Synergistic Interplay of Stimulatory Cofactors in Activation of Adenylyl Cyclase Isoform 1

Abstract

Adenylyl Cyclase isoform 1 (AC1), responsible for synthesizing the signaling molecule cyclic adenosine monophosphate (cAMP), is key in synaptic plasticity, long-term chronic pain syndromes, osteosarcoma-associated pain, and drug abuse. The protein calmodulin (CaM) and the small molecule forskolin (Fsk) stimulate AC1 to form a catalytic site for ATP catalysis; however, molecular AC1 structural changes triggered by CaM and Fsk remain poorly understood. This study developed a computational model for AC1-cofactor complexes and used all-atom molecular dynamics (MD) simulations to determine how CaM and Fsk individually and jointly affect AC1 dynamics and assess whether they exhibit any synergistic effects. Four systems were investigated: AC1-No Partner, AC1-CaM, AC1-Fsk, and AC1-CaM-Fsk. Simulations revealed that individual and joint cofactor binding induced unique structural changes within the regulatory C1b subdomain of AC1. CaM and Fsk binding results in a reduced cross section of the catalytic site, implying a tighter binding for ATP. Notably, results showed that the AC1-CaM-Fsk system exhibited unique features distinct from the AC1-CaM and AC1-Fsk systems, demonstrating synergistic effects of CaM and Fsk. Our understanding of AC1-cofactor interactions can guide future research towards modulating AC1 activity, potentially contributing to the development of novel treatments for AC1-associated diseases.

Graphical Abstract

Introduction

Adenylyl Cyclases (ACs) catalyze the conversion of adenosine triphosphate (ATP) to the secondary messenger cyclic adenosine monophosphate (cAMP).¹ The cAMP signaling pathway is a highly conserved cellular signal transduction pathway integral to cellular function throughout all kingdoms of life,² playing key roles in cell signaling-associated disease pathways including autoimmune disorders, chronic inflammation, and cancers including pancreatic adenocarcinoma.^3–5 Given the physiological significance of the cAMP signaling pathway, understanding the behavior of its components, specifically G-protein coupled receptors (GPCRs) and ACs, can increase understanding of multiple disease mechanisms. As the primary effectors of GPCRs,^6–8 ACs are modulated by G proteins and small molecule forskolin^1,2 There are nine known human membrane-bound AC isoforms, ranging from 1080 to 1353 residues in length.⁹ Despite sharing a similar structure, they are differentially modulated by distinct regulatory molecules and are associated with vastly different physiological processes due to differential expression, including inflammatory response, memory, behavior, metabolism, and heart muscle contraction.^1,2 This diversity in isoform-specific regulation poses a significant challenge to elucidate the precise molecule mechanisms underlying AC activation.

Of the nine membrane-bound AC isoforms, of particular interest is Adenylyl Cyclase isoform I (AC1), an AC isoform which is highly expressed in the nervous system and has been established as a target for modulating chronic pain and opioid use.^1,10 AC1 is sensitive to the activation of the intracellular calcium signal, which promotes the production of cAMP, generating this response through NDMA receptors and voltage-gated ion channels which activates downstream processes such as ERK and PKA to phosphorylate transcription factor CREB and various target proteins that influence synaptic plasticity and long-term potentiation.¹¹ Recent studies have associated AC1 function with long-term chronic pain syndromes (such as migraines and visceral pain), spatial memory, psychiatric disorders (such as long-term depression and hyperactivity), and drug abuse,^1,2,11–13 suggesting that AC1 could be a promising drug target for a plethora of pain and neurological function-associated diseases.¹ In mouse models, inhibition of AC1 by a small molecule, ST03437, has been found to reduce inflammatory pain and opioid-induced hyperplasia.¹⁴ In another study, inhibition of AC1 activity via the NB001 inhibitor alleviated osteosarcoma-induced pain.¹⁰ Although these inhibitors of AC1 have been identified, mechanisms of AC1 inhibition at a molecular level—such as the structural changes induced and properties of AC1 in the active state—have not been elucidated. However, understanding the molecular properties of AC1 behavior and activation can better elucidate the physiological role of AC1 as a potential drug target for several human diseases.

To understand the molecular mechanisms of AC1 behavior, it is necessary to understand the structural properties and activation pathways of AC1. AC1 contains 1119 residues, split into five functional domains (Figure 1A): the N-terminus, transmembrane domains TM1 and TM2, and cytoplasmic domains C1 (residues 234–605) and C2 (residues 806–1119).^9,15–18 The C1 region is further split into the catalytic subdomain C1a, residues 234–469, and the regulatory subdomain C1b, residues 470–605 (C1a and C1b regions were determined by aligning the AlphaFold predicted structure of human AC1 to that of bovine AC1).^9,15–18 The C2 region is similarly split into the catalytic C2a and regulatory C2b subdomains.^17,19 The C1 and C2 domains also contain coiled-coil helical regions which connect the transmembrane domains to the catalytic regions.^9,15,16,18 The activity of AC1 is stimulated by the independent presence of G-protein alpha subunit (Gα_s), calcium-activated calmodulin protein (Ca²⁺/CaM), and the small molecule forskolin (Fsk) (Figure 1).^15,20,21 Gα_s, containing 356 residues, binds to the C2a region of AC1.¹⁵ CaM, consisting of 149 residues, interacts with components of the C1b and C2a regions of AC1^22–24. Fsk binds to residues 419, 423, 878, and 924 of AC1, components of the C1a and C2a regions.²¹ Both Gα_s and Fsk increase the affinity of the C1 and C2 domains to each other, aiding in dimerization of the catalytic site which is necessary for conversion of ATP to cAMP.²⁵ Interactions between the two transmembrane domains have also been shown to play a role in AC1 assembly and function.²⁶

Figure 1. Full-length and truncated AC1 systems.

(A) Schematic illustration of full-length AC1. (B) Comparison of full-length and truncated AC1 systems via superimposition of the last frames of 500-ns simulations for each system and RMSF plot for three replicas of the last 400-ns simulation for each system. The shaded area around the average RMSF corresponds to the standard deviation across the three replicas. (C) Starting structures for MD simulations of the four truncated AC1 systems in complex with different cofactors, either Ca²⁺/calmodulin (Ca²⁺/CaM), forskolin (Fsk), or both. AC1 without cofactors serves as the control. ATP and Mn²⁺ were docked to the binding pocket in all four systems.

Given that multiple cofactors can independently activate AC1, the relationships between these cofactors are important to AC1 activation mechanisms. In prior research, the function of AC1 bearing mutations altering the Fsk-binding pocket has been rescued via binding of a calcium ionophore, A23187 (which serves the function of Ca²⁺/CaM) to AC1, restoring AC1 sensitivity to Fsk.²¹ There are no direct contacts between the CaM-binding site and the Fsk-binding pocket of AC1, suggesting that CaM and Fsk may have a synergistic impact on AC1. In fact, biochemical studies have shown that the addition of Fsk and CaM has a greater effect.²⁷ This synergism presumably results from global conformational changes to AC1 structure; however, such a molecular mechanism has not yet been investigated. Pharmacological inhibitors or activators selective toward AC1 are limited, inferring experimental overexpression and knockdown as the current predominant methods to understand AC1’s complexity in its many physiological roles, as well as the structural and functional impact of its identified cofactors.^18,21 Previous research has shown success in using molecular dynamics (MD) simulation data²⁸ to analyze the behavior of mammalian Adenylyl Cyclase isoform 5 (AC5),²⁹ demonstrating that MD simulations could provide insights into the molecular mechanisms of Adenylyl Cyclase protein-protein interactions. MD trajectories can elucidate key residues involved in the protein-protein interactions influenced by cofactors CaM and Fsk and identify AC1-specific conformational changes in activation. However, computational analysis of AC1 with respect to activating cofactors has been limited and non-specific to the relationship of CaM and Fsk. Furthermore, existing published data regarding computational modeling and MD of AC isoforms is specific to AC5.^29,30

Given the growing application of computational tools in studying protein–protein interactions, protein–ligand interactions, protein activation mechanisms, and drug discovery,^31–40 this work employed MD simulations to understand how AC1 interacts with cofactors CaM and Fsk. Specifically, we aimed to answer the following questions: (i) How does the individual binding of CaM and Fsk to AC1 affect AC1 structure and function? (ii) How do CaM and Fsk together impact AC1 structure, and does this impact differ from the effects of each cofactor alone? Our simulations of four AC1 systems (AC1-No Partner, AC1-CaM, AC1-Fsk, and AC1-CaM-Fsk) revealed different perturbations of cofactor binding on AC1 protein dynamics and explored representative global conformational structures as well as the local catalytic site. In particular, the regulatory subdomain C1b exhibited unique behavior in response to each cofactor. Additionally, cofactor binding led to reshaping the ATP-binding pocket. Notably, dual-cofactor binding impacted AC1 dynamics differently than single-cofactor binding, indicating the synergistic interplay between CaM and Fsk.

Models and Methods

AC1 Model Construction and Simulation Setup.

As there are no crystal structures of AC1 currently available, the models used to study AC1-cofactor interactions were adapted from AlphaFold-generated models¹⁷ with the input sequence acquired from UniProt.⁴¹ Specifically, the structures of Homo Sapiens AC1 and CALM1 (for CaM) were downloaded from the AlphaFold protein structure database (AC1 is UniProt Entry Q08828, and CALM1 is P0DP23). To dock ATP into the AC1 structure, a crystal structure of Adenylyl Cyclase 9 bound to MANT-GTP from the RCSB Protein Databank (PDB ID: 6R4O) was used as a reference. The secondary structures of AlphaFold-predicted AC1 and the AC9 crystal structure were aligned, and the MANT-GTP was placed into the AC1 structure at the same location as it was in the aligned AC9 structure. The MANT-GTP ligand was then modified to ATP in Maestro. Two Mg²⁺ or Mn²⁺ ions are also necessary near the ATP-binding pocket to stabilize ATP. The UniProt entry for AC1 defines the residues involved in each of the 2 binding pockets, and Mn²⁺ was docked at the center of each of these binding pockets. The full structure then underwent energy minimization and was then reviewed to ensure both ions and ATP were in locations aligning with prior literature and UniProt entry data.

Similar to the procedure for docking ATP, to prepare the AC1-Fsk model, the 6R4O PDB structure was used as the reference model. Similar to the method used to dock ATP, the secondary structures of AlphaFold-predicted AC1 and the crystal structure of AC9-Fsk were aligned, and Fsk was placed in the AC1 structure at the same position as in the aligned AC9 crystal structure.

To construct the AC1-CaM model, first, a model of Ca²⁺/CaM was constructed. The CaM structure was predicted by AlphaFold from the UniProt Sequence for CALM1. Next, Ca²⁺ ions were docked into the CaM structure. The CALM1 UniProt entry lists each of the four CaM/Ca²⁺ binding sites as 5-residue pockets. Each ion was placed in the middle of a 5-residue binding pocket defined in the UniProt database. After all 4 ions were added, the full CaM/Ca²⁺ structure underwent energy minimization, after which the positions of each Ca²⁺ ion were reviewed to ensure it remained in proximity to its 5-residue binding pocket.

Once the CaM/Ca²⁺ structure was prepared, it was docked to AC1. To do so, a fragment spanning residues 495–522, which is the experimentally determined CaM-binding region⁴², of AlphaFold-predicted AC1 was extracted and docked into Ca²⁺/CaM. The Ca²⁺/CaM complex was then structurally aligned with the AlphaFold-predicted AC1 model, and the overlapping 495–522 segment was replaced with the docked configuration. To address steric clashes, small terminal rotations of CaM were introduced. The composite model was then subjected to energy minimization and full protein preparation before molecular dynamics simulations, ensuring stereochemical compatibility and overall stability. Although these rotations were applied manually, this avoided the need for numerous distance restraints and reduced the risk of over-bias when docking a 149-residue cofactor to a 1119-residue protein, while still preserving direct-contact conformational fidelity.

Following the protocol detailed above, once CaM, Fsk, ATP, and necessary ions (Ca²⁺ for CaM and Mn²⁺ for AC1- ATP)^{21–23,41–43} were docked using Maestro (Schrödinger),⁴⁴ the complex models underwent energy minimization. As modeling full-length AC1 with its cofactors and the transmembrane domain is computationally demanding, a model with greater efficiency was achieved by removing the transmembrane domains (Figure 1B). Force restraints were applied to restrict residue behavior in the regions where residues were removed, specifically residues 60, 236, 607, and 805, thereby ensuring stability and reliable results. This truncated AC1 model was validated using simulations of full-length transmembrane AC1 bound to no cofactors. As shown in Figure 1B, the root mean square fluctuation (RMSF) of both the truncated and full-length AC1 systems were almost identical, indicating that the truncated system reflects full-system dynamics (for more detailed information, see Supporting Information).

Upon validation of the AC1 truncated model, four truncated AC1 systems were analyzed: AC1 bound to no partners (control), AC1 bound to Fsk, AC1 bound to CaM, and AC1 bound to both CaM and Fsk (Figure 1). Systems were prepared within Schrodinger’s Maestro platform.⁴⁴ Next, using System Builder⁴⁵ an orthorhombic solvent box (with SPC as the water solvent model) was constructed around each system, with the volume minimized and a padding of 15 Å in each direction. Each system was neutralized and set to a default salt concentration of 0.15 M using NaCl to reflect physiological conditions. The full-length AC1 setup follows a similar protocol by removing restraints and embedding the system into a POPC membrane with the transmembrane location determined by the PPM web server.⁴⁶ The water box was configured with a padding of 15 Å in the X- and Y-directions, and a padding of 40 Å in the Z-direction.

Simulations were run using Desmond.^44,45 Each system was minimized and equilibrated using the default multistage simulation workflow of Desmond. Subsequently, the production run of truncated systems was simulated in the NPT ensemble at 300 K for 500 ns using the OPLS5 force fields,^47,48 with three independent replicas. A harmonic force restraint of 50 kcal/mol/Ų was applied to residues 60, 236, 607, and 805 in truncated AC1 systems to ensure protein stability in the absence of the transmembrane region. Similarly, following energy minimization and equilibration, the production run of full-length AC1 was simulated in the isothermal-isobaric NPγT ensemble at 300 K for 500 ns using the OPLS4 force fields,^47,48 with three independent replicas. All system parameters are provided in Table S1.

Simulation Analysis and Visualization.

Simulations were analyzed quantitatively for the truncated AC1 systems in complex with different cofactors. For the root mean square deviation (RMSD) analysis of all systems, please see Figure S1. Unless otherwise stated, the first 100 ns of the truncated AC1 simulations were considered as further equilibration, and only the last 400 ns of each trajectory were used for data analysis. Cartoon structures of proteins were prepared in PyMOL⁴⁹ and VMD.⁵⁰

Principal Component Analysis (PCA) was performed using the ProDy Python package⁵¹ to quantify the essential features of AC1 movement during simulations. Given the flexible nature and terminal regions, the first and last 60 residues (residues 1–60, and 1060–1119) were excluded from the PCA analysis. The cumulative variance and the first four principal components of the truncated AC1 systems were compared in order to assess the significance of cofactor impact. Results were saved in NMWiz file format to generate plots that visualized the magnitude of principal components in VMD.

For each trajectory, three vectors were calculated to quantify the movement of ATP and the ATP- and Fsk-binding sites. The first vector quantified movement of the ATP-binding pocket (residues of contact: residues 308–313, 350–352, 396, 920, 997–999, 1004–1008, and 1044).^41,43 The second vector indicated the movement of the ATP molecule. The third vector illustrated the movement of the Fsk-binding pocket (residues of direct contact: 419, 423, 878, and 924).²¹ We aligned the trajectories using the coiled-coil helical region because the catalytic domain is not isolated in the cytoplasm but attached to the transmembrane domain. In the activate state, the membrane embedding of this domain slightly restricts its motion, making it a suitable reference domain for studying the binding sites’ movement. Since our truncated systems excluded the transmembrane domains, we used the coiled-coil helical region as the reference for our binding pocket motion analysis.

The movement of the ATP-binding pocket, the ATP molecule, and the Fsk-binding pocket were computed by calculating the change in center of mass (COM) location of the atoms of interest (in the x, y, and z directions). We calculated the COM vector for each frame as [xi−x1, yi−y1, zi−z1], comparing the COM at frame i to that of the frame at 100 ns (considered as frame 1, in the last 400 ns trajectories). To ensure consistency across all four systems, we translated these vectors to a common reference structure, the starting structure for AC1-CaM-Fsk. The COM of this reference structure is (x0, y0, z0). We then translated both the starting and ending points of each vector by (x0−x1, y0−y1, z0−z1), resulting in new coordinates (x1+(x0−x1), y1+(y0−y1), z1+(z0−z1)) and (xi+(x0−x1), yi+(y0−y1), zi+(z0−z1)). This preserves the original vector while shifting it to a common reference structure. We visualized their location during simulations by mapping the new coordinates of the COM positions of ATP or binding pockets at each frame onto the reference structure. We also plotted the distribution of the XYZ components of the vector for the ATP-binding pocket movement. To study the relative movement of ATP in the binding pocket, we calculated the COM distance between the ATP-binding pocket and the ATP molecule for each frame and plotted the distribution of this distance per system.

To assess the impact of each cofactor on the geometry of the ATP-binding pocket, the cross-sectional area and pocket depth were computed using COM of four key regions—residues 308–313, 350–352, 1004–1008, and 997–999—which are critical for ATP binding (see equations below). The distributions of ATP-binding pocket cross-sectional area and pocket depth were then graphed and compared between systems.

$$\overrightarrow{n}=\left(\overrightarrow{p_2}-\overrightarrow{p_1}\right)\times \left(\overrightarrow{p_3}-\overrightarrow{p_1}\right)$$ (1)

$$\mathit{\text{Cross}}-\mathit{\text{Sectional}}\mathit{\text{Area}}=\frac{\left|\right|\overrightarrow{n}\left|\right|}{2}$$ (2)

$$\mathit{\text{Depth}}=\frac{|\overrightarrow{n}·\overrightarrow{p_4}-\overrightarrow{n}·\overrightarrow{p_1}|}{\left|\right|\overrightarrow{n}\left|\right|}$$ (3)

where $\overrightarrow{p_1},\overrightarrow{p_2},\overrightarrow{p_3},\overrightarrow{p_4}$ are the position vectors of the COM of residues 308–313, 350–352, 1004–1008, and 997–999, respectively, and $\overrightarrow{n}$ is the normal vector of the plane defined by COM of residues 308–313, 350–352, and 1004–1008. As shown in Figure 2, the cross-sectional area is calculated as the area of a triangle formed by the COM of residues 308–313, 350–352, and 1004–1008. The pocket depth is defined as the perpendicular distance from the COM of residues 997–999 to the plane defined by this triangle.

Figure 2. Cartoon representation of the AC1 catalytic site with the geometric framework defining the ATP-binding pocket.

Three segments of residues—308–313, 350–352, and 1004–1008—are depicted as green cylinders, while residues 997–999 are shown as dark blue cylinders. The center of mass (COM) of the three green segments forms a triangle that defines the cross-sectional area of the pocket. The vertical distance from the COM of the dark blue segment to this triangle plane represents the pocket depth.

We also examined how cofactor binding affects the AC1-ATP interaction network. To determine whether distinct positive or negative correlations existed in different AC1-cofactor systems, we monitored the distances between ATP and key residues–Asp308, Asp352, Lys920, Asp997, Ary1008, and Lys1044–and computed the Pearson correlation coefficients among them.

Results

To systematically investigate the conformational changes of AC1 induced by cofactors CaM and Fsk and their potential synergistic effects in AC1 activation, we conducted MD simulations of different AC1-cofactor systems, namely AC1-No Partner, AC1-CaM, AC1-Fsk, and AC1-CaM-Fsk. Our results revealed distinct roles of CaM and Fsk in modulating AC1 structural dynamics and investigated their synergistic interplay.

Global Conformational and Dynamic Changes to AC1 Induced by Cofactors

We first studied the global mobility of AC1. PCA analysis performed on three replicas of the last 400-ns simulations (Figure S2A) showed that the cumulative variance of PC1 through PC4 exceeds 80% across all systems. Notably, as shown in Figure S2B, complexes with CaM, Fsk, and both cofactors demonstrated significantly different dynamics within regions around Lys497 and Asn548 (containing the CaM-binding site—residues 495–522, the C1b helix region—residues 530–542, and the subsequent loop region—residues 542–560). For example, in PC1 (Figure S2B), the mobility in this region of the AC1-Fsk complex is reduced remarkably. The CaM-bound AC1 complexes display increased movements at the CaM-binding site and the loop sequentially after the C1b region, while the movement at the C1b helix has almost the same magnitude (with modest reduced mobility at the N-terminal of the C1b helix, spanning residues 530–536, see Figure S2B and zoomed-in views in Figure 3) compared with the AC1-No Partner system. CaM-bound AC1 complexes display reduced movements in residues 560–600. This indicates that CaM introduces significant motion near its binding site while reducing motion in areas more distant from the binding site, and these dynamics are likely critical for AC1 activation.

Figure 3. Principal Component 1 (PC1) of the AC1 backbone from truncated AC1 systems.

Shown are PC1 of AC1-No Partner (A), AC1-CaM (B), AC1-Fsk (C), and AC1-CaM-Fsk (D). The terminal regions (N- and C- termini, namely the first and last 60 residues) with large fluctuations were excluded from the PCA. In the AC1-CaM system, motions were remarkably enhanced at the CaM-binding site (residues 495–522), the C1b helix (residues 530–542), and the following loop (residues 542–560), highlighted with dashed circles. In both AC1-CaM and AC1-CaM-Fsk, regions around residues 497 and 548 exhibit pronounced fluctuations, with zoomed-in views shown at the bottom.

To relate the conformational dynamics of AC1, Figure 3 illustrates PC1-associated motions. The motions in regions 490–560 and 560–600 indicate that C1b is very sensitive to cofactor binding and may play a key role in stabilizing the protein’s conformation with Fsk binding or triggering dynamics critical for AC1 activation with CaM binding. The observation suggests a dynamic and stabilizing effect, with CaM-Fsk interactions revealing a dynamic mode of motion distinct from AC1-CaM and AC1-Fsk, consistent with the current assumption that AC1-CaM-Fsk has a unique interaction that influences AC1 activity.

Relocation of the Catalytic Binding Site and the Fsk-Binding Pocket

While an understanding of global conformational changes informs the effects of cofactors on AC1 structure and dynamics, assessing cofactor-induced changes to AC1 catalytic function necessitates analysis of binding sites associated with AC1 activity. In particular, for the ATP- and Fsk-binding sites, their COM positions were calculated based on specific residues from both the C1 and C2 domains (see Methods). The movement area of the ATP-binding pocket, the ATP molecule, and the Fsk-binding pocket differ by cofactor (Figures 4A, 4B, and S3), indicating that cofactors affect the direction of motion of these binding regions. The COM coordinates of the ATP-binding pocket, visualized as a solid space in Figure 4B, illustrate that the ATP-binding site moves within a small area in the AC1-CaM and AC1-CaM-Fsk systems. By contrast, the AC1-NP and AC1-Fsk systems exhibit broader spatial movement, shifting away from the initial location, where key residues of the ATP-binding pocket are displayed as grey sticks. And the distributions of Y-direction motion of the ATP-binding pocket for both the AC1-CaM and AC1-CaM-Fsk systems are nearly identical (Figure 4B). CaM binding not only changes the motion of the ATP-binding pocket but also of the Fsk-binding pocket, as demonstrated by the similarities in the Fsk-binding pocket movement between AC1-CaM and AC1-CaM-Fsk (Figure S3). Therefore, the movements of ATP and key binding pockets (Figures 4B and S3) depend on the cofactor, and in the dual-cofactor system, CaM plays a dominant role in directing the ATP molecule and ATP- or Fsk-binding pocket motion. In addition, the relative distance between ATP and its binding pocket (Figure 4C) was monitored to show the stability of ATP in its binding pockets. In the AC1-CaM and AC1-CaM-Fsk systems, the relative distance remains within the range of 1 to 4 Å for the main peak, indicating that in most cases, ATP moves in concert with its binding pocket during the simulation. The main peak for the AC1-Fsk system appears around 5 Å, accompanied by a shoulder peak near 4 Å, suggesting that Fsk modulates the local binding environment of ATP. However, the low-density distribution of the AC1-NP system implies that ATP is quite flexible and may reside in the catalytic site in more than one state when no cofactors are bound. In addition, the distance between ATP and CaM/Fsk in the AC1-CaM-Fsk system (Figure S4) shows that the distance between ATP and cofactors remains relatively stable, which indicates that their interactions with their respective binding pockets are quite strong and that they move in a coordinated fashion relative to the coiled-coil helical region.

Figure 4. Center-of-mass (COM) movement of the ATP-binding pocket.

(A) Bottom view of AC1 shown in surface representation, with the ATP-binding pocket highlighted in dashed squares. The ATP molecule is depicted as dark gray sticks. The coordinate system is oriented with the X-axis pointing into the page. (B) The COM positions of the ATP-binding pocket for each frame were mapped onto the starting structure to visualize its motions during the simulation. All COM positions of four systems were aligned to a common reference structure for direct comparison. The graphs on the right show the distributions of the X-, Y-, and Z-components of the ATP-binding pocket COM coordinates for each system. Panel C shows the distribution of the distance between the COM of ATP and the COM of the ATP-binding pocket across all systems. Color coding is consistent across panels: AC1-No Partner (blue), AC1-CaM (yellow), AC1-Fsk (green), and AC1-CaM-Fsk (red).

Cofactor-Dependent ATP-Binding Pocket Stability

The disturbance in protein dynamics (Figures 3 and S1–S2) caused by cofactor binding inspired us to further characterize the structural features of the ATP-binding pocket. Therefore, we measured the size of the ATP-binding pocket, specifically the cross-sectional area and pocket depth (see Methods). The cross-sectional area (Figures 5A) and the pocket depth (Figures 5B) denoted in Figure 2 by the green plane and the light blue dashed line, respectively, represent the horizontal and vertical dimensions of the ATP-binding pockets. In Figure 5A, the cross-sectional area distribution of AC1-No Partner is very large, which may lead to a loose ATP binding, while the cofactor-bound systems exhibit a narrower pocket and therefore imply a stronger interaction. However, the cross-sectional area and pocket depth distribution profiles of the cofactor-bound systems AC1-CaM, AC1-Fsk, and AC1-CaM-Fsk are different, which provides evidence of the unique and synergistic effects of Fsk and CaM on AC1 catalytic site dynamics. The bimodal distribution for cross-sectional areas in the AC1-CaM system (Figure 5A) suggests that CaM induces oscillations along the C1-C2 axis, and the peak with larger cross-sectional areas in the AC1-CaM system disappeared in the AC1-CaM-Fsk system. The broad distribution for pocket depth in the AC1-Fsk system (Figure 5B) indicates that Fsk affects the pocket depth through oscillation of residues 997–999 along the vertical axis. The overall distribution profile of the cross-sectional area in the AC1-CaM-Fsk system is similar to that of the AC1-Fsk system, having a unimodal distribution, indicating the dominant role of Fsk in reshaping the local structure of the ATP-binding pocket in the dual cofactor system. And the pocket depth distribution in the AC1-CaM-Fsk is affected by both CaM and Fsk binding, exhibiting an obvious bimodal distribution. Overall, the binding of cofactors CaM and Fsk enhances the stability of the ATP-binding pocket, allowing it to accommodate ATP more tightly via interactions with the three regions denoted as green cylinders in Figure 2. Two-dimensional distributions of the cross-sectional area and pocket depth for each system are displayed in Figure S5. To explore the interactions of ATP with its binding pocket, we analyzed a few key residues that contact ATP – including Asp308, Asp352, Lys920, Asp997, Ary1008, and Lys1044 – by examining their minimal distances from ATP and the potential Pearson correlation between these distances across the four AC1 systems, as shown in Figure S6. For example, the correlation coefficients for the ATP-Asp352 and ATP-Lys920 distances are 0.74 in AC1-No Partner system, but only 0.16 in AC1-Fsk system and 0.21 in AC1-CaM-Fsk system. Conversely, the correlation coefficients for ATP-Lys920 and ATP-Lys1044 are -0.29 in AC1-No Partner system, but it becomes -0.74 in AC1-CaM system and -0.85 in AC1-CaM-Fsk system. Therefore, as shown in Figure 5C, the AC1-NP system shows a clear positive correlation for the ATP-Asp352 and ATP-Lys920 distances, while the AC1-CaM-Fsk system shows a clear negative correlation for the ATP-Lys920 and ATP-Lys1044 distances. A similar scenario was observed for Asp997 (Figure S6), as was for Lys920. Notably, as shown in Figure 5D, Lys920 and Asp997 anchor the adenosine moiety of ATP, while Asp352 and Lys1044 contact the triphosphate group. The distinct distance correlations shown in Figures 5 and S6 indicate that subtle variations in the interaction network may result from cofactor binding and play a role in the release of pyrophosphate. Furthermore, the solvent-accessible surface area (SASA) of ATP was monitored to provide a rough assessment of ATP’s interactions with AC1 and with the solvent. In the cofactor-bound AC1 systems (Figure S5F), the ATP SASA is primarily concentrated around 600 Ų, while AC1-NP exhibits no evident SASA preference, suggesting that cofactor binding may stabilize AC1 in a specific local configuration that accommodates ATP in a location more favorable for subsequent conversion to cAMP.

Figure 5.

Characterization of the ATP-binding pocket geometry and ATP-residue interactions. Distributions of the cross-sectional area (A) and pocket depth (B) of the ATP-binding pocket. Per-frame minimal distances (C) between ATP and key residues–Asp352, Lys920, and Lys1044–for the AC1-No Partner and AC1-CaM-Fsk systems. The analysis reveals a positive correlation between the distance of ATP from Asp352 and Lys920 in the AC1-No Partner system, a correlation that is absent in the AC1-CaM-Fsk system. In contrast, the AC1-CaM-Fsk system shows a negative correlation between the distance of ATP from Lys920 and Lys1044. For complete correlation coefficient results, see Figure S6. Cartoon representation of AC1 (D) with the key residues labeled.

Discussion

AC1 is an important member of the Adenylyl Cyclase family involved in the GPCR signal transduction pathway, and it plays key roles in pain mechanisms and brain function.¹ Thus, identifying molecular mechanisms behind AC1 behavior could provide new insights into drug development for multiple disease pathways. In this study, we developed a computational model for AC1 to understand how the stimulatory cofactors CaM and Fsk affect AC1 behavior, and whether CaM and Fsk have a synergistic effect on AC1 structural dynamics using in silico approaches. Analysis of key regions within the C1b domain, together with detailed characterization of the catalytic site, leads us to propose a possible role of cofactors CaM and Fsk in AC1 dynamic behaviors, which may be further validated through future computational and experimental studies. Although the AC1 model was constructed based on the AlphaFold-predicted structure, it closely resembles the crystal structure of AC9 (RCSB 6R4O), supporting the reliability of our model. To further elucidate the activation mechanism of AC1, future studies could explore more complex systems, such as the full-length AC1–cofactor complex, membranes with diverse lipid compositions, and the catalytic reaction process of ATP conversion.

Distinct Dynamic Behaviors in the C1b Region Across Different Systems

Cofactor binding induces distinct dynamic behaviors in the C1b region of AC1, which plays a pivotal role in structural rearrangement during AC1 stimulation due to its function as a CaM-binding site and its critical positions linking the catalytic site to the transmembrane domains^9,15,16,18. Protein dynamics, as revealed in the PCA results (Figures 3 and S2), further highlight the distinct behaviors within the C1b region when binding to different cofactors. For example, CaM or Fsk binding reduced the movement in C1b residues 560–600, while CaM binding enhanced the movement of C1b residues 490–560 (containing the CaM-binding pocket). This CaM-induced movement could be attenuated by simultaneous binding of Fsk, suggesting that CaM and Fsk act synergistically in modulating AC1 behaviors.

Asymmetric CaM and Symmetric Fsk Interaction Pattern at Two Interaction Sites and AC1 Catalytic Site Dynamics

Fsk binds to the C1a and C2a subdomains to stabilize the catalytic site during AC1 activation, while CaM has been reported to interact with the C1b and C2a subdomains^22–24. This dual-site binding mode across both the C1 and C2 domains suggests that Fsk and CaM may act as molecular “glue”, promoting the dimerization of the C1a and C2a subdomains. Analysis of the cross-sectional area of the catalytic site (Figure 5A) reveals that cofactor binding brings C1a and C2a closer, as indicated by a reduced cross-sectional dimension. CaM affects the cross-sectional area through oscillation along the C1-C2 axis, while Fsk affects the pocket depth through oscillation of residues 997–999. The minimal distance between CaM and C2a (Figure S4) further supports that CaM simultaneously interacts with C1b and C2a, consistent with these experimental observations of two interaction sites. Additionally, CaM introduces conformational flexibility to the C1b subdomain (Figure 3), enabling AC1 to capture conformations favorable for catalytic site formation. In contrast, Fsk reshapes the local structure of the ATP-binding pocket (Figure 5). Fsk binds symmetrically at the C1a-C2a interface, while CaM binds asymmetrically—primarily to C1b and partially to C2a. This asymmetric binding pattern in CaM-bound AC1 complexes may result in distinct effects on AC1 dynamics.

The interplay of ATP with both AC1 and solvent is worth studying, as the interaction network and the solvent exposure are of great significance for the catalytic reaction. Intriguingly, CaM and Fsk binding elicit distinct conformational changes in the catalytic domain: in cofactor-bound AC1 systems, the correlation among minimal distance between ATP and key residues and ATP SASA (Figures 5C, 5D, and S6) exhibit distinct features compared to the AC1-NP system. Specifically, the correlated movements of key residues relative to ATP suggest a dynamic and cooperative interaction network within the ATP-binding pocket, which appears to act in concert to maintain the AC1-cofactor system in its active state. We therefore hypothesize that cofactor-induced dynamics in critical regions of AC1 are linked to a concerted conformational rearrangement of key residues. This local structural change, coupled with the dual-site binding mode of CaM and Fsk, stabilizes AC1 in its active state.

Synergistic Effect on Structural Dynamics of AC1 by Fsk and CaM

Previous biochemical studies suggest that CaM and Fsk synergistically activate AC1.^27,52 The synergistic effect between CaM and Fsk is now further supported by the unique structural arrangement and protein dynamics within AC1-CaM-Fsk—a mechanism unique to this specific configuration. CaM allosterically affected the movement of the Fsk-binding pocket, resulting in an alteration in the space of motion (Figures S3). The dominant role of Fsk in the AC1-CaM-Fsk system is reflected in the reshaping of the ATP-binding pocket (Figures 5A and 5B) by inducing a unimodal distribution for pocket cross-sectional area and contributing to one peak of a bimodal distribution for pocket depth. This may be due to the proximity between the Fsk-binding pocket and the ATP-binding pocket and could be associated with the Fsk-induced AC1 stimulation. This synergistic effect is supported by experimental findings on mutant AC1, where Fsk-stimulated AC1 activity was rescued with calcium ionophore A23187,²¹ suggesting a significant role for the CaM-Fsk synergy in stabilizing AC1. The W418A and S924P mutations have been reported to cause Fsk insensitivity and affect the flexibility and shape of the binding pocket.²¹ In the same work, A23187 partially enhanced Fsk sensitivity and restored AC1 activity. As such, CaM likely mitigates the structural barrier to Fsk-binding pocket formation caused by these mutations. CaM, through its dual-site binding, facilitates the assembly of C1a and C2a subdomains in AC1 mutants, which may consequently enable the formation of the Fsk-binding pocket and enhance Fsk sensitivity. Hence, the synergy observed between CaM and Fsk in modulating AC1 activity and structure provides a promising mechanistic foundation for future studies aimed at unraveling this complex interplay.

Based on the protein dynamics and geometric analyses presented in this study, we propose a potential cascade of protein-cofactor association model. When CaM binds to the C1b region of AC1, it induces conformational dynamics within C1b, facilitating the dimerization of the C1a and C2a subdomains. Since CaM promotes the formation of the catalytic site which is near the Fsk-binding pocket, this structural rearrangement may enhance Fsk binding. Finally, the simultaneous binding of CaM and Fsk to AC1 induces distinct protein dynamics and stabilizes the catalytic site through synergistic interactions: Fsk binds internally near the catalytic site, while CaM binds externally to AC1. Together, these interactions prime AC1 for binding ATP in a favorable conformation. Given residues 418 and 924 in the AC1 Fsk-binding pocket—which are key to Fsk interactions—are conserved across the Fsk-binding pocket in all AC isoforms²¹, the synergistic interplay between CaM and Fsk observed here in AC1 could extend to the other CaM-sensitive AC isoforms, Adenylyl Cyclase 3 and 8.¹

Conclusions

In this work, we systematically explored the distinct effects of cofactors CaM and Fsk on the AC1 protein dynamics using extensive all-atom MD simulations. We found that when bound to CaM, Fsk, and CaM-Fsk, respectively, AC1 adopts unique conformations and dynamics, particularly distinguished by the dynamics of the regulatory C1b subdomain, which links C1a to TM2 and plays a critical role in AC1 activation. Both CaM and Fsk contribute to stabilizing the ATP-binding pocket by modulating its size. However, CaM and Fsk affect AC1 dynamics in different ways: CaM introduces substantial protein dynamics near its binding site at C1b, potentially promoting the conformations favorable for ATP and Fsk binding. In contrast, Fsk tends to reshape the ATP-binding pocket from the inside due to the proximity of the Fsk-binding pocket to the ATP-binding pocket. Together, CaM and Fsk can stabilize and reshape the ATP-binding pocket from different positions—CaM acting externally on the AC1 surface and Fsk acting internally on the catalytic site. In addition, the dual-cofactor-bound complex exhibits features unique from the two single-cofactor-bound complexes, indicating the potential synergistic effect of CaM and Fsk in AC1 activation. In summary, cofactor binding not only induces conformational change and alters the local environment at the catalytic site, but also exhibits synergistic effects through interactions at different sites within AC1. These findings provide the molecular basis to guide future structure-based design of small-molecule modulators for AC1.

Supplementary Material

The Supporting Information is available free of charge at https://pubs.acs.org.

Table summary of the simulation setup for truncated and full-length AC1 systems, along with extended analysis of RMSF, PCA, and the ATP-binding pocket.

Movie S1: Shows the primary PC component for all four truncated AC1 systems: AC1-CaM (yellow), AC1-Fsk (green), AC1-CaM-Fsk (red), and AC1-NP (blue), all overlaid for comparison.

Movie S2: Displays the primary PC component for the truncated AC1-NP system.

Movie S3: Displays the primary PC component for the truncated AC1-CaM system.

Movie S4: Displays the primary PC component for the truncated AC1-Fsk system.

Movie S5: Displays the primary PC component for the truncated AC1-CaM-Fsk system.