Abstract
The nanoscale organization of enzymes associated with the dynamics of second messengers is critical for ensuring compartmentation and localization of signaling molecules in cells. Specifically, the spatiotemporal orchestration of cAMP and Ca2+ oscillations is critical for many cellular functions. Previous experimental studies have shown that the formation of nanodomains of A-kinase anchoring protein 79/150 (AKAP150) and adenylyl cyclase 8 (AC8) on the surface of pancreatic MIN6 β cells modulates the phase of Ca2+-cAMP oscillations from out-of-phase to in-phase. In this work, we develop computational models of the Ca2+/cAMP pathway and AKAP/AC nanodomain formation that give rise to the two important predictions: instead of an arbitrary phase difference, the out-of-phase Ca2+/cAMP oscillation reaches Ca2+ trough and cAMP peak simultaneously, which is defined as inversely out-of-phase; the in-phase and inversely out-of-phase oscillations associated with Ca2+-cAMP dynamics on and away from the nanodomains can be explained by an incoherent feedforward loop. Factors such as cellular surface-to-volume ratio, compartment size, and distance between nanodomains do not affect the existence of in-phase or inversely out-of-phase Ca2+/cAMP oscillation, but cellular surface-to-volume ratio and compartment size can affect the time delay for the inversely out-of-phase Ca2+/cAMP oscillation while the distance between two nanodomains does not. Finally, we predict that both the Turing pattern-generated nanodomains and experimentally measured nanodomains demonstrate the existence of in-phase and inversely out-of-phase Ca2+/cAMP oscillation when the AC8 is at a low level, consistent with the behavior of an incoherent feedforward loop. These findings unveil the key circuit motif that governs cAMP and Ca2+ oscillations and advance our understanding of how nanodomains can lead to spatial compartmentation of second messengers.
Author summary
Cyclic adenosine monophosphate (cAMP) and Ca2+ are key molecules that relay signals from the cell membrane to downstream molecules. The temporal and spatial distribution of cAMP and Ca2+ within the cell can be regulated by clusters formed by specific molecules. Previous studies have shown that nanodomains of A-kinase anchoring protein 79/150 (AKAP150) and adenylyl cyclase 8 (AC8) on the surface of pancreatic MIN6 β cells modulates Ca2+-cAMP oscillations from non-zero to zero phase difference, i.e., from out-of-phase to in-phase. By developing a computational model of the Ca2+/cAMP pathway, we found that the non-zero phase difference is not an arbitrary value: it makes Ca2+ reach trough and cAMP reach peak simultaneously. We defined this out-of-phase behavior as inversely out-of-phase, and revealed that this behavior and in-phase behavior can be explained by an incoherent feedforward loop. Biophysical properties of cells and nanodomain distributions do not affect the existence of in-phase or inversely out-of-phase behaviors, but may affect the time delay for the inversely out-of-phase behavior. Furthermore, AKAP/AC nanodomains from a Turing pattern or experiments show consistent simulation results with the incoherent feedforward loop. This study improves our understanding of underlying mechanisms of Ca2+/cAMP oscillation, shedding light on controlling Ca2+ and cAMP through nanodomains.
1 Introduction
Cyclic adenosine monophosphate (cAMP) and Ca2+ are key second messengers that regulate several cellular functions, including muscle contraction, neuronal excitability, cell migration, metabolism, endocytosis, plasma membrane repair, and immune function [1–6]. The execution of these cellular functions relies on both the temporal dynamics and the spatial distribution of cAMP and Ca2+. The temporal dynamics of cAMP and Ca2+ play crucial roles in many biological systems. For example, changes in the cAMP level and downstream protein kinase A (PKA) are connected with tumorigenesis, invasion, metastasis and drug resistance [7–9], and an increase in cAMP levels inhibits hepatocellular carcinoma cell proliferation [10] and induces the apoptosis of glioblastoma cells [11]; Ca2+ frequency can be decoded by downstream Ca2+ sensors and translated into distinct cellular responses [12–16]. In addition to tightly regulated temporal control, cAMP and Ca2+ are highly spatially compartmentalized at the plasma membrane and cellular organelles [17–19]. If this spatial compartmentalization is disrupted, it may lead to changes in signaling pathways and gene expression [20–22], as well as pathological conditions such as cell apoptosis [22], abnormal insulin secretion [23], heart dysfunction [3, 24, 25], familial breast cancer [26], and schizophrenia [27].
The temporal dynamics of cAMP and Ca2+ usually demonstrate oscillatory behaviors in many cell types including neurons [28, 29], cardiomyocytes [30], and pancreatic β cells [31]. Given the importance of spatiotemporal orchestration of Ca2+ and cAMP function, understanding the biochemical and biophysical mechanisms that govern their oscillations is very important. Mathematical and computational modeling can aid in deciphering how different cellular features play a role in governing these oscillations [32–37]. Previously, we showed that the frequency of Ca2+-cAMP oscillations depends on the geometry of dendritic spines in neurons [38, 39]. In this study, we focus on cAMP and Ca2+ oscillations in pancreatic β cells because of their critical role in insulin secretion. The pathway regulating the insulin secretion in pancreatic β cells is as follows: an increase in blood glucose leads to the increase in the cytosolic ATP/ADP ratio in the pancreatic β cell [40]; then, ATP-sensitive potassium channel (KATP) channel closes, inducing the influx of Ca2+ and subsequent insulin vesicle exocytosis [41, 42]. During this insulin secretion process, the cAMP pathway is also activated: increased Ca2+ level also stimulates the activation of adenylyl cyclase type 8 (AC8), and AC8 improves the cAMP production, followed by protein kinase A (PKA) activation. Furthermore, cAMP level is also controlled by the Ca2+-dependent PDE1C (phosphodiesterase 1C; denoted as PDE for simplicity) in pancreatic β cells, because this PDE is activated by Ca2+ and then mediates the transient cAMP decreases [32]. The activated cAMP/PKA pathway affects cellular excitability, Ca2+ signals, exocytosis, cell viability, and cell–cell interactions [31, 43, 44]. Disturbances in this pathway may be associated with diabetes [45].
The cross-talk between Ca2+ and cAMP is vital for the proper function of pancreatic β cells. Such cross-talk is achieved by the highly connected spatiotemporal organization between cAMP and Ca2+ in pancreatic β cells. The temporal dynamics of Ca2+ and cAMP in pancreatic β cells show synchronized oscillations, and this oscillation is initiated and modulated by PKA [46]. Recently, experiments showed that the synchronized oscillation of these two second messengers exhibits distinct phase delay at different membrane locations. For example, the membrane nanodomain formed by plasma membrane (PM)-localized scaffold protein A-kinase anchoring protein 79 (AKAP79; rodent ortholog AKAP150) [32] (Fig 1A). AKAP79/150 is a multivalent scaffold protein that associates with itself to form nanodomains. Additionally, AKAP79/150s also can recruit signaling molecules to cell membrane, including PKA, the voltage-gated Ca2+ channel CaV1.2, Protein Kinase C (PKC), the Ca2+/calmodulin-dependent protein phosphatase calcineurin, Ca2+-sensitive ACs, and AMPA receptors [47]. On membrane nanodomains formed by A-kinase anchoring proteins 79/150 (AKAP79/150) and AC8 (referred to as the AKAP/AC nanodomains), Ca2+ and cAMP were found to oscillate in-phase, i.e., without any phase difference. Away from the membrane AKAP/AC nanodomains, Ca2+ and cAMP oscillate out-of-phase. This discovery that membrane nanodomains of AKAP79/150 and AC8 can modulate the in- and out-of-phase oscillations of cAMP-Ca2+ leads to interesting questions about the biochemical and biophysical mechanisms that regulate the formation of these nanodomains and coordinate the spatiotemporal dynamics of these critical second messengers. Can the dynamics of in- and out-of-phase oscillations of Ca2+ and cAMP be explained by a simple network motif? Does subcellular compartment size affect these phase shifts? How do patterns of nanodomains form on the membrane and how do multiple nanodomains affect these oscillatory patterns? We sought to answer these questions using computational modeling.
2 Results
In this work, we developed two classes of models—a 3D reaction-diffusion model to simulate the dynamics of cAMP and Ca2+ and a Turing pattern model with steric hindrace to simulate the formation of AKAP/AC nanodomains. For the 3D reaction-diffusion model, we defined preexisting AKAP/AC nanodomains and focused on the dynamics of cAMP and Ca2+. We used this model to investigate how activities of AC and PDE, cellular compartment size, and pattern of nanodomains affect the in- and out-of-phase oscillations of Ca2+ and cAMP (Fig 1B). To understand how nanodomains might form, we used a Turing model with steric repulsion [48] to characterize the regimes in which stable AKAP/AC nanodomains might arise. Finally, to investigate how the nanodomain patterns might affect the in- and out-of-phase oscillations, we used patterns from both the Turing model and experimentally generated STORM data as the initial distribution of AKAP/AC nanodomains and predicted that both the pattern of AKAP/AC nanodomains and AC activities are critical in controlling the phase behavior of Ca2+-cAMP. These results provide insights into the biochemical and biophysical mechanisms that regulate the spatiotemporal dynamics of Ca2+ and cAMP.
2.1 Phase oscillations of calcium and cAMP are driven by active AC and PDE
We first simulated the dynamics of Ca2+ and cAMP using the 3D reaction-diffusion model (see Reaction-diffusion model subsection in Methods for details). We briefly describe the reactions below. We considered 22 state variables in total (Fig 1C and S1 Table), where 13 chemical species are in the cytosol (non-highlighted species in Fig 1C) and the rest are on the membrane (highlighted in yellow in Fig 1C). The reactions among these chemical species are briefly introduced below (S2 and S3 Tables): the increase of Ca2+ in the cytosol leads to the activation of inactive PDE (Cyclic nucleotide phosphodiesterase) in the cytosol (v6–8) and inactive AC (Adenylyl cyclase) on AKAP/AC nanodomains through CaM (calmodulin) (v1–3 and v4–5); next, the active AC (denoted as AC *) on AKAP/AC nanodomains enhances the synthesis of cAMP (v9), and the active PDE (denoted as PDE *) improves the degradation of the cAMP in the cytosol (v10–13); the change in the cAMP level will affect PKA (protein kinase A) in the cytosol (v14–20), and the latter will regulate the dynamics of Ca2+ by interacting with the voltage module. In the voltage module (gray box in Fig 1C), the membrane is regarded as a capacitor with the capacitance Cm, and four ion channels are considered: Ca2+ channel, K+ channel, Ca2+ gated K+ channel, and leak channel. The currents for these four ion channels are represented by ICa, IK, IKCa, and IL. The current that flows through the membrane is equal to the negative of the sum of ion channel currents, i.e., , where V denotes the membrane voltage. The Ca2+ affects the Ca2+ gated K+ channel (v21), and the Ca2+ channel controls the Ca2+ influx (v22).
Kinetic parameters for the above reactions are from [32], which were listed in S4 Table. Among these parameters, those for binding/unbinding events between Ca2+ and CaM are from [49], and those for the membrane voltage model come from [46]. Other kinetic parameters in [32] were estimated empirically (labeled as “[Estimated]” in S4 Table) or obtained by fitting to the experimental data (labeled as “[Constraint]” in S4 Table). The experimental data used include not only the ranges of the molecule concentration that the sensor can monitor but also the relative changes of biosensor fluorescence for Ca2+ and cAMP in [32]. The diffusion coefficients and initial values were estimated empirically to ensure that their values are biologically plausible. The settings for AKAP/AC nanodomain and compartment are the same as the calculations in [32] (see Reaction-diffusion model subsection in Methods for details): at time 0 there is only one AKAP/AC nanodomain located at the center of the cell membrane, which is modeled by a Gaussian distribution with standard deviation of 25 nm (Fig 1D); the compartment has a height of 600 nm and width of 400 nm (Fig 2A). Furthermore, we chose periodic boundaries for the six surfaces surrounding the compartment, so that many compartments can assemble to form a large cell domain without any discontinuity. These kinetic parameters ensure that values of Ca2+ concentration, cAMP concentration, and membrane voltage are biologically plausible (panel A in S1 Fig). Specifically, the Ca2+ concentration simulated by the model is between 0 and 1.5 μM (panel A(ii) in S1 Fig), which is within the same scale of the Ca2+ concentration in internal Ca2+ store (0.1–0.5 μM from [50]). The cAMP concentration in the model ranges from 0 to 0.2 μM (panel A(ii) in S1 Fig), which falls within the detectable range for sensors ([32]). The membrane voltage in the model also shows the same scale as that in [51] (panel A(iii) in S1 Fig). The cAMP and PKA levels far from the nanodomain are lower than those on the nanodomain (panels A(ii) and (vi) in S1 Fig), forming a spatial heterogeneity that might be crucial for proper cellular functions [52].
Next, we used the above model to simulate the dynamics of Ca2+ and cAMP when there is only one AKAP/AC nanodomain existing in the center of the membrane. Due to the limited experimental data, which monitored intracellular cAMP only at the cell membrane rather than in the cytosol [32], we also only focused on the dynamics of Ca2+ and cAMP at the cell membrane. Because the AKAP/AC nanodomain is radially symmetric, we only need to focus on dynamics in the line from the center to the right edge of the cell membrane (white line in Fig 2B) and used x to denote the distance from the center to the location of interest at cell membrane. We found that, the time delay between Ca2+ and cAMP peak time is zero when x = 0 and near 25 seconds when x is larger than 50 nm (Fig 2B), indicating the in-phase behavior at the AKAP/AC nanodomain and out-of-phase behavior outside AKAP/AC nanodomain. This conclusion is consistent with the simulation results in [32]. The transition happens at x = 48 nm, which is nearly two times of standard deviation of the Gaussian distribution; this suggests that the out-of-phase behavior also occurs at a low but non-zero level of AC*.
To explore why there is a sharp transition of the time delay when changing from AKAP/AC nanodomain to the general membrane, we plotted the dynamics of Ca2+ and cAMP for different values of x (Fig 2C). The dynamics of Ca2+ remain unchanged at different locations (red lines in Fig 2C) because of the fast diffusion of Ca2+. In contrast, the dynamics of cAMP are different at different locations: its level decreases when x increases (Fig 2C). Furthermore, cAMP has two peaks for intermediate x, but the relative height varies (Fig 2C): the peak that is in-phase with Ca2+ is higher than the out-of-phase peak at small x, that is, at the center of the nanodomain, but is smaller compared with the out-of-phase peak at large x, i.e., far from the nanodomain. Here, the out-of-phase peak denotes any non-zero difference in the phase between two oscillatory trajectories. We also found that the in-phase peaks at different locations likely occur at the same time, and so do out-of-phase peaks (dashed gray line in Fig 2C). These results suggest that the sharp transition of the time delay is caused by the swapping between the in-phase and out-of-phase peaks. To better visualize the delay of peaks between Ca2+ and cAMP, we plotted the normalized dynamics of Ca2+ and cAMP (Fig 2D). Here, normalization indicates that the concentration is divided by the maximum value at a specific location during the time interval [0, 1200] seconds. It can be seen that peaks of cAMP at x = 0 are in-phase with Ca2+ at x = 0, those at x = 49 nm are either in-phase or out-of-phase with Ca2+, and those at x = 200 nm are out-of-phase with Ca2+. These results further support that the in-phase peak dominates on the AKAP/AC nanodomain while the out-of-phase peak dominates outside the AKAP/AC nanodomain.
To understand the kinetics of in-phase and out-of-phase cAMP peaks, we first analyzed the dynamics of the immediate upstream species: Ca2+, active AC (AC*), and active PDE (PDE*). We plotted the dynamics of concentrations of these three species at the cell membrane (Fig 2E), and then calculated the trough and peak time (marked by the plus sign or triangle). Here, the trough (or peak) time is defined as the time when the concentration of species reaches the minimum (or maximum) value. For Ca2+, due to its fast diffusion, the trough and peak times remain unchanged when the location x varies (upper panel in Fig 2E). As for AC*, its peak and trough times also do not change when the location x varies between 0 and 20 nm, but these peak and trough times cannot be defined at x> 20 nm because of the low AC* level outside the AKAP/AC nanodomain (middle panel in Fig 2E). The PDE* shows similar behavior as Ca2+, that is, the trough and peak times for PDE* are fixed no matter how the location x changes (lower panel in Fig 2E). Next, we compared these times between Ca2+ and AC* and between Ca2+ and PDE* (Fig 2F). We found that, the AC* trough time at x < 20 nm is the same as that for Ca2+, and so is the peak time (left panel in Fig 2F). Similarly, the trough and peak times for PDE* also match the trough and peak times for Ca2+, respectively (right panel in Fig 2F). These consistencies in the trough and peak time may be caused by the strong activation from Ca2+ to AC* (v4–5 in S2 Table) and from Ca2+ to PDE* (v1–8 in S2 Table). Furthermore, these consistencies indicate that the AC* on the nanodomain and PDE* at the entire membrane exhibit in-phase oscillation with Ca2+.
After analyzing the dynamics of upstream species of cAMP, we turned to explore how these dynamics may affect in-phase and out-of-phase behaviors of cAMP. We plotted the dynamics of the normalized cAMP level at different locations at the cell membrane (Fig 2G), and then compared the cAMP peak time with PDE* trough time or AC* peak time (Fig 2H). The cAMP concentration for each x is normalized by the maximal value during the time interval [0, 1200] seconds. From the dynamics of the normalized cAMP level (Fig 2G), we can find that the cAMP peak time shifts from left to right when the location x decreases. Moreover, the cAMP peak time is the same as the AC* peak time when x is small (left panel in Fig 2H), but equal to the PDE* trough time when x is large (right panel in Fig 2H). These results suggest the in-phase AC*-cAMP oscillation on the AKAP/AC nanodomain and inversely out-of-phase PDE*-cAMP oscillation outside the AKAP/AC nanodomain, where the “inversely” is defined as one oscillator reaching the peak while the other reaches the trough. Since AC* catalyzes the cAMP synthesis (v9 in S2 Table) and PDE* enhances the cAMP degradation (v12 in S2 Table), these inversely out-of-phase and in-phase behaviors may result from the competition between AC*-induced cAMP production and PDE*-induced cAMP degradation. On the AKAP/AC nanodomain, the AC* level is much higher than PDE*, and thus the activation from AC* dominates, leading to in-phase AC*-cAMP oscillation. In contrast, outside the AKAP/AC nanodomain, the AC* shows low values, and thus the cAMP levels are mainly governed by the action of PDE*; PDE* inhibits the cAMP level by enhancing the cAMP degradation, resulting in inversely out-of-phase PDE*-cAMP oscillation. When combined with the observation of in-phase Ca2+-AC* oscillation on the nanodomain and in-phase Ca2+-PDE* oscillation at the entire membrane, the cAMP is in-phase with Ca2+ on the AKAP/AC nanodomain but inversely out-of-phase with Ca2+ outside the AKAP/AC nanodomain.
2.2 Experimental data supports model predictions of in- and inversely out-of-phase calcium-cAMP oscillations
We summarized the mechanism of in-phase and inversely out-of-phase Ca2+-cAMP oscillations in Fig 3A. The origin of the stimulus is the Ca2+ oscillation, which is uniformly distributed spatially due to rapid diffusion. Ca2+ dynamics affects the phase of the cAMP by two species AC* and PDE*, whose dynamics are as follows: the PDE* oscillation at the entire cell membrane is in-phase with Ca2+ through the activation from Ca2+ to PDE* (v1–3 and v6–8 in S2 Table). On the AKAP/AC nanodomain, AC* is not only in high concentration but also in-phase with Ca2+ since the Ca2+ catalyzes the activation of AC* (v4–5 in S2 Table) whereas outside the AKAP/AC nanodomain, AC* is in low concentration, because AC is in low concentration outside the AKAP/AC nanodomain. As a result, on the AKAP/AC nanodomain, the enriched AC* enhances the cAMP production (v9 in S2 Table), driving the cAMP oscillation become in-phase with AC*; outside the nanodomain, the AC* level is much smaller than the PDE* level, and the latter degrades cAMP (v12 in S2 Table), causing inversely out-of-phase cAMP oscillation with PDE*. Since AC* on the nanodomain and PDE* on the entire membrane are both in-phase with Ca2+, the cAMP oscillation is also in-phase with Ca2+ on the nanodomain and inversely out-of-phase with Ca2+ outside the nanodomain. To further verify the role of the AC* and PDE* in the phase difference, we also simulated the dynamics of Ca2+ and cAMP under the two following conditions: the first is to delete the AKAP/AC nanodomain, and the second is to delete both AKAP/AC nanodomain and PDE*. For the first case, we found that the in-phase behavior disappears but the out-of-phase behavior exists on the entire cell membrane (panel B in S1 Fig). Note that no AKAP/AC nanodomain means a zero level of AC*. Thus, the simulation result in the first case suggests that the AC* contributes to the in-phase behavior instead of the out-of-phase behavior, which is consistent with the above mechanism. For the second case, we set the PDE* level to zero in the absence of AKAP/AC nanodomain, and Ca2+ still oscillates but cAMP cannot (S2 Fig), which is consistent with Figure 5E in [53] where cells were treated with the PDE inhibitor 3-isobutyl-1-methylxanthine (IBMX). Thus, this simulation result in the second case indicates that PDE* contributes to the cAMP oscillation. Taken together, AC* and PDE* are crucial for the oscillation and phase difference.
We tested the model predictions against previously published experimental data in [32]. First, we extracted the time when the response of the cAMP biosensor reaches the highest value, and then denoted this time as the cAMP peak time. In [32], two types of cAMP biosensors were used. The AKAP/AC nanodomain-specific cAMP biosensor provides the cAMP peak time on the AKAP/AC nanodomain, and the plasma membrane-specific cAMP biosensor gives the cAMP peak time outside the AKAP/AC nanodomain. These peak times from two types of cAMP biosensors are the y-coordinates in Fig 3B. Next, we extracted the time when the response of the Ca2+ biosensor reaches the highest value and lowest value, and denoted them as Ca2+ peak and Ca2+ trough time respectively (x coordinates in Fig 3B). We found that the cAMP peak time on the AKAP/AC nanodomain is close to Ca2+ peak time (left panel in Fig 3B), suggesting the in-phase Ca2+-cAMP oscillation. However, the cAMP peak time outside the AKAP/AC nanodomain is the same as the Ca2+ trough time (right panel in Fig 3B). This means that the phase difference between the Ca2+ and cAMP is exactly half of the period, which is inversely out-of-phase. Therefore, these experiment results are consistent with the numerical predictions from our model with respect to the in-phase and inversely out-of-phase Ca2+-cAMP oscillation. It should be noted that, although the in-phase and out-of-phase behaviors have been extensively studied in [32], the property that Ca2+ trough and cAMP peak occurs simultaneously in the out-of-phase behavior was first discovered in this work.
2.3 A simple incoherent feedforward loop explains in-phase oscillation and inversely out-of-phase oscillation
The above analyses demonstrate the role of AC* and PDE* in regulating the phase difference between Ca2+ and cAMP. However, the complexity of the biochemical circuit involved in the regulation of Ca2+ and cAMP dynamics (Fig 1C) prevents us from identifying if AC* and PDE* are sufficient conditions for such phase regulation. We next investigated if a simple circuit model could shed light on the main control elements of the phase difference. The signal, which mimics Ca2+ concentration, is set to be a sine wave plus a constant 1.1, where the constant 1.1 ensures the positive sign of the signal (Fig 4A). Furthermore, we assumed that the signal affects output paradoxically. On one hand, the signal improves the production of output and on the other hand, the signal increases the output degradation (Fig 4B). The circuit with two pathways with opposite effects from the input to the output is usually called as incoherent feedforward loop [54], and can exhibit different biological functions under certain conditions [54–61]. These regulatory reactions are used to mimic the effect of Ca2+ on cAMP caused by AC* and PDE*: Ca2+ enhances the cAMP production by the activation of AC; at the same time, Ca2+ improves the cAMP degradation by the activation of PDE. Therefore, the output can be regarded as the cAMP. The mathematical model of this simple incoherent feedforward loop is written as follows:
(1) |
where S(t) and x denote the concentration of signal and output, respectively. v is the activation strength caused by signal, k1 is the basal production rate constant, m is the degradation rate constant and k2 captures the feedback of x on the degradation. Without loss of generality, we can set m = 1 and obtain (see Simple ODE model subsection in Methods for details):
(2) |
The solution of this first-order linear ODE when t goes to infinity can be written as follows (see Simple ODE model subsection in Methods for details):
(3) |
where I(s) is . F(t) is defined as for simplicity.
The integral in the Eq (3) lacks a closed form solution. Therefore, we turned to numerical solutions to evaluate how the different parameters affected the output as a function of the input. We varied the value of the activation strength v, which mimics the effect of AC*, from small (away from the nanodomain) to large (center of the nanodomain). We postulated that if the oscillation behavior of x(t) is inversely out-of-phase with small v and then becomes in-phase with increased v, then this simple incoherent feedforward loop is sufficient to generate similar results observed in the 3D reaction-diffusion model. We found that, with small values of the basal production rate constant k1 and the x-induced degradation rate k2, increasing activation strength v cannot lead to the transition from inversely out-of-phase to in-phase behavior between the input and output (Fig 4C(i) and 4C(ii)), but given large k1 and k2 we can recover the proposed v dependence (Fig 4C(iii)). In Fig 4C(iii), when v is small, the output peak time is the same as the signal trough time (the dashed line in Fig 4C(iii)), indicating the inversely out-of-phase oscillation; when v is large, the output peak time is the signal peak time (the solid line in Fig 4C(iii)), suggesting an in-phase oscillation. While in-phase oscillation and inversely out-of-phase oscillation are reproduced by the simple ODE model, the transition between these two behaviors is not exactly the same as the full model in Fig 2A. As explained previously, the sharp transition in the full model is caused by the swapping between in-phase and out-of-phase peaks, but in the simple ODE model we didn’t observe the emergence of two peaks. The failure to observe two peaks in the simple ODE model may be because this simple ODE model neglects other reactions that also consume cAMP, such as the binding of cAMP and R2C2 (v14 in Fig 1C), the binding of cAMP and AKAP–R2C2 (v19 in Fig 1C).
To quantitatively identify the phase difference, we calculated the time delay of peaks between x(t) and S(t) for several combinations of k1 and k2 (Fig 4D). The relationship between the time delay and phase behavior is as follows—if the time delay is half of the signal period, that is, π, the inversely out-of-phase oscillation is achieved while zero time delay corresponds to the in-phase oscillation. We found that the x-induced degradation rate k2 is more important than the basal production rate constant k1 in producing the desired v dependence. If k2 is small, it is impossible to achieve inversely out-of-phase and in-phase oscillations even with different values of k1 (Fig 4D(i) and 4D(ii)); only the large k2 can lead to inversely out-of-phase and in-phase oscillations under different values of v (Fig 4D(iii)). Furthermore, the role of the basal production rate constant k1 is to shift the critical value of activation strength v that causes the transition from the inversely out-of-phase to the in-phase. These numerical results are also validated by analyzing the Eq (3). Since , the time when x(t) = 0 is the same as that when F′(t) = 0. Therefore, the peak time of x(t) can only be the peak time or the trough time of F(t), that is, only two values. This is consistent with the numerical simulations that there are only two values of time delay in Fig 4D with varied v. Moreover, the peak time of x(t) is the trough time of F(t) when and is the peak time of F(t) when (see Simple ODE model subsection in Methods for details), corresponding to the time delay before and after transition in Fig 4D, respectively. Especially, when k2 is large enough, F(t) has the same peak and trough time as the signal S(t) (see Simple ODE model subsection in Methods for details), and thus the peak time of x(t) is the trough time of the signal S(t) when and the peak time of S(t) when , consistent with inversely out-of-phase and in-phase behaviors observed in Fig 4D(iii). In conclusion, the exploration of the simple incoherent feedforward loop not only validates the sufficiency of the role of AC* and PDE* in driving phase difference, but also suggests the mechanism of inversely out-of-phase and in-phase oscillations—a large basal degradation rate (i.e., k2 in the Eq (2)). Thus, by simulating the dynamics of the output governed by the Eq (2), we validated that such a simple incoherent feedforward loop is sufficient to produce in-phase and inversely out-of-phase oscillations.
2.4 Simulation results predict the role of AC in driving the in-phase behavior and decreasing time delay for inversely out-of-phase behavior
Since the above analyses for a simple incoherent feedforward loop suggest the role of AC* in controlling the phase behavior of Ca2+-cAMP, we tested whether the AC activity in the 3D reaction-diffusion model also plays a similar role. Therefore, we decreased the maximal concentration of AC to 75%, 50%, and 25% of the original value, and then simulated the 3D reaction-diffusion model (Fig 5A–5C). We found that the in-phase behavior on AKAP/AC nanodomains and inversely out-of-phase behavior outside AKAP/AC nanodomains still exist for the 75% and 50% cases (Fig 5A, 5B and 5D), but only inversely out-of-phase behavior appears for the 25% case (Fig 5C and 5D). These results are consistent with the effect of activation strength v, because a small value of v usually leads to out-of-phase behavior. In addition to the effect of AC activity on the phase difference between Ca2+ and cAMP, AC activity also influences the time delay for the inversely out-of-phase behavior, that is, the higher the AC activity is, the shorter the time delay is (Fig 5E).
2.5 Cellular compartment size determines the time delay for the inversely out-of-phase calcium-cAMP oscillation
Having analyzed the biochemical mechanisms that regulate the Ca2+-cAMP phase difference, we next turn our attention to the biophysical properties that may regulate the spatiotemporal dynamics of cell signaling, including the compartment size to investigate how it influences this phase difference. Since the cell is a 3D structure, the height and width of the compartment reflect the cell size to some degree: the higher or wider the compartment, the larger the probability that a cell will have a large size. Researchers have identified the cell shape as an important factor in regulating cellular signaling [62–64], and thus we expect that the compartment size also affects the oscillation behavior. First, we tested the effect of the compartment size by doubling the height (Fig 6A), doubling the width (Fig 6B), or doubling both the height and width (Fig 6C). This effectively changes the surface-to-volume ratio of the cellular compartment. Similar to the original compartment (Fig 2A), these large compartments all lead to in-phase cAMP oscillation on the AKAP/AC nanodomains and inversely out-of-phase cAMP oscillation outside the AKAP/AC nanodomains (Fig 6A–6D). However, the time delay for the inversely out-of-phase behavior differs (Fig 6E): it is near 25 seconds for the original compartment (black dashed line in Fig 6E), 50 seconds for the compartments with the double width or double height (red and cyan lines in Fig 6E), and 120 seconds for the compartment with double width and height (yellow line in Fig 6E). The compartment with the double width and that with double height have different sizes but similar time delay, indicating that the compartment size cannot monotonically determine the time delay (also see S3 Fig). Nevertheless, by comparing these large compartments with the original compartment, we predict that the time delay for the inversely out-of-phase cAMP oscillation depends on the compartment size, and the large compartment size tends to induce a long time delay. Thus, the cell with a large volume tends to exhibit a long time delay. Although this prediction is not based on a whole-cell simulation, it fits the intuition that the large cell may take a long time to transform the signal as the molecule needs a long time to diffuse to specific locations.
2.6 Prediction of the time delay from idealized Turing pattern-based AKAP/AC nanodomains
Thus far, we focused on the impact of preexisting nanodomains. We next investigated the conditions that could lead to the formation of nanodomains. Due to the clustered distributions of AKAP and AC molecules ([32] and S4 Fig), we considered the patterning model to describe the AKAP/AC nanodomain formation. Although AC molecules do not form a periodic pattern according to the diverse distance between two nearest nanodomains and non-uniformed nanodomain size (S4 Fig), we still chose the Turing model. The reasons for this model choice are as follows: 1) the significant immobile fraction (near 42.3%) of AC molecules (Figure 3—figure supplement 3 in [32]) is consistent with the long-lived property of Turing patterns; 2) the STORM data for AKAP and AC in different cell types is limited [65, 66], and thus the identification of the pattern formed by AC and AKAP lacks solid evidence. Therefore, we chose the Turing model for mathematical tractability.
Inspired by work done by others on the formation of membrane receptor nanodomains [48, 67–70], we focused on a Turing model with steric interactions [48]. This model is summarized as follows (Fig 1E):
(4) |
In the Eq (4), r and s denote the normalized AC and AKAP concentration, respectively. AKAP works as a scaffold to recruit AC to cell membrane, and AC inhibits the recruitment of AKAP by steric repulsion similar to the model presented in [48]. and are the value of r and s at homogeneous steady state. b and β are the unbinding rates of the AC and AKAP from the cell membrane, respectively; m1 and m2 are the binding rates of the AC to cell membrane caused by itself and the AKAP, respectively; μ is the self-recruitment rate of AKAP; Dr and Ds are diffusion coefficients. The values of these kinetic parameters are listed in S7 Table. However, the assumption that AKAP/AC nanodomains are formed by the Turing model is a hypothesis, where future experiments are required to validate the existence of the above biochemical reactions. The values of these kinetic parameters may also change with the dynamics of nanodomains.
We simulated the above Turing model in a hexagonal domain of 800 nm for different combinations of kinetic parameters (Fig 7A). We found that three patterns arise: a homogeneous steady state where AC (denoted by r) and AKAP (denoted by s) are both uniformly distributed (lower left panel in Fig 7A), Turing pattern 1 where AC and AKAP are co-localized (lower middle panel in Fig 7A), and Turing pattern 2 where AC and AKAP repel each other (lower right panel in Fig 7A). We found that the Turing pattern usually occurs when the ratio of the diffusion coefficient of AC to the diffusion coefficient of AKAP (i.e., Dr/Ds) is large (area in red and yellow in the upper left panel in Fig 7A), which is consistent with the condition of Turing instability. However, the large ratio leads to the transition from Turing pattern 1 to Turing pattern 2. Taken together with the fact that AC is co-localized with AKAP in MIN6 β cells [32], our model predicts that the ratio of diffusion coefficients Dr/Ds is a critical determinant of nanodomain formation with Turing pattern 1. We note that while Turing pattern 2 is a mathematically admissible solution, its relevance to cell signaling is as yet unknown. Furthermore, colocalization and nanodomain formation are not sensitive to the unbinding rate of the AC (denoted by b) but have a strict requirement on the unbinding rate of the AKAP, binding rate of the AC to cell membrane caused by AKAP, and the self-recruitment rate of AKAP (denoted by β, m2, and μ, respectively) (upper middle and upper right panels in Fig 7A).
We next investigated the phase behavior of Ca2+-cAMP in these Turing pattern-based nanodomains. To do so, we used the Turing pattern in the lower middle panel in Fig 7A as the initial condition of AKAP/AC nanodomains (left panel in Fig 7B; also see Turing pattern-based nanodomain subsection in Methods for details). In this case, we have experimentally informed patterns but do not know the concentrations of the species. Since the simple ODE model indicates biochemical activity is important, we used the concentration of AC as free parameters and conducted a parameter variation. During this parameter variation, the largest concentration of AKAP on the AKAP/AC nanodomain was fixed to 9600 copies /μm2 to ensure the consistency with simulation results in Fig 6. The first rescaling set the largest concentration of AC on the AKAP/AC nanodomain to 9600 copies /μm2, which was used in Fig 6. Given this re-scaling, the in-phase cAMP oscillation exists not only on AKAP/AC nanodomains but also outside AKAP/AC nanodomains (upper right panel in Fig 7B). This is consistent with our previous finding of in-phase oscillations on the AKAP/AC nanodomain but not consistent with the inversely out-of-phase oscillations outside the AKAP/AC nanodomain. Next, we decreased the concentration of AC to 12.5% of the previous value, and thus the largest AC concentration on the AKAP/AC nanodomain becomes 1200 copies /μm2. We found the emergence of inversely out-of-phase cAMP oscillation outside AKAP/AC nanodomains (lower right panel in Fig 7B, 7C and 7D). Thus, our model predicted that Turing pattern 1 combined with the ratio of AC to AKAP is critical in regulating the phase behavior of Ca2+-cAMP. This prediction demonstrates the importance of both pattern formation and biochemical activity in Ca2+-cAMP oscillation.
2.7 Prediction of the time delay from a realistic distribution of AKAP/AC nanodomains
Finally, we extended our analysis to the realistic distribution of AKAP/AC nanodomains from AC8 STORM data in [32]. This analysis allows us to relax the previous assumptions of the nanodomains as either Gaussian distribution or Turing pattern and test our models in experimentally observed nanodomain distributions. To do this, we selected one square area with a length of 1000 nm from the STORM data (left and middle panels in Fig 8A); and then smoothed the STORM data in the square area by convoluting with a Gaussian function whose FWHM (full width at half maximum) is 60 nm (right panel in Fig 8A). After these two processes, we obtained a smooth distribution of AC8 on a square area with a length of 1000 nm. Next, the AC8 distribution in a hexagonal region denoted by the white dashed line (right panel in Fig 8A) is used as the initial distribution of AC in our simulations. Similar to the case with Turing pattern-based AKAP/AC nanodomain, the realistic concentration of AC is unknown. As suggested by the model results with Turing pattern-based AKAP/AC nanodomain, the AC activity plays an important role in the phase behavior of Ca2+-cAMP, and thus we also tested two values of rescaled AC here: the first is that the largest value of AC concentration on the AKAP/AC nanodomain is rescaled to 9600 copies/μm2 (upper right panel in Fig 8B), and the second is 4800 copies/μm2 (lower right panel in Fig 8B). Additionally, the initial distribution of AKAP is always rescaled to ensure the highest concentration on AKAP/AC nanodomain of 9600 copies/μm2.
The two different rescalings of the AC activity result in distinct phase behavior. The first rescaling shows in-phase Ca2+-cAMP oscillation on and outside the AKAP/AC nanodomains (upper right panel in Fig 8B). For the second rescaling of 4800 copies/ m 2, the in-phase cAMP oscillation on AKAP/AC nanodomains and inversely out-of-phase cAMP oscillation outside the AKAP/AC nanodomains arise again (lower right panel in Fig 8B, 8C and 8D). These in-phase and inversely out-of-phase cAMP oscillations also appear when smoothing the AC STORM data with a Gaussian function whose FWHM is 30 nm (S5 Fig), because the AC activity with a 30 nm FWHM is less than or equal to that with 60 nm FWHM at any location. Therefore, our model predicts that the AC activity is essential for the phase difference between Ca2+-cAMP even for a realistic distribution of AKAP/AC nanodomains.
3 Discussions
Signaling nanodomains and biomolecular condensates are ubiquitous within the cell. They regulate signaling transduction pathways by sequestering biochemical reactions. However, how these signaling aggregates precisely regulate intracellular pathways remains elusive, especially when taking the nonuniform size and distribution of these signaling aggregates in the cell into consideration. Here, we looked at the intersection of AKAP/AC nanodomain and the Ca2+/cAMP pathway in detail and studied the underlying biochemical mechanisms and potential biophysical effects.
We developed a 3D reaction-diffusion model and different types of preexisting AKAP/AC nanodomains to study how AC and PDE activities, compartment size, and pattern of AKAP/AC nanodomains affect the phase shift between Ca2+ and cAMP oscillation. Through simulating the 3D reaction-diffusion model with a Gaussian distribution-modeled AKAP/AC nanodomain, we revealed the paradoxical role of AC and PDE: active AC and active PDE drive in-phase oscillation on AKAP/AC nanodomains and out-of-phase oscillation outside AKAP/AC nanodomains, respectively (first and second columns in Fig 8E). It should be noted that the out-of-phase oscillation exhibits the same Ca2+ trough time and cAMP peak time, which is validated by the close match of Ca2+ trough time and cAMP peak time in the experiment. Such paradoxical role of AC and PDE in regulating the phase shift between Ca2+ and cAMP can be explained by a simple incoherent feedforward loop and predicted by varying AC activity in the 3D reaction-diffusion model. Furthermore, the important role of AC in regulating phase delay between Ca2+ and cAMP has been demonstrated by knocking down AC, overexpressing AC, or disturbing the interaction between AKAP and AC (Figure 2(d-e) and Figure 5(b) in [32]). Moreover, these in-phase and inversely out-of-phase behaviors persist in most cases, but the time delay for the inversely out-of-phase behavior depends on the compartment size (third column in Fig 8E). Furthermore, the simulation results with Turing pattern-based and AC STORM data-based AKAP/AC nanodomains predicted the important role of nanodomain pattern and AC activity in coordinating the phase between Ca2+ and cAMP.
Interestingly, we found that the phase shift between Ca2+ and cAMP is controlled by an incoherent feedforward loop: the activating Ca2+-AC-cAMP pathway and the inhibiting Ca2+-PDE-cAMP pathway (Fig 8F). Although the cAMP can regulate the Ca2+ level by a feedback loop through PKA (Fig 8F), it seems that the difference in oscillatory dynamics of cAMP at distinct cellular locations does not result in diverse Ca2+ dynamics due to the fast diffusion of Ca2+. Several well-known biological functions of the incoherent feedforward loop include achieving adaptation [59, 60, 71], regulating structural plasticity [61], processing oscillation [55], detecting fold-change [57], generating non-monotonic gene input functions [56], and facilitating adaptive tuning of gene expression [58], but the phase regulation of second messengers is first discovered in this work to our knowledge. Since the cell decodes information involved in second messengers to regulate gene transcription [14, 31, 72], the incoherent feedforward loop may play a broad role in gene expression. In addition, a finite phase difference between cAMP and Ca2+ has been frequently observed in experiments [46, 53], suggesting the limited contribution of AKAP/AC nanodomain in governing the average cAMP and Ca2+ dynamics within the cell.
The exploration of mechanisms of in-phase and inversely out-of-phase behavior in this work may help to understand the synchronization in oscillatory systems. The synchronization, which is widely observed in circadian clocks [73], long-term memory [74, 75], and collective behavior in cell populations [76, 77], is referred to as the locked phase for several interacting biological oscillators. Depending on the phase difference, the synchronization includes in-phase synchronization, anti-phase (or inversely out-of-phase) synchronization, and synchronization with an arbitrary phase shift. Therefore, if we regarded the Ca2+ and cAMP as two biological oscillators, their in-phase and inversely out-of-phase behavior are two types of synchronization. Our numerical results suggest that improving the activation strength from one oscillator to the other (v in Eq (2)) can only lead to a sudden change from inversely out-of-phase synchronization to in-phase synchronization rather than a continuous change. Furthermore, a continuous change in phase difference can be achieved by tuning the basal inhibition rate of the downstream oscillator (k2 in the main text), and the large rate results in the in-phase or inversely out-of-phase synchronization.
Our work demonstrates the role of biophysical properties of the cell in Ca2+-cAMP oscillation, which is another evidence that the biophysical properties of cells encode the signaling information. One important biophysical property of cells is cell shape. For example, cell aspect ratio affects cytosolic calcium levels in vascular smooth muscle cells [78]; the structure and the shape of the dendrite spine regulate AMPAR dynamics and energy landscape [79, 80]; the cell eccentricity encodes the information of growth factor receptor pathways [62]. While most studies focused on the change in the increase or decrease of intracellular signaling levels, our work observed the change in period and phase for oscillatory signaling molecules. Apart from the cell shape, the distribution of AKAP/AC nanodomains also plays a role in Ca2+-cAMP oscillation. This result implies the importance of the signaling hubs, which has been revealed in many biological systems [81–84]. The major function of signaling hubs is to sequester the chemical reactions to ensure their occurrence in the right locations, and AKAP/AC nanodomains also exhibit similar behavior: the activation of AC can only happen on AKAP/AC nanodomains, and thus guarantees the high value of cAMP on AKAP/AC nanodomains.
The phase behavior of Ca2+-cAMP in pancreatic β cell may be crucial to insulin secretion. The cAMP not only affects the Ca2+ level as modeled in this work but also regulates PKA and guanine exchange factor Epac, two types of signaling molecules that contribute to secretory granule trafficking and exocytosis. On the one hand, both PKA and Epac improve the mobilization of secretory granules, resulting in fast replenishment of readily releasable granules [85–87]. On the other hand, several proteins involved in the exocytosis process have been identified as substrates of PKA or interacted with Epac, such as SNAP-25 and Snapin [88–91]. Moreover, AKAP nanodomains can anchor both PKA and Epac [92, 93], and thus may drive the oscillation of PKA and Epac during pulsatile insulin secretion. In addition to the potential effect in insulin secretion, the out-of-phase Ca2+-cAMP behavior in pancreatic β cell encode the information in extracellular signals [53].
The AKAP/AC nanodomains mediate the phase difference between Ca2+ and cAMP. Such nanodomains have been extensively found in the intracellular space as well as biomolecular condensates [94–97]. These signaling nanodomains and biomolecular condensates are able to directly regulate cellular processes, and dysregulations of these complexes are likely indicated as potential drivers of oncogenic activity. For example, the tumor suppressor SPOP (speckle-type POZ protein) granules are disrupted in the presence of SPOP mutations that promote prostate cancer progression [98, 99]; the stress granules, which are composed of mRNA transcripts, RNA-binding proteins, and translation machinery to prevent cell apoptosis from stressful conditions, are markedly upregulated in cancer cells, resulting in more stress-resistant cancer cells [100, 101]; the promyelocytic leukemia (PML) nuclear bodies have been found to contribute to the telomere lengthening in cancer cells, thus improving the replicative ability of cancer cells [102, 103]; the flat clathrin lattice (FCL) is located on the cell membrane, providing sustained signaling hubs for epidermal growth factor receptor (EGFR), and the disruption of FCL leads to a decreased level of active EGFR on the cell membrane [81].
In addition to its application in receptor nanodomains, the Turing model is commonly employed to understand mechanisms underlying a wide range of biological patterns such as sea shell patterns, hair follicle spacing, hydra regeneration, lung branching and digits formation [104–107]. With a simple interaction network incorporating a short-range positive feedback and a long-range negative feedback (i.e., a slowly diffusing activator and a rapidly diffusing inhibitor), the Turing model is able to reproduce diverse spatial patterns. For example, a simple Turing reaction-diffusion model consisting of the phosphorylated and unphosphorylated species of Polo-like-kinase 4 (PLK4), is used to explain how mother centrioles break symmetry to generate a single daughter [108]. When taking the intrinsic noise into account, the stochastic version of Turing model can further relax parameter constraints such as the ratio of activator–inhibitor diffusion coefficients, which is validated in a synthetic bacterial population where the signaling molecules form a stochastic activator–inhibitor system [109].
Although we used the Turing model to explain the formation of AKAP/AC nanodomains, it is not the only one that can generate the pattern. For example, the particle-based model [67–70], lattice model [110, 111], and Cahn–Hilliard equation [112–114] can also produce patterns. The Cahn–Hilliard equation is well-known as the model for phase separation, but recently it has been used to model the formation of nanodomains on cell membranes. We also note that the Turing pattern in this work is a qualitative match but not a quantitative one, since the Turing pattern was not exactly fitted to experimental data: the spacing between two domains and size (black lines in the lower middle panel in Fig 7A) are 365 nm and 118 nm, respectively, while those in the experiment are 341 nm and 105 nm. This may be because the parameter settings of mean-shift algorithm when identifying the AC clusters from AC STORM data are different, or because the staining for the STORM data is not perfect, which causes significant artificial variations in domain spacing and size. Moreover, the spacing between two adjacent nanodomains in Turing pattern-based nanodomains is fixed in our study, and thus the effect of interactions between two adjacent nanodomains on Ca2+/cAMP oscillation behavior is not considered. According to the almost same time delay when the distance of two AKAP/AC nanodomains varies (S6 Fig; also see Interaction between two nanodomains subsection in Methods), we anticipate that the oscillation behavior is not affected by the distance between nanodomains under certain circumstances. In addition to the limitation in choosing the patterning model, another limitation in our work is that we did not consider the existence of RIα (a subunit of PKA) condensates in the cytosol, which can accumulate cAMP locally [84]; future work is expected to incorporate RIα condensates to fill the knowledge gap. Another limitation is that we neglected the biological noise in the Ca2+/cAMP circuit, because the number of molecules can be small within the cell. For example, there are 15 to 150 individual AC8 molecules in one AKAP/AC nanodomain [32]. We anticipate the emergence of novel behavior and mechanisms when modifying the deterministic model to a stochastic one [115–119]. Despite these limitations, our findings of Ca2+-cAMP interacting through an incoherent feedforward loop to regulate their in and out-of-phase oscillations on AKAP/AC nanodomains sheds light on how spatial arrangement of molecules can regulate temporal dynamics of second messengers in different cellular locations.
4 Methods
4.1 Reaction-diffusion model
To investigate the underlying biochemical mechanism and the effect of biophysical properties of the cell as mentioned in Fig 1B, we used the 3D reaction-diffusion model in [32] to describe the dynamics of Ca2+, cAMP, and other molecules that interact with Ca2+ and cAMP. Next, we introduced this model and listed the equations below.
4.1.1 Compartment
As mentioned before, though calcium and cAMP are mostly distributed in the cytosol, they both can interact with molecules on the cell membrane, such as AKAP and AC. To take these chemical reactions on cell membrane into consideration, we chose a small compartment in one cell, that is, a hexagonal prism (Fig 1C), where the top surface (denoted by Γ) represents the cell membrane and the volume under the top surface (denoted by Ω) the cytosol. Besides, the height and width of the compartment are set to be 600 nm and 400 nm respectively, which are consistent with those in [32]. However, in order to study the effect of the compartment size on the time delay between Ca2+ and cAMP, the height and width may vary, which are labeled in figures and illustrated in the main text.
4.1.2 Chemical reactions and governing equations
For the Ca2+-cAMP circuit, we considered 20 chemical species in total (Fig 1C). On the one hand, there are 13 chemical species in the cytosol (non-highlighted species in Fig 1C), including calcium, cAMP, CaM (calmodulin), PDE (Cyclic nucleotide phosphodiesterase), PKA (protein kinase A), and complexes of these molecules. On the other hand, there are 7 chemical species or quantities on the membrane (highlighted in yellow in Fig 1C), such as AC (Adenylyl cyclase), AKAP, and complexes of these molecules. These molecules in the cytosol and on the membrane have been proven to play crucial roles in regulating the Ca2+-cAMP pathway [8, 120, 121]. For the full list of species, we refer the readers to S1 Table.
The dynamics of these 20 chemical species are governed by chemical reactions. The Ca2+ in the cytosol can bind to CaM, leading to the activation of inactive PDE and inactive AC (v1–8). Since the inactive AC has been validated to be co-localized with AKAP by forming AKAP/AC nanodomains on the cell membrane [32], the activation of inactive AC caused by the Ca2+/CaM complex is only distributed on the AKAP/AC nanodomain. Then, the active AC (denoted as AC*) on AKAP/AC nanodomains can improve the synthesis of cAMP (v9), while the active PDE (denoted as PDE*) degrades the cAMP in the cytosol (v10–13). Under the stimulus of cAMP, PKA in the cytosol is activated (v14–15). Another pathway to activate PKA is localized at the AKAP/AC nanodomain: AKAP binds to inactive PKA (v16–18) in the absence of cAMP, and these inactive PKA will also become active and dissociate with AKAP in the presence of cAMP (v19–20). Those active PKA can increase the synthesis of Ca2+ (v23). Ca2+ level is also increased via the influx across the plasma membrane (v22), and Ca2+ regulates the Ca2+ gated K+ channel (v21). For the full list of chemical reactions, we refer the readers to S2 Table.
According to the type of reaction, these reactions can be categorized into two classes: the first type is the production, degradation, or binding event (solid arrows in Fig 1C), and the other is the regulatory reaction (dashed arrows in Fig 1C). For the first type, the species before and after the arrow are the reactant and product, respectively. For example, AC + Ca2CaM CaM AC means that the AC can bind to Ca2CaM and thus form the complex CaM AC. However, for the second type, the species before the arrow regulates the chemical reaction instead of working as the reactant; for instance, the reaction v12 means that the active PDE enhances the cAMP degradation. Due to the meaning of each type, one key difference between these two types is that the species before the arrow will be consumed for the first type of reaction but not for the second type of reaction.
To describe the dynamics of 20 chemical species in the compartment, the 3D diffusion-reaction model in [32] is adopted, which is shown as follows:
(5) |
where [⋯] denotes the concentration of species. The Γ is the top surface of the compartment, and Ω is the volume under Γ (see Fig 1C). The term on the right-hand side is the time derivative, and that on the left-hand side is composed of reaction parts and diffusion parts (DX ∇2[X], X = Ca2+, CaM, ⋯, PKA). For reaction parts, the ji (i = 1, 2, …, 20, 22, 23) represent the flux of reactions vi; (see S2–S4 Tables). Note that the flux of the production, degradation or binding event (solid arrows in Fig 1C) will lead to not only the consumption of reactants but also the production of products. However, the reaction that regulates other reactions (dashed arrows in Fig 1C) only affects the reaction rate while maintaining the concentration of regulator species. As for diffusion parts, the DX (X = Ca2+, CaM, ⋯, PKA) denotes the diffusion coefficient of species X (see S5 Table for exact values); the ∇2 is the Laplace operator representing ; the denotes the diffusion on the cell membrane Γ and is defined as .
As for the boundary condition, most of species are assumed to have the zero flux on the boundary Γ, and those with non-zero fluxes on Γ are listed as follows:
(6) |
where denotes the exterior normal vector to the membrane Γ. These non-zero fluxes result from the reactions whose reactant or product is located on the cell membrane Γ. For the six surfaces surrounding the compartment, periodic boundary conditions are applied. The bottom surface is assumed to have zero fluxes.
The initial condition for species in the cytosol is set to be uniformly distributed (see S6 Table for exact values). For those species on the membrane, their initial conditions are set to be 0 except for AKAP and AC (S6 Table). Since AKAP can form nanodomains [32], we used the following Gaussian distribution to model one AKAP nanodomain:
(7) |
where σ and A are set to be 25 nm and 6.1446E11 copies/nm respectively. The x0 and y0 denote the x and y coordinates of the nanodomain center. For example, in Fig 1D where the AKAP/AC nanodomain is at the center of the cell membrane Γ, the x0 and y0 are both zero. Moreover, the initial concentration of AC on the cell membrane is set to be the same as that of AKAP, unless otherwise specified. Besides, the derivative of the concentration with respect to time for all species is set to be 0 at time 0 (S6 Table).
4.1.3 Voltage module and governing equations
As we mentioned above, the Ca2+ level can be increased by the influx via Ca2+ channel (v22). Therefore, we used a voltage module where the Ca2+ channel is modeled by an electrical conductance. In addition to Ca2+ channel, Ca2+ gated K+ channel, K+ channel, and the leak channel are also included. Currents for ion channels are represented by Ii, i = Ca, KCa, K, and L. The cell membrane is modeled by a capacitor with capacitance Cm. In the Hodgkin-Huxley framework, the membrane current is the sum of all the contributions from ion channels, i.e., , where V is membrane voltage. Besides, the membrane voltage V is assumed to diffuse on the cell membrane with the diffusion coefficient DV. Thus, the dynamics of membrane voltage V on Γ is governed by
(8) |
where , , and w are the steady-state fraction of open Ca2+ gated K+, Ca2+, and K+ channels, respectively. For the specific i (i = KCa, Ca, K, L) channel, gi and Ei denotes the conductance and reversal potential, respectively. The governing equation for the K+ channel open probability w on Γ is as follows:
(9) |
where is the time constant, is the steady-state fraction of open K+ channel. ϕ is the factor that controls the relative time scales of V and w. See S4 Table for the value of constants KKCa, v1, v2, v3, v4, ϕ, gi, and Ei (i = KCa, Ca, K, L).
4.1.4 Coupling between chemical reactions and voltage module
The coupling between chemical reactions and voltage module is achieved by two reactions: the regulation from Ca2+ concentration to the Ca2+ gated K+ channel (v21), and the Ca2+ flux through Ca2+ channel (v22). For the reaction v21, Ca2+ concentration determines the steady-state fraction of open Ca2+ gated K+ channel in a Hill function form (See S4 Table for parameter values), i.e., in the Eq (8). For the reaction v22, the Ca2+ flux through Ca2+ channel is modeled by j22 = (− αICa − vLPM[Ca2+])(1 + kPKAV[PKA]) (See S4 Table for parameter values), which affects the boundary condition of Ca2+ in the Eq (6).
4.1.5 Numerical simulations
To simulate the dynamics of all molecules in the above mathematical model, we used COMSOL Multiphysics7.1 to numerically solve Eqs (5), (6), (8) and (9). The COMSOL files are available at https://github.com/RangamaniLabUCSD/Qiao_et_al_cAMP-Calcium-on-AKAP-AC-nanodomains.
4.2 Simple ODE model
To simplify the Eq (1), the time t is multiplied by m, and v and k1 are normalized by m. For the simplicity of notation, we kept the notation of parameters and obtained the following equation:
It can be seen that the parameter m only influences the period of the signal. Since the period of Ca2+ is the same in the entire compartment, we can fix the value of m to 1 and get the Eq (2).
Next, we will derive the solution of the Eq (2). Let , then the equation for is shown as follows:
By using the variation of parameters method, we can obtain the expression for :
where I(t) is . Note that , and thus the x(t) can be written as follows:
When t goes to infinity, the term will disappear because goes to zero. Therefore, the expression of x(t) when t goes to infinity is written as follows:
which is exactly the Eq (3).
Then, we analyzed how the peak and trough time of x(t) depends on kinetic parameters v, k1, k2 and the signal S(t). For simplicity, we use F(t) to denote . Then, we used tmax,x to denote the peak time of x(t), and the following equations for tmax,x hold:
By replacing x(t) with the Eq (3), we obtain:
Therefore, if , the peak time of x(t) is the trough time of F(t); if , the peak time of x(t) is the peak time of F(t). This explains the reason for only two values of time delay when increasing the activation strength v and the transition point in Fig 4D. Due to the close relation between x(t) and F(t), we then analyzed the peak and trough time of F(t). Recall that , and thus the due to I(t) ≪ k2t when k2 → +∞. We integrated the by parts, and then let t goes to infinity, leading to the following equation for F(t):
Thus, the peak and trough time of F(t) are the same as the signal S(t). Combined with the relation between x(t) and F(t), we obtained that if k2 is large enough, the peak time of x(t) is the tough time of S(t) for and is the peak time of S(t) for , which is consistent with results in Fig 4D(iii).
4.3 Turing pattern-based nanodomain
In order to use the Turing pattern as the initial condition of AKAP/AC nanodomain, we assumed that the formation of AKAP/AC nanodomains is much faster than reactions in the Ca2+-cAMP circuit, and thus we directly used the Turing pattern in the lower middle panel in Fig 7A. Note that we did not couple Eqs (4) into (5). We also assumed that the diffusion coefficients of AKAP and AC are zero once the Turing pattern is formed, corresponding to the zero diffusion terms in the equations for AKAP and AC in the 3D reaction-diffusion model.
4.4 Interaction between two nanodomains
We investigated how the distribution of AKAP/AC nanodomains affects the phase difference between Ca2+-cAMP. We extended our model to include two AKAP/AC nanodomains on the cell membrane but with three distinct distances 240 nm (S6 Fig), 320 nm, and 400 nm. We found that the in-phase cAMP oscillation on the AKAP/AC nanodomains and inversely out-of-phase cAMP oscillation outside the AKAP/AC nanodomains still hold (S6 Fig). Furthermore, the time delay for the inversely out-of-phase case is always near 75 seconds although the distance between two AKAP/AC nanodomains is distinct (S6 Fig). This result suggests that the distance between two AKAP/AC nanodomains has no effect on the time delay.
Supporting information
Data Availability
All the code to generate data and figures is available from: https://github.com/RangamaniLabUCSD/Qiao_et_al_cAMP-Calcium-on-AKAP-AC-nanodomains.
Funding Statement
This work was supported by the National Institute of Health Grant R01 DK073368 (to J.Z.), the Air Force Office of Scientific Research (AFOSR) Multidisciplinary University Research Initiative (MURI) Grant FA9550-18-1-0051 (to P.R.), and Army Research Office W911NF2310249 to (P.R.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
References
- 1. Beavo JA, Brunton LL. Cyclic nucleotide research—still expanding after half a century. Nature Reviews Molecular Cell Biology. 2002;3(9):710–717. doi: 10.1038/nrm911 [DOI] [PubMed] [Google Scholar]
- 2. Drago I, Stefani DD, Rizzuto R, Pozzan T. Mitochondrial Ca2+ uptake contributes to buffering cytoplasmic Ca2+ peaks in cardiomyocytes. Proceedings of the National Academy of Sciences. 2012;109(32):12986–12991. doi: 10.1073/pnas.1210718109 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Giorgi C, Marchi S, Pinton P. The machineries, regulation and cellular functions of mitochondrial calcium. Nature Reviews Molecular Cell Biology. 2018;19(11):713–730. doi: 10.1038/s41580-018-0052-8 [DOI] [PubMed] [Google Scholar]
- 4. Clapham DE. Calcium Signaling. Cell. 2007;131(6):1047–1058. doi: 10.1016/j.cell.2007.11.028 [DOI] [PubMed] [Google Scholar]
- 5. Kwok E, Otto SC, Khuu P, Carpenter AP, Codding SJ, Reardon PN, et al. The Dysferlin C2A Domain Binds PI(4,5)P2 and Penetrates Membranes. Journal of Molecular Biology. 2023;435(17):168193. doi: 10.1016/j.jmb.2023.168193 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Williams JK, Ngo JM, Lehman IM, Schekman R. Annexin A6 mediates calcium-dependent exosome secretion during plasma membrane repair. eLife. 2023;12:e86556. doi: 10.7554/eLife.86556 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Serezani CH, Ballinger MN, Aronoff DM, Peters-Golden M. Cyclic AMP. American Journal of Respiratory Cell and Molecular Biology. 2008;39(2):127–132. doi: 10.1165/rcmb.2008-0091TR [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Sassone-Corsi P. The Cyclic AMP Pathway. Cold Spring Harbor Perspectives in Biology. 2012;4(12). doi: 10.1101/cshperspect.a011148 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Zhang H, Kong Q, Wang J, Jiang Y, Hua H. Complex roles of cAMP–PKA–CREB signaling in cancer. Experimental Hematology & Oncology. 2020;9(1):32. doi: 10.1186/s40164-020-00191-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Massimi M, Cardarelli S, Galli F, Giardi MF, Ragusa F, Panera N, et al. Increase of Intracellular Cyclic AMP by PDE4 Inhibitors Affects HepG2 Cell Cycle Progression and Survival. Journal of Cellular Biochemistry. 2017;118(6):1401–1411. doi: 10.1002/jcb.25798 [DOI] [PubMed] [Google Scholar]
- 11. Chen TC, Hinton DR, Zidovetzki R, Hofman FM. Up-regulation of the cAMP/PKA pathway inhibits proliferation, induces differentiation, and leads to apoptosis in malignant gliomas. Laboratory investigation; a journal of technical methods and pathology. 1998;78 2:165–74. [PubMed] [Google Scholar]
- 12. Parekh AB. Decoding cytosolic Ca2+ oscillations. Trends in Biochemical Sciences. 2011;36(2):78–87. doi: 10.1016/j.tibs.2010.07.013 [DOI] [PubMed] [Google Scholar]
- 13. Berridge MJ, Bootman MD, Lipp P. Calcium—a life and death signal. Nature. 1998;395(6703):645–648. doi: 10.1038/27094 [DOI] [PubMed] [Google Scholar]
- 14. Smedler E, Uhlén P. Frequency decoding of calcium oscillations. Biochimica et Biophysica Acta (BBA)—General Subjects. 2014;1840(3):964–969. doi: 10.1016/j.bbagen.2013.11.015 [DOI] [PubMed] [Google Scholar]
- 15. Dolmetsch RE, Xu K, Lewis RS. Calcium oscillations increase the efficiency and specificity of gene expression. Nature. 1998;392(6679):933–936. doi: 10.1038/31960 [DOI] [PubMed] [Google Scholar]
- 16. Zaccolo M, Zerio A, Lobo MJ. Subcellular Organization of the cAMP Signaling Pathway. Pharmacological Reviews. 2021;73(1):278–309. doi: 10.1124/pharmrev.120.000086 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Calebiro D, Maiellaro I. cAMP signaling microdomains and their observation by optical methods. Frontiers in cellular neuroscience. 2014;8:350. doi: 10.3389/fncel.2014.00350 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Petersen OH. Calcium signal compartmentalization. Biological Research. 2002;35:177–182. doi: 10.4067/S0716-97602002000200008 [DOI] [PubMed] [Google Scholar]
- 19. Laude AJ, Simpson AWM. Compartmentalized signalling: Ca2+ compartments, microdomains and the many facets of Ca2+ signalling. The FEBS Journal. 2009;276(7):1800–1816. doi: 10.1111/j.1742-4658.2009.06927.x [DOI] [PubMed] [Google Scholar]
- 20. Dolmetsch R. Excitation-Transcription Coupling: Signaling by Ion Channels to the Nucleus. Science’s STKE. 2003;2003(166):pe4–pe4. [DOI] [PubMed] [Google Scholar]
- 21. Rizzuto R, Pozzan T. Microdomains of Intracellular Ca2+: Molecular Determinants and Functional Consequences. Physiological Reviews. 2006;86(1):369–408. doi: 10.1152/physrev.00004.2005 [DOI] [PubMed] [Google Scholar]
- 22. Giorgi C, De Stefani D, Bononi A, Rizzuto R, Pinton P. Structural and functional link between the mitochondrial network and the endoplasmic reticulum. The International Journal of Biochemistry and Cell Biology. 2009;41(10):1817–1827. doi: 10.1016/j.biocel.2009.04.010 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Santulli G, Pagano G, Sardu C, Xie W, Reiken S, D’Ascia SL, et al. Calcium release channel RyR2 regulates insulin release and glucose homeostasis. The Journal of Clinical Investigation. 2015;125(5):1968–1978. doi: 10.1172/JCI79273 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Kammerer S, Burns-Hamuro LL, Ma Y, Hamon SC, Cànaves JM, Shi MM, et al. Amino acid variant in the kinase binding domain of dual-specific A kinase-anchoring protein 2: A disease susceptibility polymorphism. Proceedings of the National Academy of Sciences. 2003;100(7):4066–4071. doi: 10.1073/pnas.2628028100 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Nikolaev VO, Moshkov A, Lyon AR, Miragoli M, Novak P, Paur H, et al. β2-Adrenergic Receptor Redistribution in Heart Failure Changes cAMP Compartmentation. Science. 2010;327(5973):1653–1657. doi: 10.1126/science.1185988 [DOI] [PubMed] [Google Scholar]
- 26. Wirtenberger M, Schmutzhard J, Hemminki K, Meindl A, Sutter C, Schmutzler RK, et al. The functional genetic variant Ile 646 Val located in the kinase binding domain of the A-kinase anchoring protein 10 is associated with familial breast cancer. Carcinogenesis. 2007;28(2):423–426. doi: 10.1093/carcin/bgl164 [DOI] [PubMed] [Google Scholar]
- 27. Millar JK, Pickard BS, Mackie S, James R, Christie S, Buchanan SR, et al. DISC1 and PDE4B Are Interacting Genetic Factors in Schizophrenia That Regulate cAMP Signaling. Science. 2005;310(5751):1187–1191. doi: 10.1126/science.1112915 [DOI] [PubMed] [Google Scholar]
- 28. Pasti L, Volterra A, Pozzan T, Carmignoto G. Intracellular Calcium Oscillations in Astrocytes: A Highly Plastic, Bidirectional Form of Communication between Neurons and Astrocytes In Situ. Journal of Neuroscience. 1997;17(20):7817–7830. doi: 10.1523/JNEUROSCI.17-20-07817.1997 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Eckel-Mahan KL, Phan T, Han S, Wang H, Chan GCK, Scheiner ZS, et al. Circadian oscillation of hippocampal MAPK activity and cAMP: implications for memory persistence. Nature Neuroscience. 2008;11(9):1074–1082. doi: 10.1038/nn.2174 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Addis RC, Ifkovits JL, Pinto F, Kellam LD, Esteso P, Rentschler S, et al. Optimization of direct fibroblast reprogramming to cardiomyocytes using calcium activity as a functional measure of success. Journal of Molecular and Cellular Cardiology. 2013;60:97–106. doi: 10.1016/j.yjmcc.2013.04.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Dyachok O, Isakov Y, Sågetorp J, Tengholm A. Oscillations of cyclic AMP in hormone-stimulated insulin-secreting β-cells. Nature. 2006;439(7074):349–352. doi: 10.1038/nature04410 [DOI] [PubMed] [Google Scholar]
- 32. Tenner B, Getz M, Ross B, Ohadi D, Bohrer CH, Greenwald E, et al. Spatially compartmentalized phase regulation of a Ca2+-cAMP-PKA oscillatory circuit. eLife. 2020;9:e55013. doi: 10.7554/eLife.55013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Siso-Nadal F, Fox JJ, Laporte SA, Hébert TE, Swain PS. Cross-Talk between Signaling Pathways Can Generate Robust Oscillations in Calcium and cAMP. PLOS ONE. 2009;4(10):1–10. doi: 10.1371/journal.pone.0007189 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Dupont G, Combettes L, Bird GS, Putney JW. Calcium Oscillations. Cold Spring Harbor Perspectives in Biology. 2011;3(3). doi: 10.1101/cshperspect.a004226 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Sneyd J, Han JM, Wang L, Chen J, Yang X, Tanimura A, et al. On the dynamical structure of calcium oscillations. Proceedings of the National Academy of Sciences. 2017;114(7):1456–1461. doi: 10.1073/pnas.1614613114 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Peercy BE, Sherman AS, Bertram R. Modeling of Glucose-Induced cAMP Oscillations in Pancreatic β Cells: cAMP Rocks when Metabolism Rolls. Biophysical Journal. 2015;109(2):439–449. doi: 10.1016/j.bpj.2015.06.024 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Qiao L, Zhang ZB, Zhao W, Wei P, Zhang L. Network design principle for robust oscillatory behaviors with respect to biological noise. eLife. 2022;11:e76188. doi: 10.7554/eLife.76188 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Ohadi D, Schmitt DL, Calabrese B, Halpain S, Zhang J, Rangamani P. Computational Modeling Reveals Frequency Modulation of calcium-cAMP/PKA Pathway in Dendritic Spines. Biophysical Journal. 2019;117(10):1963–1980. doi: 10.1016/j.bpj.2019.10.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Ohadi D, Rangamani P. Geometric Control of Frequency Modulation of cAMP Oscillations due to Calcium in Dendritic Spines. Biophysical Journal. 2019;117(10):1981–1994. doi: 10.1016/j.bpj.2019.10.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Tarasov AI, Semplici F, Ravier MA, Bellomo EA, Pullen TJ, Gilon P, et al. The Mitochondrial Ca2+ Uniporter MCU Is Essential for Glucose-Induced ATP Increases in Pancreatic Cells. PLOS ONE. 2012;7(7):1–12. doi: 10.1371/journal.pone.0039722 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Ashcroft FM, Rorsman P. Diabetes Mellitus and the β Cell: The Last Ten Years. Cell. 2012;148(6):1160–1171. doi: 10.1016/j.cell.2012.02.010 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Zhou Z, Misler S. Amperometric Detection of Quantal Secretion from Patch-clamped Rat Pancreatic β-Cells. Journal of Biological Chemistry. 1996;271(1):270–277. doi: 10.1074/jbc.271.1.270 [DOI] [PubMed] [Google Scholar]
- 43. Tomas A, Jones B, Leech C. New Insights into β-Cell GLP-1 Receptor and cAMP Signaling. Journal of Molecular Biology. 2020;432(5):1347–1366. doi: 10.1016/j.jmb.2019.08.009 [DOI] [PubMed] [Google Scholar]
- 44. Dou H, Wang C, Wu X, Yao L, Zhang X, Teng S, et al. Calcium influx activates adenylyl cyclase 8 for sustained insulin secretion in rat pancreatic beta cells. Diabetologia. 2015;58(2):324–333. doi: 10.1007/s00125-014-3437-z [DOI] [PubMed] [Google Scholar]
- 45. Tengholm A, Gylfe E. cAMP signalling in insulin and glucagon secretion. Diabetes, Obesity and Metabolism. 2017;19(S1):42–53. doi: 10.1111/dom.12993 [DOI] [PubMed] [Google Scholar]
- 46. Ni Q, Ganesan A, Aye-Han NN, Gao X, Allen MD, Levchenko A, et al. Signaling diversity of PKA achieved via a Ca2+-cAMP-PKA oscillatory circuit. Nature chemical biology. 2011;7(1):34–40. doi: 10.1038/nchembio.478 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Gold MG, Stengel F, Nygren PJ, Weisbrod CR, Bruce JE, Robinson CV, et al. Architecture and dynamics of an A-kinase anchoring protein 79 (AKAP79) signaling complex. Proceedings of the National Academy of Sciences. 2011;108(16):6426–6431. doi: 10.1073/pnas.1014400108 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48. Haselwandter CA, Calamai M, Kardar M, Triller A, da Silveira RA. Formation and stability of synaptic receptor domains. Physical Review Letters. 2011;106(23):238104. doi: 10.1103/PhysRevLett.106.238104 [DOI] [PubMed] [Google Scholar]
- 49. Lai M, Brun D, Edelstein SJ, Le Novère N. Modulation of Calmodulin Lobes by Different Targets: An Allosteric Model with Hemiconcerted Conformational Transitions. PLOS Computational Biology. 2015;11(1):1–28. doi: 10.1371/journal.pcbi.1004063 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50. Bagur R, Hajnóczky G. Intracellular Ca2+ Sensing: Its Role in Calcium Homeostasis and Signaling. Molecular Cell. 2017;66(6):780–788. doi: 10.1016/j.molcel.2017.05.028 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51. Fridlyand LE, Tamarina N, Philipson LH. Modeling of Ca2+ flux in pancreatic β-cells: role of the plasma membrane and intracellular stores. American Journal of Physiology-Endocrinology and Metabolism. 2003;285(1):E138–E154. doi: 10.1152/ajpendo.00194.2002 [DOI] [PubMed] [Google Scholar]
- 52. Lefkimmiatis K, Zaccolo M. cAMP signaling in subcellular compartments. Pharmacology and Therapeutics. 2014;143(3):295–304. doi: 10.1016/j.pharmthera.2014.03.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53. Landa LR Jr, Harbeck M, Kaihara K, Chepurny O, Kitiphongspattana K, Graf O, et al. Interplay of Ca 2+ and cAMP Signaling in the Insulin-secreting MIN6 beta Cell Line. Journal of Biological Chemistry. 2005;280(35):31294–31302. doi: 10.1074/jbc.M505657200 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54. Alon U. An introduction to systems biology: design principles of biological circuits. CRC press; 2019. [Google Scholar]
- 55. Zhang C, Tsoi R, Wu F, You L. Processing Oscillatory Signals by Incoherent Feedforward Loops. PLOS Computational Biology. 2016;12(9):1–16. doi: 10.1371/journal.pcbi.1005101 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56. Kaplan S, Bren A, Dekel E, Alon U. The incoherent feed‐forward loop can generate non‐monotonic input functions for genes. Molecular Systems Biology. 2008;4(1):203. doi: 10.1038/msb.2008.43 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57. Goentoro L, Shoval O, Kirschner MW, Alon U. The Incoherent Feedforward Loop Can Provide Fold-Change Detection in Gene Regulation. Molecular Cell. 2009;36(5):894–899. doi: 10.1016/j.molcel.2009.11.018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58. Hong J, Brandt N, Abdul-Rahman F, Yang A, Hughes T, Gresham D. An incoherent feedforward loop facilitates adaptive tuning of gene expression. eLife. 2018;7:e32323. doi: 10.7554/eLife.32323 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59. Qiao L, Zhao W, Tang C, Nie Q, Zhang L. Network topologies that can achieve dual function of adaptation and noise attenuation. Cell systems. 2019;9(3):271–285. doi: 10.1016/j.cels.2019.08.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60. Ma W, Trusina A, El-Samad H, Lim WA, Tang C. Defining Network Topologies that Can Achieve Biochemical Adaptation. Cell. 2009;138(4):760–773. doi: 10.1016/j.cell.2009.06.013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61. Rangamani P, Levy MG, Khan S, Oster G. Paradoxical signaling regulates structural plasticity in dendritic spines. Proceedings of the National Academy of Sciences. 2016;113(36):E5298–E5307. doi: 10.1073/pnas.1610391113 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 62. Rangamani P, Lipshtat A, Azeloglu EU, Calizo RC, Hu M, Ghassemi S, et al. Decoding information in cell shape. Cell. 2013;154(6):1356–1369. doi: 10.1016/j.cell.2013.08.026 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 63. Schmick M, Bastiaens PIH. The Interdependence of Membrane Shape and Cellular Signal Processing. Cell. 2014;156(6):1132–1138. doi: 10.1016/j.cell.2014.02.007 [DOI] [PubMed] [Google Scholar]
- 64. Haupt A, Minc N. How cells sense their own shape—mechanisms to probe cell geometry and their implications in cellular organization and function. Journal of Cell Science. 2018;131(6):jcs214015. doi: 10.1242/jcs.214015 [DOI] [PubMed] [Google Scholar]
- 65. Mo GC, Ross B, Hertel F, Manna P, Yang X, Greenwald E, et al. Genetically encoded biosensors for visualizing live-cell biochemical activity at super-resolution. Nature methods. 2017;14(4):427–434. doi: 10.1038/nmeth.4221 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 66. Zhang J, Carver CM, Choveau FS, Shapiro MS. Clustering and functional coupling of diverse ion channels and signaling proteins revealed by super-resolution STORM microscopy in neurons. Neuron. 2016;92(2):461–478. doi: 10.1016/j.neuron.2016.09.014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 67. Palaia I, Šarić A. Controlling cluster size in 2D phase-separating binary mixtures with specific interactions. The Journal of Chemical Physics. 2022;156(19):194902. doi: 10.1063/5.0087769 [DOI] [PubMed] [Google Scholar]
- 68. Varga MJ, Fu Y, Loggia S, Yogurtcu ON, Johnson ME. NERDSS: A Nonequilibrium Simulator for Multibody Self-Assembly at the Cellular Scale. Biophysical Journal. 2020;118(12):3026–3040. doi: 10.1016/j.bpj.2020.05.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 69. Pablo M, Ramirez SA, Elston TC. Particle-based simulations of polarity establishment reveal stochastic promotion of Turing pattern formation. PLOS Computational Biology. 2018;14(3):1–25. doi: 10.1371/journal.pcbi.1006016 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 70. Ullrich A, Böhme MA, Schöneberg J, Depner H, Sigrist SJ, Noé F. Dynamical Organization of Syntaxin-1A at the Presynaptic Active Zone. PLOS Computational Biology. 2015;11(9):1–22. doi: 10.1371/journal.pcbi.1004407 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 71. Takeda K, Shao D, Adler M, Charest PG, Loomis WF, Levine H, et al. Incoherent Feedforward Control Governs Adaptation of Activated Ras in a Eukaryotic Chemotaxis Pathway. Science Signaling. 2012;5(205):ra2–ra2. doi: 10.1126/scisignal.2002413 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 72. Lewis RS. Calcium oscillations in T-cells: mechanisms and consequences for gene expression. Biochemical Society Transactions. 2003;31(5):925–929. doi: 10.1042/bst0310925 [DOI] [PubMed] [Google Scholar]
- 73. Aton SJ, Herzog ED. Come Together, Right. Now: Synchronization of Rhythms in a Mammalian Circadian Clock. Neuron. 2005;48(4):531–534. doi: 10.1016/j.neuron.2005.11.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 74. Hanslmayr S, Staudigl T, Fellner MC. Oscillatory power decreases and long-term memory: the information via desynchronization hypothesis. Frontiers in Human Neuroscience. 2012;6. doi: 10.3389/fnhum.2012.00074 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 75. Hanslmayr S, Staresina BP, Bowman H. Oscillations and Episodic Memory: Addressing the Synchronization/Desynchronization Conundrum. Trends in Neurosciences. 2016;39(1):16–25. doi: 10.1016/j.tins.2015.11.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 76. Gregor T, Fujimoto K, Masaki N, Sawai S. The Onset of Collective Behavior in Social Amoebae. Science. 2010;328(5981):1021–1025. doi: 10.1126/science.1183415 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 77. Danino T, Mondragón-Palomino O, Tsimring L, Hasty J. A synchronized quorum of genetic clocks. Nature. 2010;463(7279):326–330. doi: 10.1038/nature08753 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 78. Calizo RC, Bell MK, Ron A, Hu M, Bhattacharya S, Wong NJ, et al. Cell shape regulates subcellular organelle location to control early Ca2+ signal dynamics in vascular smooth muscle cells. Scientific Reports. 2020;10(1):17866. doi: 10.1038/s41598-020-74700-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 79. Bell MK, Lee CT, Rangamani P. Spatiotemporal modelling reveals geometric dependence of AMPAR dynamics on dendritic spine morphology. The Journal of Physiology. 2023;601(15):3329–3350. doi: 10.1113/JP283407 [DOI] [PubMed] [Google Scholar]
- 80. Leung A, Ohadi D, Pekkurnaz G, Rangamani P. Systems modeling predicts that mitochondria ER contact sites regulate the postsynaptic energy landscape. npj Systems Biology and Applications. 2021;7(1):26. doi: 10.1038/s41540-021-00185-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 81. Alfonzo-Méndez MA, Sochacki KA, Strub MP, Taraska JW. Dual clathrin and integrin signaling systems regulate growth factor receptor activation. Nature Communications. 2022;13(1):905. doi: 10.1038/s41467-022-28373-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 82. Baschieri F, Dayot S, Elkhatib N, Ly N, Capmany A, Schauer K, et al. Frustrated endocytosis controls contractility-independent mechanotransduction at clathrin-coated structures. Nature Communications. 2018;9(1):3825. doi: 10.1038/s41467-018-06367-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 83. Nair D, Hosy E, Petersen JD, Constals A, Giannone G, Choquet D, et al. Super-Resolution Imaging Reveals That AMPA Receptors Inside Synapses Are Dynamically Organized in Nanodomains Regulated by PSD95. Journal of Neuroscience. 2013;33(32):13204–13224. doi: 10.1523/JNEUROSCI.2381-12.2013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 84. Zhang JZ, Lu TW, Stolerman LM, Tenner B, Yang JR, Zhang JF, et al. Phase Separation of a PKA Regulatory Subunit Controls cAMP Compartmentation and Oncogenic Signaling. Cell. 2020;182(6):1531–1544.e15. doi: 10.1016/j.cell.2020.07.043 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 85. Shibasaki T, Takahashi H, Miki T, Sunaga Y, Matsumura K, Yamanaka M, et al. Essential role of Epac2/Rap1 signaling in regulation of insulin granule dynamics by cAMP. Proceedings of the National Academy of Sciences. 2007;104(49):19333–19338. doi: 10.1073/pnas.0707054104 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 86. Renström E, Eliasson L, Rorsman P. Protein kinase A-dependent and-independent stimulation of exocytosis by cAMP in mouse pancreatic B-cells. The Journal of physiology. 1997;502(Pt 1):105. doi: 10.1111/j.1469-7793.1997.105bl.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 87. Eliasson L, Ma X, Renström E, Barg S, Berggren PO, Galvanovskis J, et al. SUR1 Regulates PKA-independent cAMP-induced Granule Priming in Mouse Pancreatic B-cells. Journal of General Physiology. 2003;121(3):181–197. doi: 10.1085/jgp.20028707 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 88. Song WJ, Seshadri M, Ashraf U, Mdluli T, Mondal P, Keil M, et al. Snapin Mediates Incretin Action and Augments Glucose-Dependent Insulin Secretion. Cell Metabolism. 2011;13(3):308–319. doi: 10.1016/j.cmet.2011.02.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 89. Vikman J, Svensson H, Huang YC, Kang Y, Andersson SA, Gaisano HY, et al. Truncation of SNAP-25 reduces the stimulatory action of cAMP on rapid exocytosis in insulin-secreting cells. American Journal of Physiology-Endocrinology and Metabolism. 2009;297(2):E452–E461. doi: 10.1152/ajpendo.90585.2008 [DOI] [PubMed] [Google Scholar]
- 90. Wan QF, Dong Y, Yang H, Lou X, Ding J, Xu T. Protein Kinase Activation Increases Insulin Secretion by Sensitizing the Secretory Machinery to Ca2+. Journal of General Physiology. 2004;124(6):653–662. doi: 10.1085/jgp.200409082 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 91. Tengholm A, Gylfe E. cAMP signalling in insulin and glucagon secretion. Diabetes, Obesity and Metabolism. 2017;19(S1):42–53. doi: 10.1111/dom.12993 [DOI] [PubMed] [Google Scholar]
- 92. Sanderson JL, Dell’Acqua ML. AKAP Signaling Complexes in Regulation of Excitatory Synaptic Plasticity. The Neuroscientist. 2011;17(3):321–336. doi: 10.1177/1073858410384740 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 93. Nijholt IM, Dolga AM, Ostroveanu A, Luiten PGM, Schmidt M, Eisel ULM. Neuronal AKAP150 coordinates PKA and Epac-mediated PKB/Akt phosphorylation. Cellular Signalling. 2008;20(10):1715–1724. doi: 10.1016/j.cellsig.2008.05.001 [DOI] [PubMed] [Google Scholar]
- 94. Mehta S, Zhang J. Liquid–liquid phase separation drives cellular function and dysfunction in cancer. Nature Reviews Cancer. 2022;22(4):239–252. doi: 10.1038/s41568-022-00444-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 95. Boija A, Klein IA, Young RA. Biomolecular Condensates and Cancer. Cancer Cell. 2021;39(2):174–192. doi: 10.1016/j.ccell.2020.12.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 96. Jiang S, Fagman JB, Chen C, Alberti S, Liu B. Protein phase separation and its role in tumorigenesis. eLife. 2020;9:e60264. doi: 10.7554/eLife.60264 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 97. Cai D, Liu Z, Lippincott-Schwartz J. Biomolecular Condensates and Their Links to Cancer Progression. Trends in Biochemical Sciences. 2021;46(7):535–549. doi: 10.1016/j.tibs.2021.01.002 [DOI] [PubMed] [Google Scholar]
- 98. Mahmud I, Liao D. DAXX in cancer: phenomena, processes, mechanisms and regulation. Nucleic Acids Research. 2019;47(15):7734–7752. doi: 10.1093/nar/gkz634 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 99. Bouchard JJ, Otero JH, Scott DC, Szulc E, Martin EW, Sabri N, et al. Cancer Mutations of the Tumor Suppressor SPOP Disrupt the Formation of Active, Phase-Separated Compartments. Molecular Cell. 2018;72(1):19–36.e8. doi: 10.1016/j.molcel.2018.08.027 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 100. Grabocka E, Bar-Sagi D. Mutant KRAS Enhances Tumor Cell Fitness by Upregulating Stress Granules. Cell. 2016;167(7):1803–1813.e12. doi: 10.1016/j.cell.2016.11.035 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 101. Somasekharan SP, El-Naggar A, Leprivier G, Cheng H, Hajee S, Grunewald TGP, et al. YB-1 regulates stress granule formation and tumor progression by translationally activating G3BP1. Journal of Cell Biology. 2015;208(7):913–929. doi: 10.1083/jcb.201411047 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 102. Lazzerini-Denchi E, Sfeir A. Stop pulling my strings—what telomeres taught us about the DNA damage response. Nature Reviews Molecular Cell Biology. 2016;17(6):364–378. doi: 10.1038/nrm.2016.43 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 103. Dilley RL, Greenberg RA. ALTernative Telomere Maintenance and Cancer. Trends in Cancer. 2015;1(2):145–156. doi: 10.1016/j.trecan.2015.07.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 104. Kondo S, Miura T. Reaction-Diffusion Model as a Framework for Understanding Biological Pattern Formation. Science. 2010;329(5999):1616–1620. doi: 10.1126/science.1179047 [DOI] [PubMed] [Google Scholar]
- 105. Sick S, Reinker S, Timmer J, Schlake T. WNT and DKK Determine Hair Follicle Spacing through a Reaction-Diffusion Mechanism. Science;314(5804):1447–1450. doi: 10.1126/science.1130088 [DOI] [PubMed] [Google Scholar]
- 106. Raspopovic J, Marcon L, Russo L, Sharpe J. Digit patterning is controlled by a Bmp-Sox9-Wnt Turing network modulated by morphogen gradients. Science. 2014;345(6196):566–570. doi: 10.1126/science.1252960 [DOI] [PubMed] [Google Scholar]
- 107. Miura T. Models of lung branching morphogenesis. The Journal of Biochemistry. 2015;157(3):121–127. doi: 10.1093/jb/mvu087 [DOI] [PubMed] [Google Scholar]
- 108. Wilmott ZM, Goriely A, Raff JW. A simple Turing reaction–diffusion model explains how PLK4 breaks symmetry during centriole duplication and assembly. PLOS Biology. 2023;21(11):1–23. doi: 10.1371/journal.pbio.3002391 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 109. Karig D, Martini KM, Lu T, DeLateur NA, Goldenfeld N, Weiss R. Stochastic Turing patterns in a synthetic bacterial population. Proceedings of the National Academy of Sciences. 2018;115(26):6572–6577. doi: 10.1073/pnas.1720770115 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 110. Ranganathan S, Shakhnovich EI. Dynamic metastable long-living droplets formed by sticker-spacer proteins. eLife. 2020;9:e56159. doi: 10.7554/eLife.56159 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 111. Wang H, Wingreen NS, Mukhopadhyay R. Self-Organized Periodicity of Protein Clusters in Growing Bacteria. Phys Rev Lett. 2008;101:218101. doi: 10.1103/PhysRevLett.101.218101 [DOI] [PubMed] [Google Scholar]
- 112. Liu QX, Doelman A, Rottschäfer V, de Jager M, Herman PMJ, Rietkerk M, et al. Phase separation explains a new class of self-organized spatial patterns in ecological systems. Proceedings of the National Academy of Sciences. 2013;110(29):11905–11910. doi: 10.1073/pnas.1222339110 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 113. Gasior K, Forest MG, Gladfelter AS, Newby JM. Modeling the Mechanisms by Which Coexisting Biomolecular RNA–Protein Condensates Form. Bulletin of Mathematical Biology. 2020;82(12):153. doi: 10.1007/s11538-020-00823-x [DOI] [PubMed] [Google Scholar]
- 114. Mahapatra A, Saintillan D, Rangamani P. Curvature-driven feedback on aggregation–diffusion of proteins in lipid bilayers. Soft Matter. 2021;17:8373–8386. doi: 10.1039/d1sm00502b [DOI] [PMC free article] [PubMed] [Google Scholar]
- 115. Mesa MH, Garcia GC, Hoerndli FJ, McCabe KJ, Rangamani P. Spine apparatus modulates Ca2+ in spines through spatial localization of sources and sinks. bioRxiv. 2023. doi: 10.1101/2023.09.22.558941 [DOI] [Google Scholar]
- 116.Qiao L, Khalilimeybodi A, Linden-Santangeli NJ, Rangamani P. The evolution of systems biology and systems medicine: From mechanistic models to uncertainty quantification. arXiv. 2024;.
- 117. Linden-Santangeli N, Zhang J, Kramer B, Rangamani P. Increasing certainty in systems biology models using Bayesian multimodel inference. bioRxiv. 2024. doi: 10.1101/2024.06.16.599231 [DOI] [Google Scholar]
- 118. Linden NJ, Kramer B, Rangamani P. Bayesian parameter estimation for dynamical models in systems biology. PLOS Computational Biology. 2022;18(10):1–48. doi: 10.1371/journal.pcbi.1010651 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 119. Bell MK, Holst MV, Lee CT, Rangamani P. Dendritic spine morphology regulates calcium-dependent synaptic weight change. Journal of General Physiology. 2022;154(8):e202112980. doi: 10.1085/jgp.202112980 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 120. Klipp E, Liebermeister W. Mathematical modeling of intracellular signaling pathways. BMC Neuroscience. 2006;7(1):S10. doi: 10.1186/1471-2202-7-S1-S10 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 121. Means AR, Dedman JR. Calmodulin—an intracellular calcium receptor. Nature. 1980;285(5760):73–77. doi: 10.1038/285073a0 [DOI] [PubMed] [Google Scholar]
- 122.Finkston B. Mean Shift Clustering; 2024. https://www.mathworks.com/matlabcentral/fileexchange/10161-mean-shift-clustering.
- 123. Comaniciu D, Meer P. Mean shift: a robust approach toward feature space analysis. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2002;24(5):603–619. doi: 10.1109/34.1000236 [DOI] [Google Scholar]