In Vivo FRET Imaging of Tumor Endothelial Cells Highlights a Role of Low PKA Activity in Vascular Hyperpermeability

Blood vessels in tumors are significantly different from those in normal tissues in terms of structure and function (1). For example, vascular permeability is markedly increased in tumor tissues, and this increase is associated with angiogenesis and metastasis (2–6). Accordingly, suppression of the vascular permeability results in the inhibition of tumor growth and metastasis (7).

The high vascular permeability in tumor tissues is caused by various factors, including the structural abnormality of blood vessels (2, 8), pressure or concentration gradients between compartments, and the properties of endothelial cells and/or pericytes (2, 6). Among a number of molecules that regulate blood vessels in tumor tissues, vascular endothelial growth factor (VEGF) has been a focus of intensive research (6, 9). VEGF increases vascular permeability by promoting transcytosis (10–14) and enlarging the intercellular gaps of endothelial cells (15). The inhibition of VEGF and VEGF receptors (VEGFR) results in a decrease of permeability and the normalization of tumor vasculature, which in turn results in the inhibition of tumor growth and improvement of anticancer chemotherapy (7).

The VEGF-induced increase in vascular permeability is mediated by various intracellular signaling pathways, including those of PTPs, Src, PI3K, uPA, PLC-γ, and eNOS (6). Another signaling molecule that has been shown to regulate vascular permeability is cAMP. In microvascular endothelial cells, cAMP mediates endothelial barrier function by suppressing vascular permeability (16, 17). Further studies have shown that increased cAMP triggers sequential activation of protein kinase A (PKA), a guanine nucleotide exchange factor Tiam1, and a small GTPase Rac1, resulting in an increase in barrier function (18, 19). On the other hand, cAMP can also contribute to the endothelial barrier function via cAMP-dependent guanine nucleotide exchange factor Epac/cAMP-GEF and a small GTPase Rap1 (20, 21). Notably, the contribution of these potential effectors downstream of VEGFR to the hyperpermeability in tumor tissue has not been assessed in live tissues.

Visualization of tumor tissues by two-photon excitation microscopy has opened new windows into the dynamics of angiogenesis and the involvement of hematopoietic cells in tumorigenesis (22, 23). Meanwhile, using cancer cells expressing Förster resonance energy transfer (FRET) biosensors, the effects of anticancer drugs on the target molecules have been visualized in the xenograft/homograft recipient mice, revealing considerable heterogeneity among cancer cells (24, 25). To conduct similar approaches for the host cells, transgenic mice expressing the FRET biosensors are needed. We and others have developed such transgenic mice, collectively designated FRET mice, for the visualization of activities of protein kinases and small GTPases (26, 27). Here, by in vivo FRET imaging of the intratumoral blood vessels, we show that the basal activity of tumor endothelial PKA is significantly lower than that of normal endothelial cells and that the cAMP analogue could alleviate the increased vascular permeability in the tumors.

Panc-02 mouse pancreatic cancer cells were obtained from the NIH (Bethesda, MD). B16-F10 mouse melanoma cells were obtained from the ATCC. Primary human umbilical vascular endothelial cells (HUVEC) were purchased from the Lonza Group, Ltd. These cell lines were acquired after 2012. Frozen stocks were prepared from initial stocks, and within every 2 months, a new frozen stock was used for the experiments. Colon-38 mouse colon cancer cells and 3LL mouse lung carcinoma cells were provided by Drs. Setoyama and Chiba at Kyoto University (Kyoto, Japan) and Drs. Shime and Seya at Hokkaido University (Sapporo, Japan), respectively, and periodically authenticated by morphologic inspection. Tumor or normal endothelial cells were isolated from FRET mice as described previously (28). A pCX4 retroviral vector was used to express the Keima fluorescent protein or the Epac-cAMP sensor (29, 30). Cancer cells and HUVECs were maintained in DMEM containing 10% fetal bovine serum and in EGM-2MV media (Lonza), respectively. Following reagents were used: Motesanib (Selleck Chemicals), Qtracker 655 (Thermo Fisher Scientific Inc.), N⁶,2′-O-dibutyryladenosine 3′,5′-cyclic monophosphate sodium salt (dbcAMP; Daiichi Sankyo Company, Ltd.), 8-(4-chlorophenylthio)-2′–O-methyladenosine 3′,5′-cyclic monophosphate (007) and N⁶-benzoyladenosine-3′,5′-cyclic monophosphate (6-Bnz) (Biolog Life Science Institute), Evans blue and isoflurane (Wako Pure Chemicals, Ltd.), Mouse VEGF164 (BioLegend), H89 and PD0325901 (Sigma-Aldrich Co. LLC).

We used PKAchu mice expressing the PKA FRET biosensor AKAR3EV-NES, PKA-NC mice expressing a negative control FRET biosensor AKAR3EV-NC, and Eisuke mice expressing the ERK FRET biosensor EKAREV-NES (26). FRET mice were back-crossed more than nine generations to C57BL/6N Jcl mice (Japan SLC Inc.). Cancer cells (2 × 10⁶ cells/50 μL PBS) were injected subcutaneously into the flanks of mice. The animal protocols were reviewed and approved by the Animal Care and Use Committee of Kyoto University Graduate School of Medicine (Nos. 12064, 13074, 14079, and 15064).

We used an FV1000MVE inverted microscope (Olympus Corporation) equipped with a 30 × 1.05 NA silicon-immersion objective lens (UPLSAPO 30xS; Olympus), and an InSight DeepSee Ultrafast laser (0.95 W at 900 nm, Spectra Physics). The excitation wavelength for cyan fluorescent protein (CFP) was 840 nm. For more information, refer to supplementary materials and methods. Acquired images were processed and analyzed with MetaMorph software (Molecular Devices) as described previously (26, 31).

Mice were anesthetized with 1.5% to 2% isoflurane inhalation and placed in the prone position. Skin flap was then placed on a cover-glass. As controls, subcutaneous capillaries with a diameter of less than 15 μm and subcutaneous arterioles or venules from 15 to 50 μm in diameter were also observed in each experiment. Images were acquired every 2 or 4 minutes at a scan speed of 2 μs/pixel. During imaging, reagents were administered in 100 μL PBS via the orbital plexus, if necessary: Motesanib (50 mg/kg), VEGF164 (5 μg/mouse), dbcAMP (50 mg/kg), 6-Bnz (50 mg/kg), 007 (50 mg/kg), and Qtracker 655 (5 μL/mouse). Images of the FRET/CFP ratio as an index of PKA or ERK activity were prepared as described in the supplementary Materials and Methods (32).

To visualize endothelial PKA activity under VEGF stimulation, mouse VEGF164 (400 ng in 50 μL PBS) was injected intradermally 30 minutes before observation. To examine PKA activity under the specific inhibitor treatment, H89 (5 mmol/L in 50 μL of 5% DMSO/saline) or vehicle (5% DMSO/saline) was added to the surface of Colon-38 tumors on the PKAchu mice.

Evans blue (10 mg/mL in PBS) was injected intravenously into tumor-bearing C57BL/6N Jcl mice at a concentration of 50 mg/kg with or without following reagents: dbcAMP (50 or 100 mg/kg), 007 (10, 25, or 50 mg/kg), 6-Bnz (10, 25, or 50 mg/kg), or Motesanib (50 mg/kg). Mice were subjected to PBS perfusion 2 hours after the dye injection and sacrificed. The tumor and skin were then excised, dried at 60°C for 24 hours, and weighed. The dye was extracted from the tissues by incubation with 0.5 mL N,N-dimethylformamide (Nacalai Tesque Inc.) at 56°C for 48 hours. The dye was quantified by measuring the absorbance at 620 nm. The extracted dye was normalized to nanograms dye per milligram tumor dry weight. In some experiments, 30 minutes after Evans blue injection, 400 ng mouse VEGF164 in 50 μL PBS was injected subcutaneously.

Lenti-X 293T cells (Takara Bio Inc.) were cotransfected with the pCSIIbsr vector encoding the FRET biosensors AKAR3EV-NES or EKAREV-NES, psPAX2 (Addgene), and pCMV-VSV-G-RSV-Rev by using 293fectin (Thermo Fisher Scientific Inc.). HUVECs were infected with the resulting lentivirus and cultured with EGM-2MV media on cover glass-bottomed dishes for 16 to 24 hours before imaging. CFP and YFP images of HUVECs were obtained by using an inverted microscope (LCV110; Olympus). Cells were stimulated with vehicle (PBS), VEGF164 (50–100 ng/mL), PD0325901 (10 μmol/L), and Forskolin (50 μmol/L)–IBMX (500 μmol/L) and analyzed as described previously (31).

Formalin-fixed tissue sections were incubated with primary antibodies to CD31 (ad22538; Abcam), GFP (JL-8; Takara Bio Inc.), VEGFR2 (55B11; Cell Signaling Technology), or pVEGFR2 (19A10; Cell Signaling Technology) followed by incubation with secondary antibodies (Alexa Fluor; Thermo Fisher Scientific Inc.). Anti-CD31 antibody conjugated with Alexa Fluor 594 (BioLegend), Hoechst33258, or DAPI (Sigma-Aldrich Co. LLC) was also used.

All data were expressed as means ± standard deviation (SD). Statistical analysis was performed using GraphPad Prism software (GraphPad Software Inc.). Student t test and one-way ANOVA with Tukey post hoc analysis were used to determine statistically significant differences. P values < 0.05 were considered to indicate statistical significance. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

To understand the roles played by PKA in the hyperpermeability of intratumoral blood vessels, we compared the endothelial PKA activities between normal subcutaneous blood vessels and intratumoral blood vessels. For visualization of the PKA activity in the tumor-bearing mice (TBM), we used PKAchu mice, which expressed a cytoplasmic FRET biosensor for PKA. The structure, mode of action, and stability of the FRET biosensors have been described previously (26, 33). Briefly, phosphorylation of the substrate domain of the FRET biosensor increases in the FRET signal, which can be quantified by the intensity ratio of 530 nm FRET channel over 475 nm CFP channel (Supplementary Fig. S1A–S1C). Syngeneic Colon-38 colon cancer cells that expressed the Keima fluorescent protein were subcutaneously implanted in the flanks of the PKAchu mice. After tumor cells grew to palpable sizes, tumors and subcutaneous tissues exposed by the skin-flap method were observed under an inverted two-photon excitation microscope (Fig. 1A). In the tumor tissue, the FRET biosensor-expressing host cells, such as vascular endothelial cells and inflammatory cells were clearly distinguished from the Keima-expressing Colon-38 cancer cells (Fig. 1B, left). In the subcutaneous tissue of TBM, capillaries were clearly detected among the dense collagen fibers (Fig. 1B, right).

The PKA activity of host cells is represented by FRET/CFP ratio images (Fig. 1C). The ratio range was fixed from 0.8 to 1.8 for consistency. As shown in the enlarged image, the PKA activity in the intratumoral endothelial cells was lower than the PKA activity in the subcutaneous endothelial cells (Fig. 1C). The PKA activity of neutrophils and platelets in the vessels could be conveniently used as internal controls (Fig. 1C, arrows): The FRET/CFP values of neutrophils and platelets in the intratumoral blood vessel and subcutaneous capillary were 1.37 ± 0.15 (n = 6) and 1.30 ± 0.07 (n = 6), respectively. We identified endothelial cells primarily by the location and shape of cells (Supplementary Fig. S1D). If necessary, Z-stack images and images of Qtracker 655 vascular tracer were used to distinguish endothelial cells from other cells such as pericytes (Supplementary Fig. S1E–S1G). Furthermore, we confirmed that the GFP-positive cells lining the internal surface of intratumoral blood vessels were positive for anti-CD31 antibody (Fig. 1D).

To verify the specificity of the biosensor, we performed in vivo FRET imaging with a PKA inhibitor, H89, a PKA-specific agonist, 6-Bnz, or an Epac/cAMP-GEF specific-agonist, 007 (Supplementary Fig. S2). H89 robustly reduced PKA activity in stromal cells and slightly in the tumor endothelial cells (Supplementary Fig. S2A and S2B). As expected, 6-Bnz, but not 007, activated PKA in tumor endothelial cells (Supplementary Fig. S2C).

Encouraged by the findings obtained using Colon-38 cells, we extended our approach by using other cancer cells with different morphological features (Fig. 2A). Because of the small cytoplasmic area of endothelial cells in the tangential image, the signal-to-noise ratios of FRET images were lower than those for the other cell types. Therefore, we time-lapse-imaged and quantified the PKA activity of at least 17 endothelial cells of more than three mice under each condition (Fig. 2B). The PKA activity was significantly lower in the tumor endothelial cells of all four cell lines examined—Colon-38 (colon cancer), Panc-02 (pancreatic cancer), B16-F10 (melanoma), and 3LL (lung cancer)—than in all normal endothelial cells of TBM and control mice. We did not find a significant difference in the PKA activity among the endothelial cells of arterioles, capillaries, and venules. We also failed to detect any difference between the subcutaneous endothelial cells in TBM and control mice. Notably, the densities of tumor cells, host cells, blood vessels, and collagen fibers were quite divergent among the tumors, strongly suggesting that the low PKA activity is a primary characteristic of the nascent intratumoral endothelial cells. To exclude the possibility that the difference in the FRET signal between the intratumoral and subcutaneous endothelial cells was caused by the difference in the physical properties of tumor tissues and subcutaneous tissues, we performed similar experiments with the PKA-NC mice, which expressed a negative control PKA FRET biosensor lacking the PKA phosphorylation site in the same FRET biosensor backbone. No significant differences were observed between the intratumoral and subcutaneous endothelial cells of PKA-NC mice (Fig. 2C). These data strongly argued for the low intratumoral endothelial PKA.

We extended the same approach to Eisuke mice, which are transgenic mice expressing a FRET biosensor for ERK (Fig. 3 and Supplementary Fig. S1A and S1C). The distribution of ERK activity among different endothelial cells was not significantly different from that of PKA activity shown in Fig. 2. Nevertheless, we failed to detect significant difference in ERK activity between the intratumoral and subcutaneous endothelial cells. We also observed no difference between the TBM and control mice. These observations indicated that the difference between tumor tissue and subcutaneous tissue was specific for PKA activity.

Previous studies have demonstrated that endothelial PKA is a key player for endothelial barrier function (18, 19). It is also known that vascular permeability is high in intratumoral blood vessels. Therefore, we hypothesized that the low PKA activity in the intratumoral endothelial cells is responsible for the high vascular permeability in the tumor. To validate this hypothesis, we conducted a modified Miles assay for measuring vascular permeability in the presence or absence of a cAMP analogue, dbcAMP (Fig. 4A).

Colon-38 or Panc-02 cells were injected into the flanks of C57BL/6N Jcl mice to prepare TBM. After intravenous injection of Evans blue to TBM, the dye extravasation from the tumor and the skin was quantified at 2 hour after administration (Fig. 4A). As expected, in the PBS-injected Colon-38-bearing mice, the extravasation of Evans blue dye was higher in the tumor than the skin, confirming the hyperpermeability of the tumor vessels (Fig. 4B). In the dbcAMP-injected mice, however, the extravasation of Evans blue was as low as that in the skin. Importantly, the effect of dbcAMP was not detectable in the skin, strongly suggesting that the PKA activity in the normal subcutaneous endothelium is sufficiently high to prevent extravasation of Evans blue. Similar results were obtained with the Panc-02 tumor model. The effect of administered dbcAMP on the endothelial PKA was validated by in vivo FRET imaging (Fig. 4C). dbcAMP robustly activated PKA in both tumor and normal endothelial cells (Fig. 4D). Taken together, these data indicate that tumor vascular permeability is in part dependent on the low endothelial PKA activity.

cAMP can regulate vascular permeability by PKA and Epac/cAMP-GEF. To assess the contribution of PKA and Epac/cAMP-GEF to the cAMP-induced decrease of tumor vascular permeability, we used a PKA-specific agonist, 6-Bnz, and an Epac/cAMP-GEF-specific agonist, 007. Mice were injected intravenously together with Evans blue dye and the agonists (10, 25, and 50 mg/kg) and subjected to a Miles assay (Fig. 5). Both 6-Bnz and 007 decreased the extravasation of Evans blue in a dose-dependent manner. Neither 6-Bnz nor 007 significantly affected the extravasation of Evans blue in the skin. Although we did not test the cAMP levels in the endothelial cells, these data strongly suggested that the low cAMP level in tumor endothelial cells at least partially contributed to the high vascular permeability in tumor tissues.

The cAMP level and PKA activity are regulated by numerous signaling molecules, including Gs-coupled GPCRs and phosphodiesterases (34). Therefore, rather than examining the contribution of direct regulators of cAMP, we focused on VEGFs and their cognate receptors (VEGFRs), which have been shown to play key roles in vascular permeability and angiogenesis in tumors (6). To examine the role of VEGFR signaling on PKA, we tested the effect of a VEGFR inhibitor on tumor vascular permeability and endothelial PKA activity by Miles assay and in vivo FRET imaging, respectively. Among the available VEGFR kinase inhibitors, we chose Motesanib based on its high water solubility. Mice were injected intravenously with Evans blue dye and Motesanib (50 mg/kg) and subjected to the Miles assay. Motesanib reduced the extravasation of Evans blue in the tumor, but not the skin, indicating that Motesanib inhibited tumor vascular permeability (Fig. 6A). This finding is consistent with previous studies that demonstrated an inhibition of VEGF-induced vascular permeability by Motesanib administration (35). By immunohistochemistry, we confirmed that Motesanib suppressed the VEGFR2 tyrosine kinase activity, but not the expression of VEGFR2 (Supplementary Fig. S3).

We next examined the response of endothelial PKA activity to Motesanib by in vivo FRET imaging. Motesanib robustly activated PKA in the endothelial cells of the tumors, but not in the endothelial cells of the subcutaneous tissues (Fig. 6B), indicating that VEGFR played a major role in the suppression of PKA. Notably, Motesanib had no effect on the ERK activities in tumor and normal endothelial cells (Fig. 6C). The lack of effect on ERK activity is probably due to the low basal ERK activity. In fact, when ERK activity in tumor endothelial cells was elevated by VEGF, Motesanib markedly decreased ERK activity (Supplementary Fig. S4A–S4C).

To confirm the role of PKA on the VEGF-induced increase in vascular permeability, PKA activity and vascular permeability were measured in the VEGF-administered skin. Evans blue was intravenously injected and 30 minutes later, PBS or VEGF164 was injected intradermally. VEGF increased the vascular permeability in the skin (Fig. 7A). In a parallel experiment, PBS or VEGF164 was intradermally injected into PKAchu mice, followed by in vivo FRET imaging for PKA activity (Fig. 7B). As shown in Fig. 7C, the PKA activity in the capillary endothelial cells was significantly lower in the VEGF-injected skin than in the PBS-injected skin. Taken together, these data strongly suggested that the VEGFR signaling pathway increases vascular permeability in part by reducing the PKA activity in endothelial cells.

Finally, we attempted to recapitulate the VEGF-induced decrease of PKA activity in human umbilical vascular endothelial cells (HUVEC) expressing the PKA FRET biosensor. In stark contrast to our in vivo observations, we found that PKA activity was increased rapidly in HUVECs (Fig. 7D and E). The increase in the cAMP level was also confirmed by the Epac-cAMP sensor (Supplementary Fig. S5; ref. 29). To examine if the discrepancy was caused by the in vitro experimental condition, we isolated intratumoral and subcutaneous endothelial cells from the transgenic and repeated the experiments (Supplementary Fig. S6). In contrast to HUVECs and similarly to the subcutaneous endothelial cells, PKA activity in the isolated subcutaneous endothelial cells was higher than the isolated intratumoral endothelial cells and decreased by VEGF stimulation. These results suggest that the isolated primary cells are more similar to the vascular endothelial cells in vivo than to the cultured HUVECs. This is likely due to the difference in proliferation ability between the two types of normal endothelial cells. In fact, the isolated endothelial cells ceased proliferation and became senescent several days after isolation.

Intratumoral vascular hyperpermeability regulates tumor progression, tumor metastasis, and intratumoral drug delivery (5). Although inhibition of PKA is known to increase vascular permeability in normal tissue (36), the role of PKA in the intratumoral vascular hyperpermeability has not been clarified. Our in vivo FRET imaging has implied that the basal PKA activity is significantly lower in the intratumoral endothelial cells than in the subcutaneous endothelial cells (Fig. 2B) and that the low PKA activity may be responsible for the intratumoral vascular hyperpermeability. Due to the technical difficulties, we needed to observe implanted cancer cells that grow faster than naturally-occurring cancer cells. It awaits further study whether this property is a common hallmark of cancer cells.

VEGF-mediated activation of endothelial VEGFRs is a canonical pathway to increase vascular permeability in diseased tissues, including tumors (35, 37). Accordingly, we observed that VEGF increased the vascular permeability of subcutaneous tissue (Fig. 7A), while a VEGFR inhibitor, Motesanib, reduced the vascular permeability of tumor tissue (Fig. 6A). Importantly, under both conditions, PKA activity in the endothelial cells was inversely correlated with the vascular permeability. In contrast to our observations, Xiong and colleagues have shown that VEGF increases cAMP and activates PKA in HUVECs (38). We have also confirmed that VEGF activates PKA in HUVECs (Fig. 7D and E). In vivo environmental factors, such as cytokines, growth factors, extracellular matrix, or 3D structures including an interaction with pericytes, may account for this discrepancy between the in vivo and in vitro observations. For example, two VEGFR-family proteins, VEGFR1 and VEGFR2, are localized at the apical and basolateral surfaces of endothelial cells, respectively, in the tissues (39); therefore, when VEGF is secreted by the interstitial cells, VEGFR2, but not VEGFR1, will primarily transduce signals in the diseased tissues. VEGFR2 may suppress, rather than activate, PKA in vivo. In contrast, VEGF will stimulate both VEGFR1 and VEGFR2 in vitro and activate PKA.

How is PKA activity regulated in the normal and intratumoral endothelial cells? In most cell types, PKA activity depends exclusively on the cAMP level (40). Therefore, the endothelial PKA activity may be regulated primarily by the activity of G protein-coupled receptors (GPCR) associated with either Gs or Gi. Previous studies have demonstrated specific expression of Gi coupled receptors such as the Apelin receptor (41) and sphingosine-1-phosphate receptors (42) in tumor endothelial cells. An endogenous PKA inhibitor (PKI) may also be involved in the regulation of PKA (16). It is important to note that not only the authentic ligand to GPCR but also other receptors, such as VEGFR, can regulate the activity of GPCR. Needless to say, we cannot exclude the possible involvement of other cAMP regulators from the regulation of the cAMP level and PKA activity in the endothelial cells. For example, phosphodiesterases have been shown to regulate the cAMP level and PKA activity under various conditions (34). Further studies will be required to elucidate the signaling pathway by which cAMP and PKA activity in the intratumoral endothelial cells are controlled and in turn control the permeability of endothelial cells.

cAMP stabilizes the endothelial barrier and antagonizes the cytokine-mediated increase of vascular permeability (17, 20). To examine which cAMP effector plays a principal role in the maintenance of endothelial barrier function, we used specific activators for PKA and Epac/cAMP-GEF. Unexpectedly, we found that both the PKA-specific activator 6-Bnz and the Epac/cAMP-GEF-specific activator 007 decreased the vascular permeability of tumor tissues (Fig. 5), suggesting that the endothelial barrier is maintained by both PKA and Epac/cAMP-GEF in vivo. Our results also underscored that in vivo FRET imaging can help not only to clarify biological processes, but also in the search for new strategies of anticancer therapy. Previous studies demonstrated that inhibition of tumor vascular permeability could lead to suppression of tumor growth (43) and improvement of chemotherapy (44). Taken these observations into account, by restoring PKA activity in the intratumoral endothelial cells to the level in normal endothelial cells, the intratumoral vascular hyperpermeability may be alleviated and thereby control tumor growth.

A drawback of the in vivo FRET imaging is that validation of the specificity is not as easy as in studies by using tissue culture cells. For example, we cannot exclude the possibility that the PKA biosensor is also phosphorylated by other kinases, unless we use PKA knockout mice. Similarly, because of the specificity issue (45), the H89-induced decrease in FRET signal (Supplementary Fig. S2) may not be sufficient for excluding the involvement of other kinases in the high basal FRET signal in normal endothelial cells. Nevertheless, we conclude that in vivo FRET imaging is a powerful approach to fill the gap between in vitro signal transduction research and in vivo cancer biology, if we understand the limitation of this technology.

No potential conflicts of interest were disclosed.

Conception and design: F. Yamauchi, Y. Kamioka, T. Yano, M. Matsuda

Development of methodology: F. Yamauchi, Y. Kamioka, M. Matsuda

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Yamauchi, Y. Kamioka

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Yamauchi, M. Matsuda

Writing, review, and/or revision of the manuscript: F. Yamauchi, Y. Kamioka, M. Matsuda

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Yamauchi, Y. Kamioka, M. Matsuda

Study supervision: Y. Kamioka, T. Yano, M. Matsuda

We are grateful to Kees Jalink for Epac cAMP-sensor, Shigetomo Fukuhara for insightful suggestions and the members of the Matsuda Laboratory for their helpful input, Y. Inaoka, K. Hirano, K. Takakura, and A. Kawagishi for their technical assistance, and Medical Research Support Center of Kyoto University for in vivo imaging.

M. Matsuda was funded by the Innovative Techno-hub for Integrated Medical Bio-imaging Project of MEXT, by the Platform Project for Supporting in Drug Discovery and Life Science Research (Platform for Dynamic Approaches to Living System) of MEXT and AMED, Japan, by a Grant-in-Aid for Scientific Research on Innovative Areas "Resonance Biology" of MEXT, and by the Naito Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.