in mammalian organs (11). Accordingly, it is not surprising that AC5 plays important roles in the regulation of multiple organ functions. These properties render AC5 an attractive drug target, but developing a clinically useful AC5 inhibitor has a number of challenges. First of all, and most importantly, there is a paucity of potent and highly selective AC5 inhibitors despite the efforts of many research groups (11, 227, 349–353). Second, the broad expression of AC5 in multiple organs (11) means that systemic application of any selective AC5 inhibitor would have physiological effects in many organs and numerous adverse effects. A solution to this problem could be the local administration of AC5 inhibitors, but such an approach is applicable only to a few organs such as skin, eye, and nose. Third, in some organs such as the heart AC5 inhibition is assumed to be therapeutically valuable (353), but in other organs such as the brain AC5 inhibition may be deleterious and cause multiple neuropsychiatric problems (354, 355). Although the therapeutic index of AC5 inhibitors may be narrow, targeting a drug to act only peripherally or only in the CNS may have therapeutic utility. A recent study suggests that annexin A4 inhibits AC5 but not AC6 (356), but at the present time it is difficult to envisage how this mechanism can be exploited pharmacologically. Based on these considerations, it is still too premature to seriously consider clinical applications of AC5 inhibitors. It has been proposed that certain AC inhibitors could be used to preferentially inhibit the elevated activity of AC5 mutants in a distinct form of human dyskinesia (357), but the insufficient selectivity of the currently available AC inhibitors is a serious concern (11).

AC5 deletion has been suggested to play a protective role in the heart, by protecting both against β-adrenoceptor (βAR) stimulation and chronic pressure overload as well as against age-related cardiac myopathy (358–362). However, this concept has been challenged (227) and still awaits confirmation by independent research groups. Intriguingly, AC5 deletion does not protect against Gα_q-induced cardiac myopathy (363). AC5 mediates the interplay of calcium and cAMP signaling in the heart under both sympathetic and parasympathetic regulation (364). AC5 deletion has also been suggested to protect against metabolic diseases by increasing insulin sensitivity and glucose tolerance, leading to decreased body weight even on a high-fat diet (365). Deletion or inhibition of AC5 may also protect against oxidative stress and reduce aging (366, 367).

In vitro, in silico, ex vivo, and in vivo experiments investigated the effect of increased extracellular glucose levels (hyperglycemia) on a cAMP signaling pathway in arterial myocytes (368). It is already established that PKA and L-type Ca²⁺ channels are associated with vasoconstriction in response to hyperglycemia, but the mechanism is unclear. This study concludes that the pathway is dependent on AC5 activity. Glucose stimulates cAMP production via AC5 in arterial myocytes, and this activity induces the L-type Ca²⁺ channel via PKA mediation and is required for vasoconstriction. AC5 and L-type Ca²⁺ channels are spatially closely associated in arterial myocytes. This pathway was verified in two established diabetic animal models, which implicates AC5 in the vasomotor response in diabetes. The aforementioned study by Ho and coworkers (365), who observed that AC5 knockout mice have increased longevity due to increased energy metabolism, glucose tolerance, and insulin sensitivity, appears to corroborate these findings and support the conclusion that AC5 links glucose to vasoconstriction.

In contrast to the cardiovascular system and the metabolic system, in the central nervous system AC5 deletions tend to be detrimental, specifically in the reward circuitry of the brain and motor systems. AC5 deletions induce impaired stress response mechanisms, impaired striatal learning and plasticity, impaired motor coordination in a Parkinson’s-like phenotype, and increased alcohol consumption (354, 369–372). Of similar or even greater concern, AC5 deletion causes autism-like symptoms (355). Considering the fact that there is no effective pharmacological treatment of autism (373), such an adverse effect as a result of treatment with an AC5 inhibitor would be unacceptable. As a beneficial effect, AC5 knockouts display an analgesic response to both acute and chronic neuropathic pain (374), suggesting that AC5 inhibition could be exploited for pain treatment, at least conceptually.

Human ADCY5 polymorphisms have been associated with metabolic diseases including diabetes and obesity (TABLE 3) (169–174, 375,376). Most prominently, ADCY5 polymorphisms are associated with neuropsychiatric and central nervous system disorders, notably alcoholism, depression, familial dyskinesia, facial myokymia, dystonia, myoclonus, and choreoathetosis (13, 15, 16, 168, 175, 176). Of particular interest are mutations in the regions of AC5 that link the C1a and C2a catalytic/G protein binding domains to the transmembrane (TM) domains, which are associated with familial dyskinesias (FIGURE 1). Mutations at R418 in the coiled-coil helical domain of AC5 that links TM1 to the C1a domain, at A726 in the C1b domain, and at M1029 in the coiled-coil helical domain linking TM2 to the C2a domain all contribute to dyskinesias (13–16). These ADCY5 polymorphisms result in AC5 variants with elevated catalytic activity causing severe neurological symptoms, providing the conceptual basis for selectively abrogating the elevated catalytic activity with appropriate inhibitors (357).

AC5’s unique regulatory pattern of Gα_i inhibition and Gα_s stimulation is an important mediator in AC5’s physiological effects. Both G proteins have the ability to independently bind with AC5, opening up the possibility of a ternary complex. Recent studies used molecular dynamics modeling and simulations to examine the ternary complex of interactions of AC5 with Gα_i and Gα_s (377). Gα_i profoundly affects structure and flexibility of AC (378, 379). The ternary complex of AC with Gα_i and Gα_s resembles an inactive conformation, suggesting that effects of Gα_i are stronger than those of Gα_s (377) The ternary Gα_i:AC5:Gα_s complex may contribute to coincident detection, particularly in corticostriatal synapses important for reinforcement learning and LTP (380). Modeling the dynamics of this complex reveals that it creates a low catalytic activity of the enzyme in the basal state, facilitating larger and synergistic increases in cAMP that enhance detection of coincident signals. Such coincident detection by AC5 is important for corticostriatal synaptic plasticity, especially for reinforcement and reward-based learning. However, as a caveat, it needs to be emphasized that molecular dynamics studies regarding the effect of Gα_i on AC5 do not replace authentic structural (crystallographic) analyses. The latter studies still need to be performed.

Another recent study presents novel evidence of another physiologically relevant regulatory heterotetramer of AC5. The proposed precoupled complex would notably account for the coincident signal detection by Gα_i and Gα_s (381). Although there is previous evidence of G protein and AC precoupling, this study reports precoupling of transmembrane domains of ACs with transmembrane domains of a GPCR. Using a peptide-based approach, Navarro et al. propose a heterotetramer of an adenosine receptor (A₂AR), a dopamine receptor (DRD2), two AC5 molecules, Gα_i, and Gα_s. When activated, there is rearrangement of membrane domains of the precomplex. The heterotetramer also allows for the canonical antagonistic interactions of Gα_i and Gα_s with AC5. These results are physiologically significant because AC5 plays key roles in both the A₂AR and DRD2 pathways, which are important for neuropsychiatric drugs (multiple GPCR antagonists; previously referred to as antipsychotics) (382). Thus, AC5 may be inhibited indirectly via targeting the GPCRs proximally located in the signaling cascade.

This A₂AR, DRD2, and AC5 homodimer heterotetramer complex has a proposed role in striatal neurons (383). The AC5 homodimer allows for the canonical antagonistic Gα_i/Gα_s interactions on AC5. The heterotetramer integrates both adenosine and dopamine signaling in the brain. The model can explain the physiological and behavioral responses of adenosine and dopamine in the striatopallidal neurons, including caffeine psychostimulation, Parkinson’s disease, schizophrenia, and substance use disorders.

AC5 also may play a role in polycystic kidney disease (PKD) (384). cAMP is known to contribute to PKD by stimulating cell proliferation and fluid secretion. Knockdown of AC5 decreased cAMP levels in renal epithelial cells. In collecting duct-specific Pkd-2 (PKD gene)-deficient mice, there is an increase in cAMP and AC5 mRNA levels in the kidney. AC5 and Pkd-2 double-mutant mice had decreased kidney injury, decreased kidney enlargement, decreased cyst index, and improved kidney function. cAMP levels were reduced in the kidneys of the double mutants. In renal epithelial cells, cilium length was decreased upon ablation of AC5 in the Pkd-2 mutant mice. Overall, these results imply that AC5 inhibition is a viable therapeutic approach for PKD, but, as stated above, selectivity of AC5 inhibitors and the widespread expression of this isoform make adverse effects a serious concern.

5.2. Adenylyl Cyclase 6

Structurally, AC5 and AC6 are very closely related to each other (11). Accordingly, it is not surprising at all that pharmacologically it has been very difficult if not impossible to develop AC6 inhibitors with selectivity relative to AC5 (385). Like AC5, AC6 is also broadly expressed in heart, brain, kidney, pancreas, and bone. Interestingly, despite the large structural similarities between AC5 and AC6 (11), AC5 and AC6 possess distinct functions. Specifically, AC6 deletion in the heart limits β₁AR-stimulated ventricular contractions and calcium signaling (386–388). Although AC5 deletion was protective, AC6 overexpression is protective of heart failure, specifically by increasing cardiac response and contraction and decreasing detrimental left ventricular remodeling (389–392). This difference between AC5 and AC6 in the heart may be due to the fact that AC6 is uniquely regulated by constitutive Gα_i activity. To help tease out these isoform-specific effects on cardiac contractility, AC5 and AC6 knockout mice were compared (393). In particular, it was studied whether Gα_i inhibits cAMP-regulated contractile force in the ventricle in an AC isoform-specific manner (393). With the use of Förster resonance energy transfer (FRET) to determine cAMP levels, in hearts from AC5 but not AC6 knockouts cAMP increased after inhibition of PDE3 and PDE4 in cardiomyocytes. This indicates that AC6 is regulated by Gα_i activity and helps maintain basal cAMP in the compartment that regulates contractility of the rat ventricle. When PDE3, PDE4, and Gα_i are inhibited in the AC6 knockout mice, an increase in contractility occurs. These data demonstrate that the inotropic and diastolic effects are dependent on AC6 and that AC6 controls contractile function in a localized cellular compartment. The study also shows that Gα_i inhibition of AC6 limits the cAMP increase via β₁AR.

The beneficial effects of AC6 overexpression are not just due to increased cAMP signaling, as AC6 also improves cardiac myocyte calcium handling (394), likely through interaction with a specific phosphatase, PH-domain leucine-rich protein phosphatase (PHLPP), that regulates Akt (395). Human gene therapy trials delivering AC6-expressing adeno-associated virus to patients with heart failure improved left ventricular function, particularly in patients with nonischemic failure (396). This is at odds with studies in mice, where transgenic overexpression is not protective against chronic pressure overload, particularly in female mice, and where AC6 deletion is beneficial (397, 398). These differences have not been reconciled but may reflect critical differences between humans and mice. Another explanation may come from recent studies in which just the catalytic domains of AC6 were expressed in transgenic mice (399, 400). These mice were protected from pressure overload-induced systolic and diastolic dysfunction as well as the deleterious effects of isoproterenol infusion. This same group of investigators expressed a mutant form of AC6 that lacks catalytic activity (401). These mice also displayed less dysfunction after isoproterenol infusion. Thus, the physiological effects of AC6 in the heart appear to be more complex than just its ability to catalyze the synthesis of cAMP.

β₂AR agonists are used to reduce airway smooth muscle tone for treatment in asthma and chronic obstructive pulmonary disease, and AC6 has been linked as the primary isoform mediating these effects by studies of AC6 knockout mice (402). In the brain, AC6 is also implicated in hippocampal learning processes and age-related neurodegenerative diseases (403). In bone, AC6 in the primary cilia is involved in loading-induced bone formation (404). In the pancreas, AC6 regulates amylase and fluid secretion (405). In marked contrast to ADCY5, ADCY6 polymorphism studies are limited and have only linked AC6 to sickle red cell adhesion (TABLE 3) (406).

In the kidney, AC6 functions in water homeostasis, renin secretion, urine osmolarity and output, and cyst growth in PKD (407–414). More recently, it was found that AC6 also plays an important role in regulating acid-base homeostasis (415). Globally deleted AC6 mice and renal tubule and collecting duct-specific AC6 knockout mice were used to study AC6’s role in acid-base homeostasis. AC6 knockout mice have lower urinary pH and higher blood pH at baseline, implying that AC6 is required for normal acid-base homeostasis. However, when renal tubule and collecting duct-specific AC6 knockout mice were examined there was no disturbance in baseline acid-base homeostasis, suggesting that extrarenal AC6 has a role in energy expenditure. When the investigators challenged mice with $\mathrm{HC}{\mathrm{O}}_3^{\mathrm{-}}$ they observed alterations in blood pH, supporting the idea that AC6 plays specific roles in the kidney tubule and collecting duct that enable acid-base homeostasis.

AC6 may also play a role in inducing inflammatory responses and chronic pain in various diseases (416). Knockout of AC6 increased TNF release and inflammatory responses by BMDMs, indicating that AC6 is important in inhibiting TNF release. More recently, alpha7 nicotinic acetylcholine receptors (CHRNA7) were found to directly interact with and signal through AC6 to induce anti-inflammatory effects (417). CHRNA7-selective agonists reduced production of inflammatory mediators by macrophages stimulated with LPS. Knockdown of AC6 prevented the anti-inflammatory response to CHRNA7 agonists, and overexpression of AC6 promoted these responses. The authors were able to observe that AC6 activation by CHRNA7 agonists led to degradation of Toll-like receptor 4, accounting for the anti-inflammatory action. Prolonged activation of the inflammatory response is implicated in many chronic diseases such as rheumatoid arthritis and Crohn’s disease, so characterization of AC6’s role in regulating inflammatory pathways may be useful in developing new treatment options.

AC6 is also important in regulating ciliary length, with consequences for both the airway and intestinal systems (418). AC6 knockout mice have longer cilium length and decreased flow in the epithelial cells compared with wild type. The AC6 knockouts exhibit decreased levels of Kif19A, a depolymerizing kinesin that controls cilium length. In vitro studies teased out a pathway in which AC6 inhibits AMP-activated kinase (AMPK), suggesting that AMPK can be more readily activated in the AC6 knockout (418). The AMPK targets Kif19A for autophagy, and the decreased levels of Kif19A then lead to increased cilium length. This study implicates AC6 in an important pathway for regulation of cilial length and motility that could play a role in cilia-related disorders specifically of the respiratory tract, reproduction tract, brain ventricle, and middle ear function. However, the role of AC6 has yet to be directly examined in ciliopathies of these various organs. In addition, a recent study implies that AC3, not AC6, is the primary resident isoform in primary cilia (419).

In intestinal epithelial cells, AC6 is a necessary component of cholera toxin-induced diarrhea (420, 421). In cholera, diarrhea results from increased cAMP levels, which activate cystic fibrosis transmembrane conductance regulator (CFTR)-mediated chloride transport. Thomas et al. (420) first proposed a role of AC6 in cholera toxin-induced diarrhea by using RNA seq to identify AC6 as the predominant isoform that associates with CFTR in epithelial intestinal cells. These investigators then created epithelial AC6 knockout mice and observed an abolishment of the CFTR-dependent fluid secretion. Intestinal epithelial cell-specific AC6 knockout mice displayed similar effects (421). AC6 was essential for cAMP production leading to cholera toxin-induced diarrhea via the movement of water into the intestinal lumen. Together, these results suggest that AC6-CFTR complexes may mediate diarrhea in cholera, and inhibition of AC6 may be useful as a treatment.

Intestinal epithelial cell differentiation appears to be regulated by AC expression. Treatment of colon epithelial cells in culture with sodium butyrate induces epithelial differentiation, which coincides with a large decrease in expression of AC3, AC4, AC6, and AC7 (but not AC5 or AC9) (422). Similar expression patterns were also observed in colon epithelia from mice, albeit involving some different AC isoforms that likely reflect the species differences. Expression of Gα_s was also reduced in these studies, implying that global cAMP signaling is altered during differentiation. The differentiation of epithelia appears to require a reduction in cAMP generating capacity, likely to liberate cells from the antiapoptotic effects of cAMP (422).

5.3. Adenylyl Cyclase 5 and 6 Redundancies

There are many similarities between the expression patterns and functions of AC5 and AC6 (11), leading them to be studied together in double knockouts. AC5 and AC6 are the predominant isoforms in the heart, so most functional redundancies have been studied in this tissue. Both AC5 and AC6 have been previously reported to act in the β₁AR pathway. A recent study used both AC5 and AC6 knockout mice to investigate their roles in β₁AR-mediated inotropic responses by the heart (423). The results indicate a functional redundancy between AC5 and AC6 in mediating β₁AR inotropic responses. Pertussis toxin inactivation of Gα_i reveals that both isoforms are subject to tonic inhibition by Gα_i. In the AC6 knockout, β₁AR switched from predominant AC6 coupling to AC5 coupling. This switch was associated with alteration in the PDE isoform regulating the cAMP pool, indicating that AC6 knockout results in reorganized signaling compartments.

AC5 and AC6 knockout mice have also been used to investigate their respective localizations in ventricular myocytes and their effect on L-type Ca²⁺ current (424). AC6 is differentially localized to the plasma membrane outside of the T-tubular region, and it enhances β₁AR-mediated L-type Ca²⁺ current. In contrast, AC5 is localized to the membrane in T-tubular regions, and its influence on L-type Ca²⁺ current is limited by PDEs. AC5 is responsible for the β₁AR enhancement of L-type Ca²⁺ current, further supporting the hypothesis of compartmentation of AC isoforms in the heart. AKAP5 participates in this βAR-AC5-PKA complex to regulate L-type Ca²⁺ current in the T tubule (424). Cardiac myocytes also express AKAP6 (also known as mAKAP), which interacts with AC5 to coordinate cAMP, calcium, and MAP kinase pathways to regulate cellular hypertrophy (26). These specific AKAP-AC interactions are critical to the spatial regulation of cAMP signaling and encoding specific cellular responses such as cardiac repolarization, calcium release, and cell stress responses (FIGURE 8). There is also evidence for caveolin-3 as a scaffolding protein for AC5 (426). This scaffolding interaction may be important for compartmentalizing the β₁AR signaling pathway.

FIGURE 8.

Adenylyl cyclase (AC)-A kinase anchoring protein (AKAP) signaling complexes regulate specific cardiac functions. AC isoforms associate with specific AKAPs to form signaling complexes. These complexes are localized in specific structures and spaces within the cardiac myocyte, likely based on localization of AC. AKAP interactions with downstream effector and regulator proteins allow a complex to regulate unique physiological functions. Cardiac repolarization is facilitated by an AKAP9 (Yotiao) complex that includes KCNQ1, PKA, protein phosphatase 1 (PP1), and phosphodiesterase 4DE3 (PDE4D3). Disruption of the complex prolongs the action potential. Calcium-stimulated calcium release is regulated by a macromolecular complex consisting of AC5/6 and AKAP5 (AKAP79/150). AKAP5 anchors a cAMP regulatory unit in the transverse tubule, where it is in register with phospholamban (PLN), ryanodine receptors (RyRs), and the sarco(endo)plasmic reticulum calcium ATPase (SERCA). Molecules involved in cardiac hypertrophy, including protein phosphatase 2B (PP2B), PDE4D3, and phospholipase Cε (PLCε), are anchored by AKAP6 (mAKAP) to both AC5 in the transverse tubule and the nuclear membrane via nesprin. The close proximity of the transverse tubules and the nucleus allows functional coupling of the AC5 complex to nuclear calcium signaling. The cardiac stress response is further regulated by binding of AC9 to heat shock protein (HSP)20, facilitating PKA regulation of HSP20 phosphorylation and promoting cardioprotection. LTCC, L-type Ca²⁺ channel. Figure adapted from Ref. 425, with permission from Journal of Cardiovascular Development and Disease.

Collectively, both AC5 and AC6 are broadly expressed AC isoforms. Despite their structural similarities (11), they possess different functions, most prominently in the heart. The different functions of AC5 and AC6 may be attributable, at least in part, to differential cellular localization. Unfortunately, it is still impossible to therapeutically exploit the different functions of AC5 and AC6 because of the lack of AC6 inhibitors with high selectivity relative to AC5 (and, of course, AC isoforms from other AC families) (11). Several studies have exploited the Ca²⁺-inhibitable nature of AC5 and AC6 (427) to infer roles for these isoforms. Such an approach is unable to distinguish between AC5 and AC6 but does establish the involvement of these group III ACs. Endothelial cell gap junction formation and vascular permeability (428–431), regulation of vascular tone (432), and catecholamine-regulated renin secretion (408) are a few prominent examples.

A general problem in the AC field, and a problem particularly relevant for the AC5/6 field, is the paucity of well-characterized and AC isoform-specific antibodies. Some commercial AC5 and AC6 antibodies are available and have been used in certain studies (433, 434). However, in general, authors rely on company descriptions of selectivity. Rarely, antibodies are validated by studying corresponding AC knockout animals, AC knockdown experiments, or recombinant AC isoforms. This is the major reason for the fact that most studies on the (patho)physiological functions of AC5 and AC6 rely on AC5 and AC6 knockout animal studies. An issue not sufficiently known in the scientific community and contributing to the difficulties of valid studies with AC5 and AC antibodies is that the expression levels of these signaling proteins are very low compared with GPCRs and, particularly, G proteins (11).

6. GROUP IV (AC9)

6.1. Adenylyl Cyclase 9

Until recently, what little was known about physiological functions of AC9 was gleaned from ADCY9 polymorphisms and studies of microRNAs that potentially target AC9. Human miR-142-3p and miRNA-181-5p reduce AC9 expression and have implicated AC9 in sepsis (TABLE 3) (435), cervical cancer (436), promyelocytic leukemia (437), and immune function (438). It is of note that AC9 is expressed in many immune cells, including regulatory T cells, neutrophils, and monocytes, and reducing AC9 expression decreases Gβγ-mediated neutrophil chemotaxis (338, 439–442). ADCY9 polymorphism studies have also associated AC9 with asthma (211, 212), pancreatic cancer (213), liver cancer (443), mood disorders (444), body weight (214), malaria susceptibility (215, 216), and efficacy of the cholesteryl ester transfer protein (CETP) inhibitor dalcetrapib (217).

The link between the serum lipid-altering drug dalcetrapib and AC9 is complicated. Clinically, dalcetrapib (and to a lesser extent evacetrapib) had better efficacy and 20% fewer cardiac events when a particular polymorphism of AC9 was present in the second intron of the ADCY9 gene (217, 445). However, the presence of this polymorphism had no effect on the related drug anacetrapib within the REVEAL trial (446). A larger trial, Dal-GenE, will hopefully determine the actual benefit-risk ratio for this genotype (447). In mice, inactivation of ADCY9 by gene-trapping (AC9KO) gave rise to reduced atherosclerosis, macrophage accumulation, and plaque formation after treatment with an atherogenic protocol that consisted of AAV8 viral expression of a gain-of-function mutant of Pcsk9 followed by a high-cholesterol diet to induce hypercholesterolemia (448). AC9KO also increased body mass and fat mass after the atherogenic protocol. However, effects on at These results have implications both for protection from atherosclerosis and for treatment of atherosclerosis.

Mouse studies of AC9KO have also recently uncovered important roles for AC9 in heart, including regulation of stress responses, repolarization, and heart rate control (448, 449). AC9 activity is detected in SAN, and two different laboratories report bradycardia for AC9KO mice, suggesting a role in heart rate control (449). AC9 is also recognized for its ability to associate with the AKAP Yotiao, which is the smallest of the splice variants of AKAP9 (450). Yotiao also anchors PKA and the alpha subunit (KCNQ1) of the slowly activating delayed-rectifier K⁺ current (I_Ks), critical components for the late-phase repolarization of the cardiac action potential in humans during sympathetic stimulation (451). Mutations of KCNQ1 or Yotiao that disrupt their association can cause potentially fatal arrhythmias, known as long QT syndrome. In a transgenic mouse line that expresses the KCNQ1/KCNE1 components of I_Ks, AC9 was the only AC isoform to associate with Yotiao and KCNQ1 (452). Deletion of AC9 resulted in reduced isoproterenol-stimulated KCNQ1 phosphorylation in vivo and in decreased I_Ks currents in adult cardiomyocytes, suggesting that AC9 is necessary for the regulation of the I_Ks potassium current. By bringing together AC9, PKA, and I_Ks, Yotiao facilitates the generation of local pools of cAMP to activate anchored PKA, increase PKA phosphorylation of the channel, and enhance potassium currents necessary for cardiac repolarization (115) (FIGURE 8).

AC9 may also play an important role in basal cardiac stress responses, regulating the PKA phosphorylation of heat shock protein 20 (Hsp20) at baseline but not upon sympathetic stimulation (319). AC9 activity associates with Hsp20 in heart, suggesting that these proteins are in complex with one another. The close interaction of AC9 with downstream targets of PKA (e.g., KCNQ1 and Hsp20) may be required because of the very low levels of AC9 activity in heart compared with the major cardiac isoforms, AC5 and AC6 (319).

AC9 was originally categorized as forskolin insensitive (group IV), but several studies demonstrate that AC9 is weakly conditionally stimulated by forskolin in the presence of Gα_s (72, 453). Based on cellular studies, AC9 is currently thought to be stimulated by Gα_s, PKC βII, and calcium calmodulin kinase II (CaMKII) and inhibited by Gα_i/o, novel PKC isoforms, and calcium calcineurin (78, 454–456). However, direct regulation of AC9 was limited to Gα_s (although regulation by calcineurin could not be tested) (72). The other regulators likely either act indirectly or require additional scaffolding proteins in order to facilitate regulation. AC9 can also form homodimers and can heterodimerize with Gα_i-regulated AC5 and AC6 when overexpressed, potentially adding to the complication of interpreting cellular overexpression studies of AC9 (72).

Interestingly, a regulatory method of autoinhibition for AC9 has been proposed with physiological implications (457). A short motif in the C2b regulatory domain of AC9 was identified that reduced Gα_s activation. Suppression was lost upon proteolytic cleavage of AC9 at the COOH terminus. The resulting truncated AC9 protein was prevalent in the left ventricle of the heart, as detected by antibodies. This unusual mechanism was further elucidated with the solution of the cryogenic electron microscopy structure of full-length AC9 bound to Gα_s (10). Density from a region from the COOH-terminal C2b domain of AC9 (aa 1246–1275) was present within the active and forskolin binding sites. By comparing a truncated AC9 with the full-length protein, it was clear that this region had little effect on basal activity, but its presence dramatically increased the K_m for ATP by 18-fold when the enzyme was activated by Gα_s (48 vs. 880 μM ATP). Thus, proteolytic cleavage of AC9 may regulate GPCR activation of AC9 activation in heart (457, 458).

Another mechanism of regulating GPCR activity relates to the trafficking of GPCRs. All transmembrane ACs are present on the plasma membrane (PM), although localization to internal membranes has certainly been reported (reviewed in Refs. 11, 459). This is consistent with the need for GPCRs to couple G protein activation and AC regulation at the PM. However, upon ligand binding, many GPCRs internalize yet continue to activate G proteins and enhance cAMP accumulation at endomembrane sites (131) (reviewed in Ref. 116). AC9, but not AC1, can internalize with hormone-stimulated β₂AR (460). Although both the GPCR and AC9 utilize a dynamin-dependent membrane pathway, AC9 traffics independently of the GPCR. Gα_s activation was sufficient to promote AC9 accumulation in endosomes; the trafficking of AC9 did not require β-arrestins or AC activity (FIGURE 9). Thus AC9, and potentially other AC isoforms, may promote cAMP generation from internal sites to differentially regulate downstream events.

FIGURE 9.

Internalization and endosomal signaling by adenylyl cyclase (AC)9 but not AC1. β₂ adrenergic receptors (β₂ARs) activated by agonist stimulate cAMP at the plasma membrane but then internalize via arrestin-mediated endocytosis into endosomes. When complexed with AC1, the receptor ceases cAMP signaling once internalized, creating a short-duration cAMP signal delimited to the plasma membrane. When complexed with AC9, the receptor internalizes with the AC and continues to generate cAMP in the endosomal compartment (131, 460). AC9 internalization occurs independently of the receptor and does not require arrestins or AC activity (460).

7. GENETIC REGULATION OF AC ISOFORMS

The genetic structure and chromosomal location of the different AC isoform genes have been studied (6, 56, 461–466). However, details about the genetic regulation and promoters of those genes are still largely missing (48, 56). Studies employing sequence homology and expressed sequence tags (ESTs) have shown that human AC genes are formed of 11–26 exons and have a length of 16–430 kb (56). The genetic organization of AC isoforms appears to be well conserved among group II and III ACs, with ADCY5 and ADCY6 presenting 21 exons and ADCY2, ADCY4, and ADCY7 being composed of 25, 25, and 26 exons, respectively. The numbers of exons and introns are less conserved among group I ACs, with 20, 22, and 18 exons for ADCY1, ADCY3, and ADCY8, respectively. ADCY9 has 11 exons (56).

Regarding the promoter regions of human AC genes, ADCY2, ADCY3, ADCY5, ADCY6, ADCY8, and ADCY9 do not contain a TATA box, suggesting that the expression of these isoforms is not regulated by TATA box-binding factors (56). ADCY1, ADCY2, ADCY3, ADCY4, ADCY5, and ADCY9 present putative GC boxes, which may be regulated by transcription factor specificity protein 1 (Sp1) (56, 467). In addition, the translation initiation ATG codons are only within the context of good Kozak sequences for ADCY6 and ADCY9, which may result in lower expression levels of the other AC isoforms. The promoter regions of rat ADCY3 and mouse ADCY8 possess a CRE motif and therefore can be regulated by changes in intracellular cAMP (468–470). Notably, point mutations in the promoter region of ADCY3 (in sites different from the CRE motif) have been linked to decreased glucose-induced insulin secretion in a rat model of diabetes (469).

In rats, the expression of AC8 is also enhanced by the farnesoid X receptor (FXR), a nuclear receptor that binds to the promoter region of ADCY8 (471). The expression levels of both AC8 and FXR are reduced in a rat model of diabetes compared with control animals. A more detailed study analyzed the genetic regulation of mouse AC1. In addition to reporting the lack of a TATA box, promoter sequences for potential binding of glucocorticoid receptors, POU transcription factors, the immediate-early gene zif268, Gcn4, E-box-binding factors, as well as the previously mentioned Sp1 were identified (472).

The methylation state of the promoter regions of ADCY3 and ADCY4 have been studied. For ADCY3, hypomethylation has been detected in gastric cancer cell lines, where AC3 transcript levels are increased (473). For ADCY4, hypermethylation of the promoter region has been shown in patients with breast cancer compared with control subjects (317). This profile was correlated with a decrease in AC4 mRNA transcripts in those patients.

Processing of mRNA through alternative splicing represents another natural method for regulating enzymatic activity. Notably, splice variants for AC isoforms have also been reported (56, 474–478). Some variants, such as those from AC4, AC5, AC6, and AC8, lack certain exon regions and may present different activity compared with the full enzymes (56). In the case of AC8, short variants dimerize with full-length AC8 to downregulate cAMP production (475). Notably, splice variants for AC1, AC5, and AC9 were reported to encode for nearly half of the AC (56, 474). These near-half variants may serve as another way to regulate AC activity and cAMP signaling.

8. CONCLUDING REMARKS

ACs have the potential to regulate a vast array of physiological responses since their enzymatic product, cAMP, regulates an immense number of cell functions. The fact that most cells express several of the nine AC isoforms creates a redundancy of function that makes it difficult to discern the role of one specific isoform. Drugs to specifically stimulate or inhibit a single AC isoform have largely been lacking, making knockdown and overexpression the mainstay of understanding the physiological roles of an AC. A number of early biochemical studies of ACs defined how these isoforms are regulated by different coincident signals, allowing a framework for understanding how different isoforms could be central to the physiological functions of highly differentiated cells. More recent studies make clear that these isoforms exist in distinct signaling complexes where they couple to specific upstream receptors and generate compartmentalized cAMP pools downstream. The concept that ACs are not redundant but are capable of playing roles in specific physiological responses at the cellular, tissue, and organismal levels is now widely accepted. As such, studies to examine these specific responses are beginning to accelerate.

Resolving AC and cAMP compartmentation in a spatial and temporal manner will help further tease out the physiological roles of the AC isoforms. Historically, measures of cAMP levels in cells have been limited to end-point assays that, in most cases, required the inclusion of broad PDE inhibitors so that cAMP could accumulate to levels sufficient for detection. These whole cell measures of cAMP do not reflect the biochemical reality that the second messenger activates local effectors, is degraded locally, and does not freely diffuse throughout the cell (479–481). cAMP biosensors, most of which are based on FRET, have been developed recently that allow the detection of cAMP in living cells and can be monitored in real time (482–484). FRET-based approaches are limited by relatively low efficiency, so other approaches are needed for more sensitive and rapid readouts. One approach uses cyclic nucleotide-gated channels to report cAMP levels in near-membrane regions (485, 486), whereas others have developed fluorescent biosensors that quench or unquench upon binding cAMP (487). Targeting these various biosensors to discrete subcellular compartments can report localized cAMP dynamics, representing a major advance in understanding compartmentalized GPCR-AC-cAMP signaling (108, 113, 488–493). Combining these powerful new cAMP detection tools with molecular interventions that disrupt expression of specific AC isoforms should accelerate the discovery of how individual ACs regulate specific physiological processes.

GRANTS

Support was provided by a New Investigator Award from the American Association of Colleges of Pharmacy and an IntegraConnect Grant from the Lloyd L. Gregory School of Pharmacy at Palm Beach Atlantic University (T.F.B.); NIH Grant GM60419 (C.W.D.); NIH Grants NS119917 and NS111070 and the Purdue University College of Pharmacy (V.J.W.); and NIH Grant GM107094 (R.S.O.).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

K.F.O., T.F.B., V.J.W., and R.S.O. prepared figures; K.F.O., J.E.L., T.F.B., R.S., C.W.D., V.J.W., and R.S.O. drafted manuscript; K.F.O., J.E.L., T.F.B., R.S., C.W.D., V.J.W., and R.S.O. edited and revised manuscript; K.F.O., J.E.L., T.F.B., R.S., C.W.D., V.J.W., and R.S.O. approved final version of manuscript.