Microdomain organization and frequency-dependence of CREB-dependent transcriptional signaling in heart cells

Abstract

Voltage-gated Ca_v1.2 calcium channels couple membrane depolarization to cAMP response-element-binding protein (CREB)-dependent transcriptional activation. To investigate the spatial and temporal organization of CREB-dependent transcriptional nuclear microdomains, we combined perforated patch-clamp technique and FRET microscopy for monitoring CREB and CREB-binding protein interaction in the nuclei of live cells. The experimental approach to the quantitative assessment of CREB-dependent transcriptional signaling evoked by cAMP- and Ca_v1.2-dependent mechanisms was devised in COS1 cells expressing recombinant Ca_v1.2 calcium channels. Using continuous 2-dimensional wavelet transform and time series analyses, we found that nuclear CREB-dependent transcriptional signaling is organized differentially in spatially and temporally separated microdomains of 4 distinct types. In rat neonatal cardiomyocytes, CREB-dependent transcription is mediated by the cAMP-initiated CaMKII-sensitive and Ca_v1.2-initiated CaMKII-insensitive mechanisms. The latter microdomains show a tendency to exhibit periodic behavior correlated with spontaneous contraction of myocytes suggestive of frequency-dependent CREB-dependent transcriptional regulation in the heart.—Kobrinsky, E., Duong, S.Q., Sheydina, A., Soldatov, N. M. Microdomain organization and frequency-dependence of CREB-dependent transcriptional signaling in heart cells.

In a variety of cells, gene expression is under the control of cell excitation. Excitation-transcription coupling is a functional link between signaling events in the plasma membrane and transcriptional response in the nucleus (Fig. 1). The predominant role in initiating this important chain of signaling cascades is attributed to the voltage-gated Ca_v1.2 calcium channels (1–3). Several consecutive events are found to be important for transmitting excitation information from the plasma membrane to the nucleus. Depolarization of the plasma membrane initiates opening of Ca_v1.2 channels. Ca²⁺ ions passing through the channel inside the cell bind to calmodulin (CaM) tethered to the C-terminus of the Ca²⁺ channel pore-forming α_1C subunit. This binding induces a conformational rearrangement of the α_1C C-terminus and a concomitant inactivation of the Ca²⁺ conductance. In neurons, these events are followed by the activation of Ca²⁺/CaM-dependent protein kinase II (CaMKII). It is interesting that the above initial steps induce multiple signaling pathways (3, 4) leading to activation of cAMP response element binding protein (CREB)-dependent transcription (CDT).

Figure 1.

Ca_v1.2- and cAMP-dependent cell signaling pathways mediating CDT. Activation of CREB by phosphorylation of Ser-133 induces interaction with CBP, binding to CRE, and activation of transcription of the target gene.

CREB mediates both Ca²⁺- and cAMP-dependent transcription of certain genes. Both cAMP and Ca²⁺-dependent mechanisms require phosphorylation of serine-133 in CREB but involve different kinases (5, 6). In neurons, protein kinase A mediates cAMP-dependent phosphorylation, and Ca²⁺/CaM-dependent kinases (CaMKs), ribosomal S6 kinases, and mitogen- and stress-activated protein kinases likely mediate Ca²⁺-dependent phosphorylation (7). In the heart, much recent attention was focused on NF-AT transcriptional regulation (8, 9), but CREB-dependent excitation-transcription coupling has not been studied. However, CDT regulation is essential for a variety of clinically important cardiac issues, including cardiac remodeling and ischemic preconditioning (10–12). In this study, we filled this gap by using, for the first time, a quantitative approach to transcriptional signaling.

Most of the existing approaches to studying the cell information processing underestimate the necessity for quantitative measurements of spatial and temporal organization of signaling pathways. Recently, we have developed a system biology tool based on wavelet analysis for quantitative global and local analysis of signaling events in the plasma membrane and inside the cell (13, 14). Based on wavelet transform analysis of FRET microscopy images, this method can discriminate signaling microdomains (submicrometer discrete regions inside the cell that have a distinct function) in noisy local and global environments, which is currently beyond the reach of existing techniques. We have combined this methodological innovation with FRET microscopy imaging in real time and in live cells and applied them to the investigation of CDT signaling involved in excitation-transcription coupling (14). To correlate the time dependency of excitability of the plasma membrane with the coupled response of transcriptional signaling events in the nucleus, we have extended our analysis with the wavelet transform coherence method (15). This approach allows us to characterize different types of discrete subnuclear CDT signaling microdomains in model COS1 cells expressing recombinant calcium channels and in rat neonatal cardiomyocytes. Interestingly, we discovered transient cardiac transcriptional domains where CREB and CREB-binding protein (CBP) interactions are modulated by the Ca_v1.2 calcium channel activity in a coherent manner and correlate with spontaneous contraction of cardiomyocytes. These results reveal the existence of a built-in, robust CDT signaling pathway for excitation-transcription coupling in cardiac myocytes and suggest frequency-dependent CDT regulation in the heart.

MATERIALS AND METHODS

Cell culture, transfection and chemicals

COS1 cells were grown on poly-d-lysine-coated coverslips (MatTek, Ashland, MA, USA) in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. Before transfection, COS1 cells were kept for 18 h under serum-deprived conditions. Rat neonatal cardiac myocytes were isolated from the hearts of 1- to 3-d-old Fisher 334 rats using collagenase and pancreatin treatment according to Zheng et al. (16). Briefly, the ventricles were dissected, separated from attached vessels and atria, cut into small pieces, and digested with a solution containing collagenase type II (0.2 mg/ml; Worthington Biochemical, Lakewood, NJ, USA) and pancreatin (0.6 mg/ml; Sigma-Aldrich, St. Louis, MO, USA) in PBS. Neonatal ventricular myocytes, separated by preplating for 1 h, were plated at a density of 6 × 10⁴ cells/cm² on gelatin-covered poly-d-lysine-coated coverslips (MatTek), kept in culture conditions according to Kobrinsky et al. (17) for 24 h in a medium composed of 4 parts (by volume) of DME containing 1.0 g/L glucose and 1 part of medium 199 (Earle's salts; Gibco Life Technologies, Inc., Gaithersburg, MD, USA) supplemented with 10% preselected horse serum, 5% heat-inactivated fetal bovine serum, penicillin (100 U/ml), and streptomycin (100 mg/ml). Following this step, cells were transferred into a serum-free medium supplemented with 1x ITS (Insulin-Transferrin-Selenium-A, Gibco Life Technologies, Inc.). In the initial series of experiments, cultured myocytes were maintained in the presence of bromodeoxyuridine (10⁻⁴ M) to inhibit cell replication of a small percentage of fibroblasts that was present. Immunocytostaining with antibodies to cardiac α-actin or sarcomeric myosin heavy chains identified 90–95% positive cells. The cell cultures were maintained in 5% CO₂ at 37°C. COS1 cells and cardiac myocytes were transfected with cDNA constructs by Effectene kit (Qiagen, Valencia, CA, USA) as described previously (2). To monitor CREB transcriptional activity, we used the KIX domain of CBP and the kinase-inducible domain (KID) involved in interaction with KIX. Enhanced cyan fluorescent protein (ECFP) and enhanced yellow fluorescent protein (EYFP) were fused to the C termini of KIX (KIXCN) and KID (YCREBDQ1) probes, respectively. The latter construct contains the basic leucine zipper domain that targets CREB to chromatin. The pore-forming Ca_v1.2 α_1C subunit (z34815) was coexpressed with auxiliary subunits β_1b1 (M92302) and α₂δ-1 as described earlier (18). In FRET experiments, all constructs were expressed at 1:1 M ratio under conditions optimized by the Effectene kit manufacturer for a total amount of 0.2 μg of DNA/35 mm Petri dish. Relative expression levels of the Ca²⁺ channel were approximately the same, as judged from the mean amplitude of Ca²⁺ currents. To decrease background CREB phosphorylation, cells were kept in serum-free medium for 18 h before measurements. The water-soluble CaMKII inhibitor KN93 was purchased from Calbiochem (EMD4Biosciences, Gibbstown, NJ, USA) di-8-ANEPPS was from Molecular Probes (Invitrogen, Carlsbad, CA, USA), cell-permeable analog of cAMP and all other chemicals were from Sigma-Aldrich.

Electrophysiology

Ion currents were recorded using the Axopatch 200B amplifier (Axon Instruments, Foster City, CA, USA) at 20–22°C by the perforated whole-cell patch-clamp method 48–72 h after transfection. To monitor transcriptional activation under voltage-clamp conditions, we used the perforated patch-clamp technique (19). Amphotericin B (75 μM) was added to the pipette solution containing (in mM) 130 CsCl, 20 TEA-Cl, 10 EGTA, and 5 HEPES, adjusted to pH 7.2 with CsOH. External solution contained (in mM) 150 NaCl, 5.4 KCl, 1 MgCl₂, 5 HEPES, 2 CaCl₂, and 5.5 glucose, pH 7.4. Voltage protocols were generated, and data were digitized, recorded, and analyzed, using pClamp 8.1 software (Axon). Depolarization to +20 mV from the holding potential of −90 mV was applied to elicit maximum Ca²⁺ currents.

Imaging

Images were recorded using a 14-bit Hamamatsu C9100-12 digital camera (Hamamatsu, Bridgewater, NJ, USA) mounted on a Nikon TE2000 epifluorescent microscope (Nikon, Tokyo, Japan) equipped with a 60 × 1.45 numerical aperture (n.a.) oil objective and multiple filter sets (Chroma Technology, Rockingham, VT, USA). Corrected FRET intensity was calculated using MATLAB from data acquired using the 3 filter sets (CFP, YFP, and FRET), as described previously (18). Excitation light was delivered by a 175-W xenon lamp. Excitation filter sets were changed by a high-speed filter wheel system (Lambda 10-2; Sutter Instrument, Novato, CA, USA). The Dual-View system (Optical Insights, Santa Fe, NM, USA) was used for the simultaneous acquisition of 2 fluorescence images (donor and FRET). Images were collected and analyzed using C-Imaging (Compix, Cranberry Township, PA, USA) and MATLAB 7.0.4 (The Mathworks, Natick, MA, USA). Two-color FRET was quantified with 3 filter sets: for the YFP cube, excitation filter 500/20 nm, dichroic beam splitter 515 nm, emission filter 535/30 nm; for the CFP cube, excitation filter 436/20 nm, dichroic beam splitter 505 nm, emission filter 480/40 nm; for the FRET cube (CFP/YFP), excitation filter 436/20 nm, dichroic beam splitter 505 nm, emission filter 540/30 nm. To avoid photobleaching artifacts, we used minimal possible exposure (from 5 ms in fast acquisition time series to 27 ms in standard experiments) and monitored fluorescence intensities of donor and acceptor. To control images for possible positional changes, we used a standard approach based on comparison of images and region of interest (ROI) with Pearson correlation coefficient. Image correlation is very sensitive to the shift of object, it is independent of the gain and offset and it is widely used in colocalization studies (20, 21). We performed image correlation of all images within the selected time frame to align them. Because all FRET constructs contained nuclear localization signals, the most accurate correlations were obtained if the ROI was the edge of the nuclei in the images selected for the alignment. The experiments where aligned images shifted >3 pixels (pixel size ∼250 nm) in either direction were excluded from analysis. Contraction of cardiac myocytes was measured optically (22) as sensitive to L-type calcium channel blockers and fully reversible nonrandom positional changes of the edge (23). To record muscle cell contraction, the plasma membrane of cardiomyocytes was labeled by lipophilic fluorescent dye di-8-ANEPPS (5 μM; Molecular Probes). Fluorescence from this plasma membrane marker (excitation at 436 nm and measurement at 535 nm) was spatially well separated from CDT signaling events and was collected simultaneously with FRET recording. The voltage sensing by di-8 ANEPPS, causing a few percent change in relative fluorescence intensity, did not influence optical measurement of cell contractions.

Mathematical tools of analysis

One-dimensional continuous wavelet transform (1D-CWT)

The 1D-CWT is given by the following equation:

$$W\left(a,b\right)=\frac{1}{\sqrt{a}}{\displaystyle {\int }_{-\infty }^{\infty }h\left(x\right)\psi *\left(\frac{x-b}{a}\right)dx}$$

where a represents scale (inversely related to frequency), b is the translation parameter, h(x) is the signal, and ψ is a wavelet basis function (note that * indicates complex conjugation). For 1D-CWT, we used a Morlet wavelet given by the following equation:

$$\psi \left(\eta \right)={\pi }^{-1/4}{e}^{i\omega 0\eta }{e}^{-{\eta }^2/2}$$

where η is a nondimensional time parameter, and ω₀ is the nondimensional frequency, taken to be 6. The cone of influence (COI) is defined as the area in which wavelet power dropped to e⁻² of its value at the edge of the time series. To exclude edge effects in the signal, regions within the COI were omitted from the analysis. Torrence and Compo (24) have developed methods to define statistical significance of the 1D CWT based on the background power spectrum of the signal. Our data are well modeled by a first-order autoregressive (AR1) process. The Fourier power spectrum (P_k) of an AR1 process is given by the following equation:

$$P_k=\frac{1-{\alpha }^2}{{\left|1-a{e}^{-2i\pi k}\right|}^2}$$

where α is the estimated lag-1 autocorrelation (estimated from the observed time series) and k is the Fourier frequency index (15). Torrence and Compo (24) have used Monte Carlo methods to show that the probability of the wavelet power of a process with given power spectrum P_k being >p is

$$D\left(\frac{{\left|W\left(a,b\right)\right|}^2}{{\sigma }_x^2}<p\right)=\frac{1}{2}{P}_k{\chi }_v^2\left(p\right)$$

where σ_x² is the variance within the signal, and χ_ν² is a χ² distribution where ν is equal to 1 degree of freedom for real and 2 degrees of freedom for complex wavelets.

Wavelet transform coherence (WTC)

Torrence and Webster (25) define wavelet coherence of two time series W^X(a,b) and W^Y(a,b) as follows:

$$R_n^2\left(s\right)=\frac{{\left|S\left(a^{-1}\left(W^X\left(a,b\right)W^Y\left(a,b\right)*\right)\right)\right|}^2}{S\left(a^{-1}{\left|W^X\left(a,b\right)\right|}^2\right)\times S\left(a^{-1}{\left|W^Y\left(a,b\right)\right|}^2\right)}$$

where S is a smoothing operator.

In accordance with Grinsted et al. (15), we have applied Monte Carlo methods to ascribe statistical significance levels to coherence calculations. We used 10 scales/octave and 1000 surrogate dataset pairs in our significance level estimate, as suggested by Grinsted et al. (15). MATLAB codes were downloaded from Dr. S. Jevrejeva's website (26).

2D-CWT

The 2D-CWT was implemented using the YAW toolbox (27). Briefly, the 2D-CWT is readily extended to 2 dimensions by the following equation:

$$W\left(a,b,\theta \right)=\frac{1}{a}({\displaystyle {\int }_{-\infty }^{\infty }h\left(x\right)\psi *\left(\frac{r-\theta \left(x-b\right)}{a}\right)}{d}^{2}x$$

where θ is the wavelet angle rotation (r − θ is the rotation operator).

For our wavelet basis function, we used the 2D Mexican hat function, given as follows:

$$\psi \left(x\right)=\left(2-{\left|x\right|}^2\right)\times \exp \left(-\frac{1}{2}{\left|x\right|}^2\right)$$

Data analysis

Corrected FRET values were calculated on a pixel-by-pixel basis as described before. To calculate corrected FRET in time-lapse images, we recorded images from the YFP cube right before continuous FRET imaging and used it for YFP bleedthrough correction for each image in the sequence. Each image from FRET and CFP cubes in the sequence was spatially corrected to the image obtained from YFP cube by finding the maximum Pearson's linear correlation coefficient between signals.

Microdomain identification and verification was done as described before (14). Normalized difference (ND) images with values between 0 and 1 were computed from an image at time t W(a,b,θ)_t and control image W(a,b,θ)_C as follows:

$$\text{ND}=\frac{\left[W{\left(a,b,\theta \right)}_C-W{\left(a,b,\theta \right)}_t\right]}{\max \left[W{\left(a,b,\theta \right)}_C-W{\left(a,b,\theta \right)}_t\right]}$$

The potential microdomains were defined as groups of ≥2 pixels above a normalized value of 0.25. These potential microdomains were verified by comparing the pixels within the potential microdomain across all time points using 1-way ANOVA with Tukey-Kramer multiple comparisons test. Those time points showing statistical significance from the control in response to drug application were classified accordingly.

Normalized-difference images were calculated for FRET images of cardiomyocytes taken in time series just as described above, using as the reference image the average of 5 successive FRET images during a noncontracting period within the sequence. Potential microdomains were defined as groups of ≥2 pixels above a normalized value of contraction. Average pixel intensity within each potential domain was calculated for each FRET image recorded within the sequence, and the resultant time series of FRET signal within microdomain vs. time was analyzed with ANOVA to confirm statistically significant changes in FRET signal vs. control time point.

We measured contraction data from the cardiomyocyte by recording movement of the cell surface membrane delineated with lipophilic fluorescent dye di-8-ANEPPS (5 μM; ref. 28). The FRET time series and the contraction data were standardized by subtracting the mean of the entire series from each individual value in the series and dividing it by the sd of the series. The time series were then smoothed using a 10-point moving-average smoothing operation to attenuate noise artifact. We filtered potential microdomains for only those that showed a statistically significant wavelet power within the expected scales (corresponding to a period of 1/3 and 2 s) during periods of cell contraction in the wavelet transform. We then computed the cross wavelet coherence between the contraction time series and the FRET data. Domains showing phase coherence of contraction and FRET intensity signals were selected.

RESULTS

Spatiotemporal organization of L-type Ca²⁺ channel-activated CDT signaling in COS1 cells

We have recently thoroughly documented that COS1 cells lack endogenous calcium channels (29). However, functional calcium channels can be assembled in COS1 cells from transiently expressed α_1C, β, and α₂δ subunits, providing thereby a carefully controlled experimental cell system in which to study CDT signaling. Indeed, we showed previously (2) that periodic stimulation of Ca²⁺ channels expressed in COS1 cells induced CDT signaling. We confirmed the importance of the Ca²⁺-CaM “IQ” binding sites on the α_1C cytoplasmic C-tail (1), and we discovered a crucial role of the apo-CaM “pre-IQ” site in a cascade utilized by the cell for signal transduction from the plasma membrane Ca_v1.2 calcium channel to CDT in the nucleus. We demonstrated the importance of voltage-dependent conformational rearrangements of α_1C for excitation-transcription coupling. Because two major second-messenger systems, cyclic AMP and Ca²⁺ (7), are upstream of CDT, we decided to perform a detailed analysis of cAMP- and Ca²⁺-dependent coupling between the Ca_v1.2 activity and CDT signaling in COS1 cells and rat neonatal cardiac myocytes.

The ability of CREB to activate target gene transcription is coordinated by phosphorylation of Ser-133 of the KID, which confers sensitivity to elevation of the cAMP concentration due to the presence of 2 functional motifs, a conserved consensus phosphorylation site for PKA and other kinases, and the constitutive Gln-rich Q2 domain, which is responsible for basal CREB activity. Ser-133 phosphorylation triggers the KID-mediated recruitment of the pleiotropic transcriptional coactivator CBP, via their KIX domains (30, 31).

To visualize CREB and CBP complex formation with spatial and temporal resolution, we employed a FRET-based detection system that uses the fluorescently labeled CDT model substrates KID and KIX (32, 33). To load nuclei, EYFP-KID and ECFP-KIX probes used for transfection were supplemented with nuclear localization sequence (32). FRET images were obtained continuously during application of cell-permeable analog of cAMP (0.25 mM). FRET was calculated from the whole nuclear region (2). No changes in FRET signal were observed during the 5-min control period preceding the application of cAMP (Fig. 2A). However, a progressive increase of FRET signal due to KID/KIX interaction was observed consistently in response to cAMP application similar to that induced by the application of adenylyl cyclase activator forskolin (32). We observed statistically significant changes in FRET between KID and KIX after 8 min of cAMP treatment. After 15 min of cAMP application, net FRET signal increased more than 50% in comparison to the control level. In control experiments, we did not observe an increase in FRET signal during cAMP application in COS1 cells expressing a phosphorylation-insensitive Ser-133Ala mutant of KID domain, corroborating previously published results (32). Taken together, these data show that the KID/KIX FRET system is adequate for the study of CDT in COS1 cells.

Figure 2.

Potentiation of CDT signaling by cAMP and calcium current in COS1 cells expressing recombinant Ca_v1.2 calcium channel (whole-cell perforated patch conditions). A) Effect of cAMP on CDT at −90 mV. Application of cAMP (0.25 mM) induced interaction between CREB and CBP (solid circles) in COS1 cells (n=6), while Ser-133Ala mutant of CREB did not show significant cAMP-induced response (open circles; n=3). B) Effect of calcium current stimulation on CDT. Calcium channel current activity was stimulated by periodical depolarization pulses applied from the holding potential of −90 to +20 mV for 600 ms every 10 s. Solid circles, increase in FRET between CREB and CBP constructs. Inhibition of L-type calcium channels by (+)PN200–110 (5 μM) inhibited potentiation of CDT signaling (triangles, n=3). The mutant form of CREB construct was not sensitive to the Ca²⁺ current stimulation (open circles). *P < 0.05.

We next showed that calcium channel currents activate CDT signaling by stimulating the interaction between CREB and CBP, Ser-133 phosphorylation being critical for this interaction. The recombinant Ca_v1.2 calcium channel was expressed in COS1 cells. Consistent with the previously established dependence of CDT on the activity of Ca_v1.2 calcium channels (1), periodic stimulation of the L-type Ca²⁺ current caused progressive increase of FRET signal (Fig. 2B). At the end of a 3-min window of the Ca²⁺ current stimulation, FRET signal increased 3.22 ± 0.70 times (n=8) compared with the control level before the stimulation. After termination of the calcium channel stimulation, CDT signal returned to the prestimulation basal level within 1–2 min. In control experiments with the mutant form of the KID, the FRET signal did not change significantly and was 0.94 ± 0.08 compared with the control level before the stimulation (n=5). It has been shown in neurons (1) that dihydropyridine (DHP) blocker of L-type calcium channels nimodipine completely inhibited CREB-dependent gene expression. Respectively, potentiation of CDT was abolished in COS1 cells (0.97±0.03, n=3) by inhibition of calcium channels with 5 μM (+)PN200–110 (Fig. 2B). These results indicate that the number of interacting CREB and CBP molecules significantly increased during periodic stimulation of Ca_v1.2 calcium channels. Phosphorylation of Ser-133 is important not only for cAMP-dependent CREB activation, but also for the Ca²⁺ channel-dependent activation of CREB signaling pathway.

Microdomain organization of transcriptional signaling in COS1 cells

All transcriptional signaling data in previous works were obtained by integrating methods that rely either on biochemical studies of a large amount of cells or on imaging experiments where intensity of a net signal was obtained by averaging the intensity of all pixels in the nucleus (3). However, it is clear that in most cases signals are distributed nonhomogenously inside the nucleus, and standard integral methods do not resolve the dynamics of spatial and temporal organization of cell signaling, characteristics that provide new insight into development and distribution of distinct local CDT events. We recently developed a new approach to identify nuclear CDT signaling domains by using 2D-CWT analysis of FRET imaging data (14).

We conducted the perforated patch-clamp experiment in COS1 cells expressing Ca_v1.2 calcium channels and KID and KIX probes of CDT, and performed both the global and local 2D-CWT microdomain analyses of the results. Figure 3A shows the averaged FRET data obtained from the whole nucleus. We did not observe significant changes in FRET signal under control patch-clamp conditions at −90 mV maintaining all calcium channels closed. However, we registered an increase in net FRET intensity during the periodic stimulation of L-type Ca²⁺ channels by depolarizing pulses applied from −90 to +20 mV for 600 ms every 10 s (Fig. 3A), indicating significant activation of CDT signaling by stimulation of the Ca²⁺ channel activity that reached steady state approximately by the third minute. Then we held the membrane potential constant at −90 mV to keep channels silent and applied cAMP (0.25 mM). cAMP induced a continuous additional increase in FRET that was observed for as long as 10 min in the absence of the Ca²⁺ current. These data show that effects of the Ca²⁺ channel stimulation and cAMP on CDT signaling might be partially independent. Repeated activation of the Ca²⁺ current in the presence of cAMP evoked only a small additional FRET, probably because CREB and CBP interaction reached saturation after prolonged stimulation of CDT signaling pathways.

Figure 3.

Global and local microdomain analysis of the effects of the Ca²⁺ channel activation and cAMP on CDT signaling in COS1 cells. A) Increase in FRET intensity of CDT signaling from the whole nucleus in an exemplary COS1 cell expressing recombinant Ca_v1.2 calcium channel. FRET was recorded during the indicated episodes of periodic stimulation of the Ca²⁺ current in whole-cell perforated patch clamp before and after cAMP application at the holding potential of −90 mV maintaining all Ca²⁺ channels closed. Arrow indicates moment of establishing whole-cell (perforated) patch mode at −90 mV. Ca bars indicate episodes of calcium channel stimulation by depolarizing pulses applied from −90 to +20 mV for 600 ms every 10 s. B) FRET signals recorded within the nucleus of the same representative COS1 cell during selected time points. Axes show pixel numbers. Identified microdomains are circled in red. Color bar represents FRET value normalized to the maximum. C) Nucleus and the microdomains outlined to illustrate activation of microdomains during Ca²⁺ channels stimulation and cAMP application. Open circles show stable CDT signaling microdomains activated only by cAMP application. Solid circles show stable CDT signaling microdomains that depend on both the Ca²⁺ channel activity and cAMP (red), Ca²⁺ channel-dependent stable CDT signaling microdomains that were not activated by cAMP (yellow) and transient microdomains (green). D–G) Four categories of the identified CDT signaling microdomains, including stable CDT signaling microdomains activated by Ca²⁺ and cAMP (D), microdomains activated only by Ca²⁺ (E) or cAMP (F), and transient microdomains activated by Ca²⁺ and cAMP (G). Values are averaged FRET of individual microdomains normalized to maximal microdomain FRET. Time is indicated inside bars. C, control measured before stimulation. At bottom of graphs (in relation to specific stimuli indicated by arrows) are typical appearances of the respective microdomains at maximal development of sustained transcriptional activation (D–F) or at initial and late recording time points (G). Recombinant α_1C, β_2d, and α₂δ-1 subunits were coexpressed in COS1 cells. Nuclear CREB signaling microdomains were identified using 2D Mexican hat wavelet. *P<0.05 for Ca²⁺ channel stimulation; **P<0.05 for cAMP stimulation; n=5.

By applying 2D-CWT analysis to every tested cell, we characterized (Fig. 3B) individual microdomains responsive to both cAMP and Ca²⁺ channel stimulation (Fig. 3C, solid circles), as well as those responsive to only cAMP or Ca²⁺ (Fig. 3C, open circles). We use the term microdomain because of the submicrometer size of these CDT signaling domains (usually these domains are a few pixels large, pixel size ∼250 nm) and to avoid confusion with nanodomains (scale of a single molecule, from few to 100 nm). Augmentation of FRET signals within the individual domains and an increase in the number of new domains during stimulation underlies the continuous increases in average FRET intensity obtained from the whole nuclei (Fig. 3A). Figure 3D–G summarizes results of CWT analysis that revealed 4 categories of CDT signaling domains, each exhibiting different Ca²⁺- and cAMP-dependence. The majority of the detected nuclear CDT microdomains showed a long-lasting increase in FRET intensity during stimulation of CREB signaling. These microdomains of stable CDT activity can be divided into 3 groups. The major group (65% of the total number of domains, n=5) is Ca²⁺- and cAMP-activated microdomains showing continuous increase in FRET signal during Ca²⁺ current stimulation and during application of cAMP (Fig. 3D). Some nuclear microdomains responded with continuous increase in FRET signal only to Ca²⁺ current stimulation (15%, Fig. 3E), or to cAMP application alone (5%, Fig. 3F). Intriguingly, we also observed transient CDT microdomains (15%, Fig. 3G). Intensity of FRET signal in these microdomains was characterized by a significant initial increase in response to Ca²⁺ or cAMP stimulation that, however, returned to the control level during the stimulation. We did not observe such transient domains sensitive to one of the stimuli. This finding suggests that CREB and CBP interactions in these nuclear regions are activated transiently, possibly by mechanisms other than those responsible for the stably activated CDT domains. All these results were obtained in COS1 cells transiently expressing the recombinant cardiac Ca²⁺ channel. Having described microdomain organization in the recombinant COS1 system, we sought to characterize the CDT signaling microdomain regulation in a native cell, the neonatal cardiac myocyte, where CREB serves as an important transcriptional activator (34).

Microdomain organization of L-type Ca²⁺ channel-activated CDT signaling in neonatal cardiac myocytes

CREB signaling in the heart regulates the transcription of genes encoding proteins that are important for cardiac myocyte function, such as contractility, energy production, and growth (34). We decided to use spontaneously contracting neonatal cardiac myocytes to study Ca²⁺ channel- and cAMP-dependence of CDT regulation. These cells contract due to spontaneous activation of L-type Ca²⁺ channels (33, 35). Spontaneously contracting neonatal cardiac myocytes do not need external stimulation and thus can be used as a model to study excitation-transcription coupling in heart cells. Another advantage of neonatal cardiac muscle cells over adult cells is their tolerance to harsh conditions of the transfection procedure.

Continuous application of cAMP (0.25 mM) produced a significant increase in average intranuclear FRET signal. In the example presented in Fig. 4A, FRET signal significantly increased to 1.35 ± 0.06 of control level after 10 min of cAMP application, and after 15 min of application was 1.37 ± 0.15 of control level (P<0.05; n=4). Blockade of the L-type Ca²⁺ current by DHP calcium channel blocker (+)PN200–110 (5 μM) inhibited spontaneous contraction of cardiomyocytes and completely abolished the cAMP-induced FRET increase. In fact, FRET signal reduced to 0.62 ± 0.03 compared with control level (P<0.05, n=4). This result indicates that cAMP-induced CDT signaling in cardiomyocytes depends in part on the activity of endogenous Ca²⁺ channels.

Figure 4.

Contribution of Ca_v1.2 Ca²⁺ channel activity, cAMP, and CaMKII in CDT signaling in rat neonatal cardiac myocytes. A) Effect of (+)PN200–110 (5 μM) on cAMP-induced CDT signaling. B) Effect of CaMKII inhibitor KN93 on cAMP-induced CDT signaling. C) Identified microdomains of CDT signaling in the nucleus of a representative neonatal rat cardiac myocyte before (control) and after 10 and 15 min of cAMP (0.25 mM) application followed by inhibition of Ca²⁺ channels by (+)PN200–110. Top panels: images of corrected normalized FRET between KIX and KID probes. Intranuclear microdomains of CDT signaling (red circles) were identified by 2D-CWT analysis. Color bar represents FRET value normalized to the maximum. Bottom panels: nucleus and microdomains outlined to illustrate different types of Ca²⁺ channel- and cAMP-dependent CDT signaling microdomains. Solid red circles represent stable CDT signaling microdomains that depend on both Ca²⁺ channel and cAMP. These microdomains can be activated by application of cAMP and are inhibited in the presence of 5 μM (+)PN200–110. Open circles represent stable CDT signaling microdomains activated by application of cAMP and retaining activity in the presence of (+)PN200–110. Solid yellow circles represent Ca²⁺ channel-dependent stable CDT signaling microdomains that do not require cAMP for their activity and are sensitive to (+)PN200–110. Solid green circles represent transient microdomains. D) Relative presentation of CDT signaling microdomains in rat neonatal heart cells (n=4). *P < 0.05.

We next investigated critical components of the signaling pathway that mediate CDT activation in rat neonatal cardiomyocytes. It was already shown in neurons that pretreatment with CaMKII blocker KN93 (2 μM) inhibits cAMP-induced interaction between CREB and CBP proteins (3). Figure 4B shows that FRET signal did not change significantly in the presence of KN93 (2 μM). FRET was not significantly changed in response to the subsequent cAMP application (1.02±0.8 after 10 min and 0.95±0.1 after 15 min, n=8). Thus, cAMP-induced activation of CDT signaling in cardiac myocytes depends on CaMKII phosphorylation. However, the application of the L-type Ca²⁺ channel blocker decreased FRET signal to 0.67 ± 0.1 of control value (P<0.05; n=8), showing that ∼1/3 of the basal CDT signaling is due to Ca²⁺ channel activity but does not depend on CaMKII phosphorylation. These results suggest existence of divergent cAMP- and Ca²⁺ channel-initiated regulation pathways of CDT signaling in heart cells.

We next characterized the respective nuclear CDT signaling microdomains in spontaneously contracting cardiac myocytes using 2D-CWT analysis. We routinely observed stable and transient cAMP- and Ca²⁺-activated CDT microdomains (Fig. 4C) that could be discriminated by application of cell-permeable cAMP and calcium channel blocker (+)PN200–110, respectively. During spontaneous contraction, openings of Ca²⁺channels provide the signal for Ca²⁺-dependent CDT. Indeed, no CDT nuclear microdomains and cell contraction were seen in the presence of (+)PN200–110. In spontaneously contracting cardiac myocytes 30.9 ± 6.0% of domains showed stable activity when activated by both Ca²⁺ and cAMP (Fig. 4D); a similar number of stable CDT domains were activated only by Ca²⁺ (30.0±6.0%). However, the number of stable CDT domains activated only by cAMP was significantly smaller (11.7±4.0%; P<0.05; n=4). The proportion of transient domains activated by Ca²⁺ and cAMP in neonatal heart cells was slightly higher than in COS1 cells (24.2±4.0%). Thus, 4 major types of CDT signaling microdomains underlie the architecture of CDT signaling in COS1 cells expressing recombinant Ca_v1.2 calcium channels as well as in native cardiac myocytes.

The presence of transient CDT signaling subnuclear domains in cardiac myocytes raises the possibility of a functional link between the transcriptional signaling and L-type Ca²⁺ channel activation during spontaneous muscle contractions. To examine this link, the plasma membrane of cardiac cells expressing KID and KIX constructs was labeled with di-8-ANEPPS, enabling simultaneous recording of muscle cell contraction and CDT activity. Time-series and WTC analysis was used to find the microdomains that exhibited a phase relationship with contractile (and thus excitation) activity. In the experiment presented in Fig. 5, we compared FRET and contraction signals at different acquisition rates of 27 and 200 ms (Fig. 5A). 1D-CWT analysis (13) was used to expand time series into a time and frequency space with Morlet transform. This wavelet function is a good choice for feature extraction purposes because it is reasonably localized in time and frequency (15). WTC can be used for local correlation between two CWTs. Microdomains showing locally phase-locked behavior were identified by WTC with a significance level determined using Monte Carlo methods. The results of this analysis are presented in Fig. 5B–D. Comparison of contraction activity (Fig. 5A, blue line) and average FRET intensity of a single CDT transient microdomain (Fig. 5A, red line) in time suggest that the contraction and transcription activity are linked on both small (Fig. 5A, left panel) and large time scales (Fig. 5A, right panel, box). Indeed, our analysis of these time series with WTC confirmed that they exhibit a statistically significant phase-locked relationship (Fig. 5D). Significant areas on wavelet transform and coherence scalograms are outlined with black lines in Fig. 5D. These are the areas where FRET and contractility signals are phase locked with high significance level, P < 0.05, in a broad range of oscillation frequencies from a period of 0.2 to 3 s. Similar results were obtained in all recorded cardiac cells (n=8). Stable domains of a comparable size during contraction did not change significantly (e.g., see black traces in Fig. 5A, representing changes of FRET signal in a stable microdomain during spontaneous contraction). The fast acquisition (Fig. 5A, left panel) shows that the timing of CDT transient activity peaks coincides with contractions, although the two series are some what out of phase, probably because the time scales of the development of these processes are different. The amplitude of CDT oscillations (related to the number of interacting KID and KIX probes) was relatively constant. Most of the significant regions of phase-locked behavior are close to 0° phase angle (in phase), suggesting that transcriptional activation is localized tightly in time with contraction. We observed periodic changes in interaction between CREB and CBP in segregated nuclear microdomains. This transient and periodic interaction of transcriptional factors coincided with periodic heart cell contractile activity (excitability). These results suggest the existence of excitation-transcription coupling built-in signaling pathways regulating CDT signaling microdomains.

Figure 5.

Coherence time-series analysis of spontaneous contractility and CDT signaling activity in neonatal rat cardiac myocytes. A) Simultaneous recording of contraction (blue) and FRET between the KID and KIX probes within an identified transient (red) and stable (black) nuclear CDT microdomain (≈0.3 μm² in area) with fast (left panel) and slow (right panel) acquisitions. B) CWT scalogram of spontaneous contraction. C) CWT scalogram of FRET signal between ECFP-KID and EYFP-KIX. Color bars represent scale of CWT. D) Wavelet coherence between spontaneous contractile activity and FRET signal. Regions outlined in black represent statistically significant areas based on Monte Carlo analysis (P<0.05). Arrows represent phase difference between contractile and transcriptional signaling data. Arrows pointing to the right are in phase, arrows pointing to the left are in antiphase. COI where edge effects might distort the picture is shown as a lighter shade. Color bar represents wavelet coherence.

DISCUSSION

Here, we characterized the spatiotemporal organization of CDT signaling using wavelet transform analysis of KID-KIX FRET assay that reproduces with fidelity CDT events. To investigate microdomain organization of transcriptional signaling, we developed quantitative tools for spatial and temporal analysis in FRET microscopy, where 2D-CWT identifies microdomains, while 1D-CWT and WTC find the areas of local correlation with phase-locked behavior between time series in time frequency space. This approach provided a framework for analysis of complex spatial and temporal dynamics of cell signaling. Our data indicate that CDT signaling is organized in stable and transient nuclear microdomains activated by calcium channel current, cAMP, or both of these stimuli. A similar structure of CDT signaling was observed in COS1 cells expressing recombinant Ca_v1.2 calcium channels and in neonatal rat cardiomyocytes, although there are probably differences in nuclear distribution and other parameters that our FRET assay has not detected. Thus, characteristic spatiotemporal organization of CDT signaling is inherent in wide range of different cells and can be rendered even in cells initially deprived of calcium channels, such as COS1. In spontaneously contracting rat neonatal cardiomyocytes, we observed frequency-dependent CDT signaling that matched contractions and were sensitive to DHP calcium channel blocker.

Gene expression is organized in spatially heterogeneous subnuclear domains (36). The mammalian cell nucleus is structurally and functionally complex and contains morphologically distinct chromatin domains and numerous protein subcompartments constrained within a defined nuclear volume. It is generally accepted that the spatial organization of these nuclear compartments is inherently connected to their role in gene expression and cell regulation (37). These microdomains might be related to the recently described transcriptional factories (38, 39), discrete subnuclear sites where multiple active RNA polymerase and signaling molecules are concentrated and anchored to a nuclear structure. This domain organization of transcription is an important part of dynamic 3D nuclear organization. However, individual behavior of spatially separated domains is overlooked by the traditional integral approach of averaging FRET intensity among the pixels from the whole nuclei.

The cellular correlate of CDT signaling domain is molecular complex of signaling molecules involved in CREB-dependent initiation of transcription, specifically CREB and CBP. CREB and CBP interaction reflecting CDT activation is downstream of signal sensing and serves as the first major step for execution of gene expression. In our study, CREB and CBP interaction was activated by periodic stimulation of L-type Ca²⁺ channels and/or application of cAMP. In COS1 cells expressing recombinant Ca_v1.2 calcium channels, we observed 4 different groups of spatially separated nuclear microdomains of CDT signaling. Most of the domains responded to both cAMP and Ca²⁺ channel stimulation, showing stable (65%) and transient (15%) activity. These domains constitute changes in discrete FRET signals that essentially mirror the average nuclear FRET changes. Only small subsets of stable CDT microdomains showed differential responses to either cAMP or to the Ca²⁺ current stimulation.

In cardiac myocytes, we detected microdomains showing periodic CREB-dependent activity (Fig. 5). Periodic oscillations in gene expression were observed in other native cells (40) and reconstituted in engineered gene circuits (41). In the case of NF-κB-controlled transcription, inhibition of the oscillatory nuclear shuttling blocked transcription from a NF-κB–dependent promoter (42). Oscillation frequency of NF-κB activation determines which downstream genes are expressed (43, 44). Another example is the control of yeast gene expression through nuclear translocation frequency by the Ca²⁺-dependent transcription factor Crz1 (45). The frequency, but not the amplitude, of short nuclear bursts of Crz1 increased with increased extracellular Ca²⁺ concentration. Taken together, these observations suggest that a similar periodic activity might be a property of other important signaling pathways. Frequency of Ca²⁺oscillations might also modulate a transcriptional signaling. For example, NFAT-dependent transcriptional activation in lymphocytes is sensitive to Ca²⁺oscillation frequencies (46, 47). Periodic Ca²⁺ influx through L-type Ca²⁺ channels in cardiac cells provided the basis for frequency modulation of transcriptional signaling. It has been shown that frequency of Ca²⁺ oscillations during spontaneous contractions of neonatal cardiomyocyte can be decoded by Ca²⁺-dependent phosphatase calcineurin (48). The decoding of alterations in the frequency even of rapid Ca²⁺ oscillations in spontaneously beating cardiomyocytes resulted in the translocation of NFAT into the nucleus and the downstream activation of the cardiac hypertrophy program. Pacing-induced changes in CREB were also observed in heart cells (49). CREB is an important factor in cardiac memory (50). Induction of short-term and long-term cardiac memory is attenuated by L-type Ca²⁺ channel blockers (51), suggesting the functional link between the two most important players of excitation-transcription coupling in the heart, L-type Ca²⁺ channel and CREB. We showed that CDT signaling is localized in spatially segregated nuclear microdomains. CREB and CBP interactions in some of these domains are transient in nature and might play the most important part of the pacing (frequency)-dependent CDT regulation. A number of earlier studies point to the transient nature of CREB signaling. Thus, transient kinetics of CREB phosphorylation has been reported to be induced by growth factors (52) and erythropoietin (53). Transient responses were associated with the stimulation of CREB activity by phosphorylation of serine-133 as well as by the activity of protein serine/threonine phosphatase (54). Phosphorylation of serine-142 and 143 of CREB can also induce the inhibition of CREB/CBP interaction as reported by Kornhauser et al. (55). Phosphatase activity as well as phosphorylation may provide a negative feedback in Ca²⁺-dependent regulation of CREB activity and CREB/CBP interaction. It is clear that the dynamics of phosphorylation/dephosphorylation of CREB regulatory sites is crucial for the appearance of transient microdomains of CREB and CBP interaction. How fast is the interaction (binding/unbinding) between the two molecules, and how fast is phosphorylation/dephosphorylation of the regulatory sites? A proximity of CREB and CBP is the prerequisite of the fast binding event. It is known that phosphorylation can be a very fast process, and phosphoryl transfer step might occur within <20 ms (for review see ref. 56). Indirect electrophysiological experiments with excised patch showed that phosphatase-mediated dephosphorylation might be very fast as well (57). Binding and unbinding of transcription factors are also characterized by a very fast kinetics (58). Overall, a time scale of few hundred milliseconds is enough to complete either phosphorylation or dephosphorylation reaction. Nonlinear organization of transcription initiation might be one of the factors of complex temporal pattern of transcription and protein expression dynamic in eukaryotic cells. A recent modeling study (59) showed that target proteins are more efficiently activated by the oscillatory signals at low levels of stimulation, irrespective of whether the information was encoded in the amplitude or in the frequency of oscillations. This study provides a theoretical support for our experimental findings of Ca²⁺-dependent CDT signaling.

It was shown recently that in neurons CaMKII integrates Ca²⁺ channel activity to CDT activation (3). CaMKII acting near the channel couples transient increase in local Ca²⁺ concentration to CDT signal transduction, reacting to the frequency of Ca²⁺ channel openings rather than to the integrated Ca²⁺ flux. Our data showed that inhibition of CaMKII by KN93 completely abolished cAMP-induced CDT signaling. About 1/3 of basal CREB/CBP interactions in heart cells depend on Ca²⁺ channel activity but are not affected by CaMKII inhibition. Therefore, in cardiac myocytes, the CREB-dependent excitation-transcription coupling is partially executed through CaMKII-independent pathways that have not been observed in neurons (3). Using wavelet analysis, we observed a more complicated behavior of the transcriptional signaling regulation. We found that different types of CDT signaling regulation are localized in spatially different microdomains, which might represent separated signaling units of transcriptional regulation.

Recent studies indicate that signal sensing and transcriptional induction by CREB are separate events, the latter of which requires cooperative interactions with other upstream activators (60). Interaction of CREB and CBP is an important step in transcription initiation, but even transcriptional initiation very often fails to complete gene expression (61). This condition is due to a different level of regulation, which occurs subsequent to initiation and involves transcript elongation and stability. It remains to be investigated whether and how the observed CDT signaling domains are related to specific cellular functions.

This study was supported by the National Institute on Aging Intramural Research Program (Z01 AG000294-08 to NMS). The authors thank Dr. Mark R. Montminy (The Salk Institute for Biological Studies, La Jolla, CA, USA) for providing KID and KIX constructs and Dr. Kenneth R. Boheler (National Institute on Aging, U.S. National Institutes of Health, Baltimore, MD, USA) for providing rat cardiac myocytes.