A signal transduction blind spot: the function of adenylyl cyclase transmembrane domains

Abstract

Signal transduction of external primary signals into intracellular elevations of the second messenger cyclic AMP is an ancient and universal regulatory mechanism in biology. In mammals, 9 of the 10 adenylyl cyclases (ACs) share a common topology that includes a large transmembrane (TM) domain assembled from two clusters of six helices. This domain accounts for ~ 35% of the coding sequence but, remarkably, its function is still an open question. In this viewpoint, we consider how the first primary AC sequences spurred ideas for the purpose of AC TM domains, including voltage‐sensing and transporter functions. In the original conceptions of second messenger signalling, ACs were put forward as potential receptors, and we discuss emerging evidence in support of this function. We also consider growing evidence that cyclase TM helical bundles help to organise multiprotein signalling complexes by engaging in interactions with other membrane‐embedded proteins. Cyclase TM regions are more diverse between isoforms than the catalytic domain—we conclude by considering how this might be exploited in therapeutic strategies targeting specific cyclase isoforms.

Transmembrane adenylyl cyclases (tmACs) are essential for synthesising the universal second messenger cyclic AMP. All tmACs share a common topology that includes a large 12‐helix transmembrane (TM) domain. This domain accounts for ~ 35% of the coding sequence but its function is still an open question. In this viewpoint, we consider whether cyclase TM domains might function as transporters, voltage sensors, receptors and interaction modules.

Abbreviations

ABC

ATP‐binding cassette

AC

adenylyl cyclase

AKAP

A‐kinase anchoring protein

CAI‐1

cholerae autoinducer‐1

cAMP

cyclic 3′‐5′ adenosine monophosphate

CB1

cannabinoid receptor 1

CREB

cAMP response element‐binding protein

GPCR

G protein‐coupled receptor

GWAS

genome‐wide association study

NBD

nucleotide‐binding domain

PDE

phosphodiesterase

PKA

protein kinase A

PPAR

peroxisome proliferator‐activated receptors

SDPA

1‐stearoyl‐2‐docosahexaenoyl‐phosphatidic acid

TM

transmembrane

tmAC

transmembrane adenylyl cyclase

Introduction

Cyclic adenosine 3′,5′‐monophosphate (cAMP) is a universal second messenger that mediates intracellular responses including changes in synaptic strength, cell contractility, permeability, migration and differentiation [1]. In the first conception of a second messenger in the 1960s, membrane‐embedded adenylyl cyclase (AC) was ear‐marked as a potential receptor for hormones (primary messengers), with the cyclase directly transducing these signals into the production of the intracellular second messenger cAMP [2]. Gradually a canonical pathway has developed with extracellular signalling molecules acting on G protein‐coupled receptors (GPCRs) that activate transmembrane ACs (tmACs) via G proteins [3]. There are nine tmACs in mammals—all nine share a similar structure (Fig. 1A) [4, 5]. The cyclases are comprised of two major regions—catalytic (green) and transmembrane (TM, blue) regions that are linked by ~ 40‐amino acid extended helices (grey). Both catalytic and TM regions are assembled from two halves that are separated in the primary sequence [6, 7]: subdomains C1A and C2A form a pseudo‐dimer that constitutes the catalytic domain (green, Fig. 1A) whereas two clusters of 6 helices assemble into a contiguous 12‐helix TM domain (blue) [8, 9, 10]. The first report of an adenylyl cyclase primary sequence in 1989 triggered a wave of studies focusing on possible functions for the TM regions beyond membrane tethering [11]. These studies were spurred by sequence similarities to ATP‐binding cassette (ABC) transporters [11] and from the intuition that evolution would be unlikely to maintain ~ 35% of the coding sequence to a domain that served no purpose beyond what could be achieved with a short signal sequence. The importance of these domains is underscored by several pathological mutations located in AC TM domains (Fig. 1B) [12, 13, 14, 15, 16, 17]. However, with no clear function emerging, research has focused on more tractable aspects of AC regulation involving intracellular elements. For example, the catalytic mechanism is well understood [4], as are regulation of the catalytic region by direct binding to G proteins [6] and the basis of interactions between intracellular cyclase elements and A‐kinase anchoring proteins (AKAPs) that organise cAMP signalling in subcellular compartments [18].

Fig. 1.

Transmembrane adenylyl cyclase (tmAC) topology and locations of transmembrane (TM) domain pathogenic mutations. (A) Mammalian tmAC structure showing TM (blue), helical (grey) and catalytic (green) domains. The structure corresponds to AC8 (PDB 8BUZ). (B) Pathogenic mutations falling within the TM domain are shown mapped onto the AC8 structure (PDB 8BUZ) with reference to the AC5 cryo‐EM structure (PDB 8SL3), and AlphaFold models for AC2, 3 and 6. The mutations have the following disease associations: AC2 V147L, bipolar disorder [12]; AC3 S107P, childhood obesity [13]; AC5 P399L, early‐onset generalised dystonia [14]; AC5 E908K, spastic paraparesis [15]; AC6 R739W, autism [16]; and AC8 A728P, schizophrenia [17]. Structural representations were generated using pymol (Schrodinger Inc, New York, NY, USA).

The last 5 years have seen the first high‐resolution structures of full‐length tmACs [8, 9, 10, 19]. The structures confirm that the TM domain is physically separated from the catalytic region and comprised of two integrated six‐helix bundles, with each bundle connected to half of the catalytic domain via half of a coiled‐coil helical domain (Fig. 1A). The structures have the potential to serve as a catalyst for finding answers to the old problem of assigning a function to the mammalian cyclase TM domain. They provide more information to assess the likelihood of different theories and simply serve as a potent visual reminder that the problem still exists. The aim of this viewpoint article is to introduce different possible functions put forward for cyclase TM domains before considering two functions in detail and how this research area could support isoform‐specific AC pharmacology.

Origin of putative functions for the TM domain

The first mammalian tmAC primary sequence revealed a striking topological similarity to the transporter P‐glycoprotein [11]. P‐glycoprotein is a prototypic member of the ABC transporter family, now known to contain 48 members in humans [20], and all transporters in this family share the same tmAC‐like topology. The ATP hydrolysis site in the transporters is located between two catalytic subdomains—identical to the structural arrangement of the tmAC catalytic interface for ATP cyclisation. ABC transporters also contain polytopic TM domains comprised of two integrated six‐helix bundles. ATP hydrolysis drives the transition of the ABC transporter TM domain between inward and outward‐facing conformations, triggering the translocation of substrates across the plasma membrane [21]. This similarity hints at a transporter function in mammalian tmACs (Fig. 2A)—if so, what might be the cargo? Interestingly, the Dictyostelium discoideum (D. discoideum) ABC transporter, abcB3, is one of several transporters known to export cAMP as a chemotactic signal for aggregation [22]. There are some emerging functions for extracellular cAMP in mammals, however, four ABC transporters including MRP4 are already thought to export cAMP [23]. Structures of MRP4 have been determined at different stages of the export cycle [24], focusing on the transporter's ability to export prostanoids. In the apo state with neither ATP nor substrate bound, the two pseudo‐dimeric halves of the transporter flex open. In this state, the two nucleotide‐binding domains (NBDs) are separated by ~ 25 Å, and a deep wedge‐shaped cavity is revealed between the two TM subdomains. The substrate binds at the base of this cavity, bringing about partial closure of the clamshell‐like structure—closure is completed upon binding of ATP to the NBDs [24]. It is notable that no substrate exit channel is evident in the closed ATP‐bound MRP4 structure [24]. To date, all tmAC structures resemble the ‘closed’ MRP4 conformation, with the catalytic subdomains bound to one another and the two TM subdomains oriented in parallel. We await tmAC structures in the absence of activators that might reveal whether the cyclases can open to reveal cavities on the internal face of the TM domain that could support transport. After 35 years, there are still no reports of any transporter activity for a mammalian tmAC, and it should be noted that the topographical similarity to ABC transporters is not complemented by substantial primary sequence similarity.

Fig. 2.

Putative functions of the adenylyl cyclase (AC) transmembrane (TM) domain. (A) ATP hydrolysis could be coupled to a change from an outward to an inward‐facing pose to support transport of a molecule across the membrane. (B) Changes in membrane polarisation could be sensed by the TM domain to alter the sensitivity of the catalytic regions to, for example, regulatory G proteins. (C) The TM domains may serve a receptor function, for example, enabling allosteric inhibition as shown. (D) The 12 helices of the TM domain are well‐placed to engage in interactions with helices present in other integral membrane proteins including GPCRs and ion channels.

Krupinski et al. [11] also immediately noted a similarity to ‘dihydropyridine‐sensitive Ca²⁺ channels and K⁺ channels’. The possibility of an ion channel or voltage‐sensor function was reinforced by the report of a Paramecium tetraurelia AC that responds to membrane hyperpolarisation [25]. Remarkably, the Paramecium cyclase is not only regulated by the membrane potential but is itself the ion channel responsible for setting the resting potential [25]. However, it appears that these dual‐function cyclases are restricted to protozoans [25, 26]. The TM domains of ion channels typically contain channels lined by charged amino acids that enable ion flux—such features are absent in mammalian tmACs. Could the TM domains at least enable cyclase regulation via changes in membrane potential (Fig. 2B)? Experiments in primary neurons from rat cerebellum and hippocampus showed that KCl‐mediated membrane depolarisation can cooperate with isoproterenol to elevate cAMP [27], suggestive of the presence of a voltage‐sensing cyclase. However, it was subsequently shown that this effect resulted from Na⁺ entry through L‐type Ca²⁺ channels [28]. Voltage‐gated ion channels are typically activated by positively charged ‘S4’ TM helices that move in response to changes in membrane potential [29]. Although such sequences are absent in tmACs, a voltage‐sensing function cannot be ruled out since the ligand‐sensitivity of several GPCRs is regulated by membrane depolarisation in the absence of S4‐type helices [30].

The following sections focus on functions where the balance of evidence indicates that TM domains can act in this way to at least some extent: as receptors for ligands whose binding is coupled to changes in cyclase activity (Fig. 2C); and as modules for interaction with other integral membrane proteins (Fig. 2D).

Receptor function

How plausible is it that binding of a ligand to the TM region could be coupled to either inhibition (Fig. 2C) or activation of cyclase activity? Experiments with chimeric and split tmACs demonstrate that both halves of the TM domain must be properly aligned to support cyclase activity. Isolated C1 and C2 subdomains from AC8 co‐expressed in HEK293 cells are not able to combine to support catalysis whereas co‐expression of AC8 M1‐C1 and M2‐C2 fragments generates an active cyclase [31]. Reinforcing the idea that the two TM subdomains are able to dictate whether the catalytic domains are positioned in a conformation compatible with catalysis, AC5/AC6 chimeras containing swapped TM subdomains exhibit normal localisation and expression levels but are effectively catalytically dead [32]. The helical domain appears well‐placed to conduct structural changes between the TM and catalytic regions. Interestingly, several pathogenic mutations have been identified within the helical domain [7] including gain‐of‐function mutations at residue R418 in AC5 that have been linked to familial dyskinesia [8]. Studies of helical domain mutations in Mycobacterium intracellulare AC Cya [33], and D. discoideum AC [34] reinforce the idea that changes in the conformation of the helical domains can control cyclase activity in both directions. The equivalent domain is also critical for catalysis in guanylate cyclases [35]. Proof‐of‐concept studies with synthetic cyclases sensitive to TM domain ligands reinforce the idea that a receptor function for the TM domains is possible. Swapping the TM domain of the Mycobacterium tuberculosis AC Rv1625c for the equivalent region of the Vibrio harveyi quorum‐sensing histidine kinase CqsS generates a chimeric cyclase that can be directly stimulated by the lipophilic ligand cholerae autoinducer‐1 (CAI‐1) [36]. A similar strategy has also been applied to AC2. In this case, exchanging the TM domains with the equivalent CqsS sequence generates a cyclase that is inhibited by CAI‐1 [37]. Taken together, these studies indicate that structural changes elicited by a TM domain ligand could conceivably be transmitted to the catalytic domain in mammalian tmACs to either increase or decrease cyclase activity.

A limited number of molecules have been put forward as allosteric modulators that act through the TM domain. Miconazole, and the related fungicidal imidazole‐containing compounds econazole and clotrimazole, have been shown to alter activity in HEK293 membrane extracts, with a proposed site of action outside of the catalytic domain [38]. Whereas miconazole inhibits AC1 and AC2 and activates AC9, clotrimazole inhibits AC1 and AC9, and econazole stimulates AC9 [38]. It should be noted that there is no conclusive evidence that these compounds act via the TM domain, and the related imidazole compound calmidazolium inhibits a cyclase construct limited to the catalytic region [39]. In recent years, the Schultz laboratory has identified several lipids that alter AC activity. Lipids are obvious candidates for TM domain modulation given their ability to regulate many other integral membrane proteins [40]. Building on the observation that G_sα stimulation of several tmACs is inhibited by serum [37], both AC inhibitors and activators have been identified in lipid extracts from serum. The fatty acid 1‐stearoyl‐2‐docosahexaenoyl‐phosphatidic acid (SDPA) was found to potentiate G_sα‐stimulated AC3 activity [41] but it did not affect an isolated catalytic domain pairing (AC1 C1 subdomain with AC2 C2 subdomain). The authors proposed an intracellular TM domain binding site in this study [41]. Heme b was identified in a low‐pH extract from the lung as an inhibitor of multiple cyclase isoforms [42] and hemin (heme b chloride) inhibited multiple AC isoforms in HEK293 cells. Hemin inhibition of membranes containing cyclase was found to be reversible, suggesting that heme exerts its effects by binding directly to ACs rather than by remodelling of the membrane. Most recently, several fatty acids related to oleic acid were shown to alter tmAC activity [43]. For example, oleic acid elevates cAMP in HEK293 cells expressing AC3 but not AC5. Conversely, arachidonic acid selectively attenuates AC1 and AC4. The physiological relevance of AC regulation by fatty acids is not immediately clear, since circulating levels of free fatty acids are low [44]. Despite these exciting discoveries, the field awaits definitive mapping of any cyclase modulator to a TM domain binding site. Furthermore, no conventional primary messengers, such as a hormone or neurotransmitter, have yet been shown to allosterically modulate AC activity via the TM domain.

Although TM domain sequence conservation across the tmAC family is low [45], more highly conserved pockets and surfaces are evident in alignments across species for specific isoforms [45]. For example, Fig. 3A shows sequence conservation computed using ConSurf [46] overlaid on the structure of AC8 [9]. Intracellular surfaces including the helical and catalytic domains are more highly conserved (purple), whereas the extracellular face of the TM domain is the least conserved (green) with an intermediate level of conservation for the membrane‐embedded surfaces of the TM domain. However, at least four clusters of higher conservation are evident on the AC8 TM domain surface: along helix 1 (region a, Fig. 3A), at the C‐terminus of helix 11 (b), at the C‐terminus of helix 8 (c), and in a pocket formed by helix 4 and a loop between helices 3 and 4 (d). Regions b‐d all include negatively charged side chains (Fig. 3B)—region d is particularly promising as a potential ligand binding site [9]. Molecular dynamics simulations with Rv1625c suggest that a similar negatively charged pocket in this cyclase can bind to positively charged ligands including divalent metal ions [47]. Research on neurosteroid regulation of GABA‐A receptors is instructive [48]. Progress in this field was laboured with extensive use of chimeras, site‐directed mutagenesis and photo‐affinity labelling applied to map binding sites before the advent of cryo‐EM structure determination enabled more conclusive answers. Mapping ligands to AC TM domains should theoretically proceed faster since tmAC research has already decisively entered the cryo‐EM structure era.

Fig. 3.

Surface analysis of AC8. Side‐on (left and middle columns) and top‐down (right column) views of the AC8 surface are shown coloured by either sequence conservation (A) or charge (B). Conservation scores for AC8 (PDB 8BUZ) were computed using ConSurf [46], and structural representations were generated using ChimeraX [79]. Homologues with > 50% sequence identity were obtained from the Clean UniProt database, with a total of 482 sequences included in the analysis. The average pairwise distance was 0.36. Four clusters of higher conservation are highlighted: along helix 1 (region a), at the C‐terminus of helix 11 (region b), at the C‐terminus of helix 8 (region c), and in a pocket formed by helix 4 and the loop linking helices 3 and 4 (region d).

Interactor function

Pre‐assembled signalling complexes are an emerging general feature of signal transduction [49]. cAMP signalling is organised in nanodomains, with close coordination of cyclases, phosphodiesterases and cAMP receptors in complexes often nucleated by AKAPs [50]. However, given that cyclases are the source of all cAMP signals, it might be expected that these enzymes would play a more prominent role in coordinating multiprotein complexes for precise spatiotemporal regulation of cAMP signalling [51]. Given that many important channels and transporters—in addition to all ~ 800 GPCRs [52]—feature bundles of TM helices, there is tremendous scope for interactions between these elements and AC TM helices (Fig. 2D). Multiple tmACs interact with AKAPs including Yotiao [53, 54], AKAP79 [55, 56] and mAKAPβ [57] but these interactions typically involve the cytosolic N‐terminal regulatory regions of the cyclases. There are a few reports of cyclase interactions involving the TM domain. Bioluminescence resonance energy transfer (BRET) and bimolecular fluorescence complementation show that AC5 interacts with both adenosine 2A (A_2A) and type‐2 dopamine (D₂R) receptors [58]. Furthermore, disruption experiments with peptides derived from specific TM helices suggest that the interface between AC5 and these receptors depends on the receptor activation state. For example, in the presence of the A_2AR agonist CGS21680, peptides corresponding to any of the first six TM helices from AC5 disrupt AC5‐A_2AR interactions, whereas helices 1 and 12 are the most effective disruptors prior to receptor activation [58]. AC9 interacts with POPDC1/2 in the heart through an interaction that involves both TM and cytosolic elements in POPDC [59]. POPDC is thought to bridge AC9 to TREK‐1 channels in this context [59], although the details of the membrane interface have not been established.

Dimerisation is an important feature of tmAC regulation that involves TM domain interactions at least for some isoforms. AC5 undergoes cooperative activation that can be explained by binding to an AC5 dimer [60]. Experiments with truncated cyclases show that the second TM cassette mediates dimer formation for AC8 [31], although a recent homodimeric structure of AC5 suggests that a hairpin loop between the fifth and sixth helices of the C1B domain forms the primary interface rather than the TM helices [8]. Another structure from this same study—a monomeric AC5–Gβγ complex—reveals another interesting potential function for the TM domain. The lipophilic geranylgeranyl group attached to the C‐terminus of Gγ packs against helices 1 and 7 of the TM domain and may thereby support the interaction of the two proteins. It should be noted that although several AC isoforms are localised to lipid rafts, surprisingly the C1 and C2 subdomains are thought to be critical for targeting AC5 and AC8 to these cholesterol‐rich regions [61].

Identifying and studying protein–protein interactions in membranes is challenging [62] not least because complexes may dissociate when cellular membranes are solubilised during conventional techniques for studying protein interactions. There is a growing awareness that GPCRs are regulated by membrane‐embedded accessory proteins [63]. For example, receptor activity‐modifying proteins (RAMPs) are single‐pass proteins that bind family B GPCRs [64]. RAMP2 acts as a negative allosteric modulator of the glucagon receptor with interactions involving membrane‐spanning helices [65]. Similarly, the single‐pass accessory protein MRAP2 biases ghrelin receptor signalling away from β‐arrestin [65]. Research on ion channels including AMPA receptors is also revealing that membrane‐spanning accessory proteins are critical regulators [66]. There are likely to be many unexplored areas in the full landscape of cyclase regulation by membrane accessory proteins. New in situ labelling methods are likely to support progress in this endeavour [67].

Implications for pharmacology

Many potential applications of therapeutic agents targeting ACs have been proposed, for example AC1 is a putative target for treating chronic pain, and AC5 has been put forward as a target for cardioprotection [68]. However, the suitability of ACs as drug targets is limited unless isoform‐selective agents can be developed [68]. A handful of previous screens have been conducted to identify compounds that act on tmACs within crude membranes extracted from HEK293 cells [69, 70, 71]. These screens have identified inhibitors of AC1 [69, 70] and AC2 [71] that compete with ATP for binding to the catalytic domain. A further study focused on fungicide imidazole compounds and identified clotrimazole as a pan‐AC inhibitor without determining its locus of action [38]. Sequence divergence is much higher between tmAC isoforms in the TM than the catalytic subdomains [45], therefore agents that target the TM region are more likely to possess isoform‐specific characteristics. One can speculate on different approaches that could be taken. For example, if drugs can be developed that alter cyclase activity via allosteric sites on the TM domain (Fig. 2C), they might be applied in isolation to selectively elevate or depress cAMP signalling in the desired system. Alternatively, they could be applied in combination with GPCR agonists/antagonists to bias signalling towards either the G protein or arrestin branch [72]. TM domain binders that do not alter cyclase activity might also be useful—they could be fused to GPCR‐directed drugs to increase specificity. Many antibody‐drug conjugate agents are in development [73] including anti‐CD33 antibody conjugated to calicheamicin through acid‐labile linker for the treatment of acute myeloid lymphoma [74]. The recent structure of AC5 highlights how multiple structurally complex polypeptide loops extend outside of the cell membrane for this isoform, which could be amenable to antibody targeting [8]. Finally, interactions mediated by TM domains (Fig. 2D) might be targeted by peptide or small molecule disruptors. This approach has received attention from researchers focused on AKAPs. For example, a peptide has been developed to disrupt interactions between TRPV4 channels and AKAP79 with the aim of treating chronic pain [75], and proof‐of‐concept experiments with a peptide targeting the intracellular AKAP79‐AC5 interface support the viability of this approach [76].

AC TM domains are emerging as important protein interaction hubs, and evidence is accumulating that the domains can serve as receptors. To date, cyclases have been neglected as drug targets in comparison to other components of the cAMP signalling cascade including GPCRs [77] and PDEs [78]. Further efforts to illuminate the AC TM domain function ‘blind spot’ may enable therapeutic strategies targeted at these domains to partly redress this balance in the future.

Conflict of interest

The authors declare no conflict of interest.

Author contributions

MGG and RSD wrote the manuscript together; MGG prepared the first two figures; RSD performed the surface analysis of AC8.