Research ResourceTechniques

cAMPr: A single-wavelength fluorescent sensor for cyclic AMP

See allHide authors and affiliations

Science Signaling  06 Mar 2018:
Vol. 11, Issue 520, eaah3738
DOI: 10.1126/scisignal.aah3738

Live-cell imaging of cAMP

The second messenger cAMP is found in many cell types and organisms. Increases in the cytosolic concentration and localization of cAMP mediate various cellular functions through the activation of the kinase PKA and the transcription factor CREB; thus, the ability to detect and image cAMP in live cells would help to understand its mechanism of action. Hackley et al. developed a genetically encoded fluorescent sensor of cAMP (cAMPr), which was expressed in mammalian neurons and in the Drosophila brain and responded to stimulation of adenylyl cyclase, the enzyme that generates cAMP. Furthermore, cAMPr was used together with a fluorescent Ca2+ sensor to simultaneously detect cAMP and Ca2+ in the Drosophila brain through two-photon imaging, suggesting its potential use to study cAMP signaling in diverse tissues and organisms.


Genetically encoded fluorescent sensors enable cell-specific measurements of ions and small molecules in real time. Cyclic adenosine monophosphate (cAMP) is one of the most important signaling molecules in virtually all cell types and organisms. We describe cAMPr, a new single-wavelength cAMP sensor. We developed cAMPr in bacteria and embryonic stem cells and validated the sensor in mammalian neurons in vitro and in Drosophila circadian pacemaker neurons in intact brains. Comparison with other single-wavelength cAMP sensors showed that cAMPr improved the quantitative detection of cAMP abundance. In addition, cAMPr is compatible with both single-photon and two-photon imaging. This enabled us to use cAMPr together with the red fluorescent Ca2+ sensor RCaMP1h to simultaneously monitor Ca2+ and cAMP in Drosophila brains. Thus, cAMPr is a new and versatile genetically encoded cAMP sensor.


Live imaging is a powerful tool to understand the dynamics of cell signaling because it enables constant monitoring of an individual cell over time (1). Numerous sensors exist that report real-time changes in the amounts of small molecules and ions or the activities of specific biochemical pathways. Genetically encoded sensors are useful because they can be targeted to specific cell types within tissues and even to subcellular regions. Most genetically encoded sensors have either one or two fluorescent proteins fused to one or two protein domains. These protein domains alter conformation in response to changes in small-molecule or ion concentrations. Single fluorescent protein sensors typically use switches in conformational states of the fused protein(s) to alter the chromophore environment of the fluorescent protein and change the amount of fluorescence. Sensors with two fluorescent proteins link conformational changes of the fused protein domain to changes in the distance between the two fluorescent proteins, which can be monitored by changes in Förster resonance energy transfer (FRET) or fluorescence lifetime imaging.

The genetically encoded sensor GCaMP is a single-wavelength Ca2+ sensor (2). GCaMP consists of a circularly permuted green fluorescent protein (cpGFP) with the M13 peptide from myosin light-chain kinase fused to the N terminus of cpGFP and calmodulin (CaM) fused to the C terminus. CaM interacts with M13 in a Ca2+-dependent manner, which leads to increased cpGFP fluorescence that reflects intracellular Ca2+. GCaMP has gone through multiple rounds of design over 15 years, culminating in GCaMP6, which is sensitive enough to record single action potentials, shows comparable performance to small-molecule Ca2+ indicators, and can be used to map neuronal circuits (3, 4).

Cyclic adenosine monophosphate (cAMP) is a widely used second messenger in living organisms. The amount of cAMP differs between cells, and increasing cAMP in different subcellular compartments can result in different outputs (5, 6). Many neurotransmitters and neuropeptides activate heterotrimeric guanine nucleotide protein–coupled receptors (GPCRs) that regulate adenylate cyclase activity and either increase or decrease the concentration of intracellular cAMP. Changes in cAMP concentration have numerous effects in neurons, including regulating gene expression through the transcription factor CREB (cAMP response element binding protein) (79). Because changes in cAMP concentration can be locally triggered at synapses, it may be possible to monitor the strength and location of synaptic connections by detecting changes in cAMP abundance.

Multiple genetically encoded cAMP sensors are available, but the ones with the best kinetics, intensity, and fold change are based on Exchange protein directly activated by cAMP 1 (Epac-1) and Epac-2. Epac proteins are guanine nucleotide exchange factors (GEFs) with a cAMP-binding domain; GEF activity depends on cAMP binding. Epac-based cAMP sensors use FRET between cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP), in which CFP is excited and emissions from CFP and YFP are monitored. Increasing the amount of cAMP decreases FRET, leading to an increase in CFP emission and a decrease in YFP emission (10, 11). These FRET-based Epac-cAMP sensors have also undergone several rounds of improvements, mostly by switching the fluorescent proteins fused to Epac-1 or Epac-2 (12, 13). Other cAMP sensors include FRET sensors based on protein kinase A (PKA). However, these sensors respond more slowly and with a smaller fold change than Epac-based sensors and require transfecting cells with two separate constructs, which limits their usefulness (11). A-kinase activity reporters (AKARs) are another family of FRET sensors, although these detect the activity of PKA signaling pathways rather than changes in cAMP (14).

Although these sensors have helped understand cAMP signaling, CFP/YFP FRET imaging has some drawbacks. For example, the shorter wavelengths required to excite CFP tend to penetrate tissues poorly, making FRET sensors more difficult to image. Other limitations of FRET include the differential scattering of photons from CFP and YFP, the tendency for CFP and YFP to bleach, and CFP also emitting fluorescence in the YFP channel (15). In addition, simultaneously measuring cAMP and Ca2+ in a single neuron is desirable, but this is challenging with a FRET sensor that already uses two fluorescent proteins. A single-wavelength sensor for cAMP Difference Detector in situ (cADDis) based on Epac-2 was developed, but it remains to be fully characterized in vivo and is only available commercially (16). Although this sensor is a step forward for cAMP imaging, there are still substantial issues to be resolved. cADDis has only been tested in mammalian cell culture, and it is unknown whether cADDis adversely affects cellular physiology. cADDis also has a relatively high Kd (dissociation constant) for cAMP (~10 to 100 μM) in vitro; thus, this sensor may only detect relatively high amounts of cAMP. To combat the limitations of the various FRET and Epac-based sensors, we developed a new single-wavelength sensor for cAMP.

We constructed a sensor using PKA because the cAMP-dependent protein interactions of PKA have been extensively studied (1719). The strategy that we used was a rapid preliminary screen in bacteria followed by transfecting embryonic stem (ES) cells with constructs encoding the brightest variants. With this second step, we tested sensor variants in rapidly dividing cells and then in post-mitotic neurons derived from ES cells. We called the resulting sensor cAMPr. We tested cAMPr in circadian pacemaker neurons in intact Drosophila brains and demonstrated that cAMPr responds robustly to endogenous signals within an intact neural circuit. Finally, we showed that cAMPr can be used with RCaMP1h, a red-shifted fluorescent Ca2+ reporter, to simultaneously monitor cAMP and Ca2+ in individual neurons in the intact fly brain. Thus, cAMPr is a new and versatile genetically encoded sensor for cAMP.


Rationale for using PKA-R amino acid residues 91 to 244

PKA exists as a holoenzyme of two catalytic (PKA-C) and two regulatory (PKA-R) subunits. PKA-R subunits dimerize through the dimerization/docking (D/D) domain, which can also direct the subcellular localization of the PKA holoenzyme. We used amino acid residues 91 to 244 of PKA-R, which lack the D/D domain, to ensure that the sensor would remain in the cytoplasm (19). Removing the D/D domain changes the overall stoichiometry of the PKA-R/PKA-C complex to 1:1 and should also prevent interaction with endogenous PKA subunits. Full-length PKA-R has two cAMP-binding domains, which function cooperatively. However, the second cAMP-binding domain is missing in PKA-R residues 91 to 244 (19). This increases the in vitro Kd for cAMP from 10 nM for full-length PKA-R to 1 μM and reduces the cooperativity of cAMP binding from n = 1.5 to n = 1 (19). Together, these properties should extend the range of a sensor based on PKA-R residues 91 to 244, which is important because cAMP concentrations span a wide range in different cell types (20).

Initial sensor construction and linker library screening

We initially constructed two versions of the sensor. C-G-R consists of PKA-C fused to the N terminus of cpGFP with a Pro-Gly linker and PKA-R residues 91 to 244 fused to the C terminus of cpGFP with an Ala-Cys linker (Fig. 1A). R-G-C had the opposite configuration of PKA subunits relative to cpGFP and lacked linkers. R-G-C fluoresced in bacteria but showed no response to forskolin in ES cells and thus was not pursued.

Fig. 1 Design and selection of a cpGFP-based cAMP sensor.

(A) Schematic and putative mechanism of cAMPr. PKA-C (orange) is the full-length catalytic subunit of protein kinase A, cpGFP (green) is circularly permuted GFP, and PKA-R 91-244 (blue) contains amino acid residues 91 to 244 of the regulatory subunit of PKA. The binding of cyclic adenosine monophosphate (cAMP; yellow) to the PKA-R peptide triggers a conformational change that releases PKA-C from PKA-R and increases fluorescence. (B) Flowchart of linker library generation and screening as a fusion protein with the signal peptide from TorA in Escherichia coli (gray). PG is a two–amino acid linker inserted between PKA-C (C, orange) and cpGFP (G, green). AC is a two–amino acid linker inserted between cpGFP and PKA-R amino acids 91 to 244 (R, blue) to generate C-G-R. Three different linker libraries were made with four random amino acids (X)4 inserted between PG and cpGFP (left), between PG and cpGFP and between cpGFP and AC (middle), or between cpGFP and AC (right). The most strongly fluorescent bacteria were selected by flow cytometry analysis. The chosen bacteria were then grown before plasmid DNA was extracted.

We expressed C-G-R in Escherichia coli where it was targeted to the periplasm. The first C-G-R construct with the shortest linkers did not fluoresce in E. coli at either 18° or 37°C. Because amino acid linkers can affect both basal fluorescence and the fold change of cpGFP-based sensors (2, 21), we constructed linker libraries consisting of ~30,000 individual clones (Fig. 1B). Many bacteria fluoresced under the microscope, albeit most quite weakly. However, the fluorescence occurred in inclusion bodies, and we could not detect any change in fluorescence upon adding cAMP. Thus, we selected the brightest variants by flow cytometry and then moved to a eukaryotic system to test their responsiveness to cAMP. We isolated eight bacteria that were ~5- to 10-fold brighter than most of the ~1 million bacteria that were sorted.

Isolating forskolin-responsive cAMP sensors in ES cells

We chose ES cells to characterize cAMP sensor variants for two reasons. First, site-specific recombination can be used to integrate DNA into a defined genomic location (22). This should remove variation in expression due to chromosomal location, thereby facilitating comparison between clones expressing different variants. Second, ES cells can be differentiated into numerous cell types, enabling a cAMP sensor to be tested in different cellular environments, including neurons.

The variants from the eight bacteria selected by flow cytometry were cloned as a pool into the P2Lox vector (Fig. 2A) (22). A17iCRE ES cells were transfected, and DNA integration was selected by antibiotic resistance. We induced sensor expression by adding doxycycline for 2 days. Imaging revealed a mix of fluorescent and nonfluorescent colonies. We then applied forskolin to stimulate adenylate cyclase activity to increase the concentration of intracellular cAMP. Adding forskolin increased the fluorescence of the fluorescent colonies by ~30 to 40%; no change was seen in the nonfluorescent colonies (Fig. 2B). Adding the vehicle dimethyl sulfoxide (DMSO) did not change the fluorescence (Fig. 2B).

Fig. 2 Initial testing and characterization of cAMPr in ES cells.

(A) Overview of the screening process in embryonic stem (ES) cells. Forskolin was added to ES cells expressing the cAMP sensor variants, and those colonies that exhibited an increase in fluorescence were selected and sequenced. The sequences of the linkers of the forskolin-responsive clones are shown. (B) Left: Response of ES cell colonies expressing cAMPr to 40 μM forskolin (red; n = 15 colonies) or dimethyl sulfoxide (DMSO) vehicle (blue; n = 16 colonies) from two independent experiments. The gray box indicates the time of drug addition. Light red and blue shading represents the SEM. Data are normalized to the fluorescence at time zero. Right: Images show representative single ES colonies before (Pre; left) and after (Post; right) the application of 40 μM forskolin (top) or DMSO (bottom). P was calculated using a standard unpaired two-sided t test. (C) Response of cAMPr in digitonin-permeabilized ES cells to 1 mM cAMP or 1 mM cGMP. Left: Graph shows the average of at least 13 colonies from two independent experiments, and the data were plotted as described in (B). Right: Bar graph showing the maximum fold changes in fluorescence plotted for 1 mM cAMP (red) and 1 mM cGMP (Blue). (D) Response of ES cell colonies expressing cAMPr (left) or cAMP Difference Detector in situ (cADDis) (right) to the indicated concentrations of forskolin. Cells expressing cADDis that were exposed to DMSO are shown in black. Data are from 14 to 20 colonies from two independent experiments for all treatments. Data are plotted as described in (B). (E) Maximum average (max avg) fold change in fluorescence in response to the indicated concentrations of forskolin in ES cells expressing cAMPr (left) or cADDis (right) normalized to the maximum response to 40 μM forskolin using data from (D). Two-sample unpaired t tests for cAMPr showed P < 0.05 for each value plotted compared to the next highest concentration (40 μM versus 10 μM, P < 0.05; 10 μM versus 1 μM, P < 0.001; 1 μM versus 100 nM, P < 0.001; 100 nM versus 10 nM, P < 10−5). For cADDis, two-sample unpaired t tests only showed statistical significant differences for 1 μM versus 100 nM when compared to the next highest value (40 μM versus 10 μM, P = 0.7; 10 μM versus 1 μM, P = 0.5; 1 μM versus 100 nM, P < 10−8, 100 nM versus 10 nM, P = 0.10; 10 nM versus DMSO, P = 0.44).

To identify the genotype of the forskolin-responsive clones, we selected six fluorescent clones. Sequencing revealed that four of the six fluorescent clones had the original Pro-Gly sequence as the N-terminal linker and Thr-Ala-Pro-Pro-Ala-Cys as the C-terminal linker. Surprisingly, the other two clones had the original Pro-Gly and Ala-Cys linkers as N- and C-terminal linkers, respectively (Fig. 2A). There were no mutations in either variant. To determine which variant should be fully characterized, we directly compared the performance of the two forskolin-responsive variants. Cells from two clones of each variant were plated in 24-well plates and imaged before and after the addition of 40 μM forskolin. The two variants had similar fold changes in fluorescence and kinetics (fig. S1). We selected the sensor with the Trp-Ala-Pro-Pro-Ala-Cys C-terminal linker because it had a slightly faster response time to forskolin than the shorter linker variant and because it fluoresced in both ES cells and bacteria. We named this variant cAMPr.

ES cells expressing cAMPr exhibited a ~50% increase over baseline fluorescence in response to forskolin and did not exhibit a response to DMSO (Fig. 2B and movies S1 and S2). Furthermore, cAMPr had a cytosolic distribution in ES cells before and after exposure to forskolin or DMSO, as predicted from deleting the PKA-R D/D domain (19). The excitation and emission spectra for cAMPr expressed in ES cells with and without forskolin had the expected excitation and emission maxima for green fluorescent protein (GFP; fig. S2). Thus, increasing the cAMP concentration with forskolin changed the fluorescence output of cAMPr but not its excitation or emission profiles.

Specificity of cAMPr for cAMP in ES cells

For cAMPr to be a genuine cAMP sensor, the increase in fluorescence with forskolin must be due to increases in cAMP abundance, not secondary or indirect changes in other cyclic nucleotides. To test this, we used digitonin to permeabilize the membranes of ES cells expressing cAMPr and exposed them to either 1 mM cAMP or 1 mM cyclic guanosine monophosphate (cGMP). cAMPr showed a robust ~40% increase in fluorescence in response to cAMP but no change in response to cGMP (Fig. 2C). Thus, cAMPr exhibited specificity for cAMP over cGMP and directly responded to changes in cAMP concentration.

Comparison of cAMPr to cADDis in ES cells

Next, we compared the performances of cAMPr and cADDis. To obtain ES cells expressing cADDis, we infected A17iCRE ES cells with a baculovirus expressing cADDis. Because this produced cells with a range of fluorescence intensities, we chose cells with a similar fluorescence to that of cAMPr-expressing ES cells. We plated the cells in 24-well plates and imaged them 2 days later. We compared the average normalized response to decreasing amounts of forskolin for ES cells expressing cAMPr or cADDis (Fig. 2D). The data showed that cAMPr had a more linear response to different forskolin concentrations, whereas cADDis had a more sigmoidal response (Fig. 2E). Although cAMPr had a slightly lower maximum fold change than that of cADDis, we concluded that cAMPr was a better sensor for detecting small changes in cAMP concentration.

Kinetics and imaging of cAMPr in ES cells

Next, we tested how accurately cAMPr fluorescence intensity reflected changes in cAMP concentration in real time by measuring cAMPr responses to transient forskolin application. To dynamically control cAMP amounts, we plated ES cells on glass coverslips and performed the imaging in a flow chamber. ES cells expressing cAMPr showed no response to DMSO applied for 180 s (Fig. 3A, blue bar), whereas 40 μM forskolin (Fig. 3A, red bar) increased fluorescence by ~25%. The fluorescence gradually dissipated back to baseline by ~9 min. This time frame is consistent with the expected rates of cAMP hydrolysis, suggesting that cAMPr fluorescence intensity mirrored the changes in cAMP concentration (23).

Fig. 3 cAMPr kinetics and two-photon compatibility in ES cells in a flow chamber.

(A) Response of cAMPr-expressing ES cells to DMSO (blue) and 40 μM forskolin (red) applied for the indicated periods to cells in a flow chamber. Data are plotted as the fold change in fluorescence normalized to that at time zero. Graph shows an average of 10 colonies from a single experiment imaged by single-photon excitation. (B) Response of cAMPr-expressing ES cells to sequential exposure to forskolin. Forskolin (40 μM; red) was applied to cAMPr-expressing ES cells in a flow chamber from t = 120 to 240 s and from t = 660 to 800 s. The graph shows an average of 15 colonies from a single experiment using single-photon excitation, and the data are plotted as described in (A). (C) Response of cAMPr-expressing ES cells to 40 μM forskolin (red) in a flow chamber and imaged by two-photon microscopy. Graph shows an average of 15 colonies from a single experiment, and the data are plotted as described in (A).

To test whether cAMPr retained sensitivity over multiple stimulations, we sequentially applied 40 μM forskolin to ES cells expressing cAMPr in a flow chamber after the fluorescence values returned to baseline. The responses to the first and second applications of forskolin were similar in both fold change and kinetics (Fig. 3B), indicating that cAMPr can undergo more than one activation cycle without losing sensitivity. Because two-photon imaging enables better sample penetration than single-photon imaging, we tested whether cAMPr was compatible with two-photon imaging and thus could be useful for deep tissue imaging. We excited ES cells expressing cAMPr in a flow chamber with a two-photon laser at 920 nm and exposed the cells to forskolin. We found that cAMPr produced a similar fold change and kinetic profile to forskolin under two-photon (Fig. 3C) and single-photon (Fig. 3, A and B) excitation. Thus, cAMPr is compatible with two-photon imaging and should be useful for in vivo applications.

Testing cAMPr in mammalian neurons

ES cells can be differentiated into different cell types. Because we were particularly interested in using cAMPr to study neuronal signaling, we differentiated ES cells expressing cAMPr into neurons. We induced ES cells to form embryoid bodies (EBs) and then added either retinoic acid (RA) or a Smoothened agonist (SAG) to differentiate the cells into two different populations of neurons (Fig. 4A) (24, 25). RA-induced differentiation produces a mixture of different types of ventral spinal cord neuron progenitors (24), whereas activation of Hedgehog signaling pathways directs ES cells to become ventral forebrain neurons (25). To test whether cAMPr functioned in mammalian neurons, we measured the response of RA- and SAG-differentiated neurons to forskolin. Both RA-differentiated neurons (Fig. 4B and movie S3) and SAG-differentiated neurons (Fig. 4C and movie S4) responded to forskolin with an increase in fluorescence. Next, we assessed whether cAMPr could detect changes in cAMP stimulated by endogenous GPCRs. For this, we treated the differentiated neurons with dopamine, which can either increase or decrease cAMP concentration, and γ-aminobutyric acid (GABA), which either decreases cAMP or activates a chloride channel. Most of the SAG-differentiated neurons (>95%) responded to dopamine with increased fluorescence (Fig. 4C and movie S5). The response was specific for dopamine because GABA did not cause any change in fluorescence (Fig. 4C).

Fig. 4 cAMPr responses in ES cells induced to differentiate into neurons.

(A) Schematic of the protocols used to induce ES cells to differentiate into neurons. Doxycycline was added to induce cAMPr expression. (B) Left: Response of retinoic acid (RA)–differentiated neurons expressing cAMPr to either 40 μM forskolin (red) or DMSO (blue). Data are from 14 and 18 neurons, respectively, from two independent experiments. Data are plotted as described in Fig. 2B. Right: Images of individual representative RA-differentiated neurons expressing cAMPr showing fluorescence before (Pre) and after (Post) the addition of 40 μM forskolin (top) or DMSO (bottom). (C) Left: Response of Smoothened agonist (SAG)–differentiated neurons expressing cAMPr to 10 μM forskolin, 1 mM dopamine, or 100 μM γ-aminobutyric acid (GABA). Data are from 18 to 20 neurons per treatment, with two replicates per treatment, from two independent experiments. Data are plotted as described in Fig. 2B. Right: Representative images of a typical SAG-differentiated neuron expressing cAMPr before (Pre) and after (Post) the addition of 1 mM dopamine. The same neuron is shown in grayscale (top) and pseudocolor (bottom). (D) Responses of SAG-differentiated neurons expressing either cAMPr (left) or cADDis (right) to the indicated concentrations of dopamine. Fluorescence was quantified from 15 to 20 neurons for each dopamine concentration from two independent experiments. Two-sample unpaired t tests showed a statistical significance of P < 0.05 for all dopamine concentrations in neurons expressing cAMPr except for 1 mM versus 100 μM when compared to the next highest value (1 mM versus 100 μM, P = 0.90; 100 μM versus 10 μM, P < 0.05; 10 μM versus 1 μM, P < 10−8; 1 μM versus 100 nM, P < 0.01). For cADDis-expressing neurons, only 10 μM versus 1 μM was statistically significantly different (1 mM versus 100 μM, P = 0.07; 100 μM versus 10 μM, P = 0.90; 10 μM versus 1 μM, P < 10−7; 1 μM versus 100 nM, P = 0.75).

RA-differentiated neurons expressing cAMPr also responded to dopamine but not to GABA (fig. S3, A and B, and movie S6). However, the responses were more variable than those of SAG-differentiated neurons, which is consistent with the diversity of neurons resulting from RA differentiation protocols (24). The results from experiments with neurons derived from ES cells expressing cAMPr indicate that this cAMP sensor is sufficiently sensitive to detect neurotransmitter signaling mediated by endogenous GPCRs in cultured neurons. We exposed SAG-differentiated neurons expressing cAMPr to increasing concentrations of dopamine and calculated a median effective concentration (EC50) of these neurons for dopamine of ~5 μM (fig. S3C), which is similar to the reported Kd of 2.5 to 4 μM of the dopamine receptor D1 for dopamine (26, 27). We also performed a similar experiment with SAG-differentiated neurons infected with cADDis-expressing baculovirus (Fig. 4D and fig S3D). In contrast to the cAMPr-expressing neurons, the cADDis-expressing neurons appeared to give an all-or-none response to decreasing amounts of dopamine, whereas cAMPr showed a more sigmoidal response (Fig. 4D). Thus, cAMPr produced a more accurate quantitative response to dopamine concentrations than cADDis in neurons differentiated from ES cells.

Testing cAMPr in Drosophila pacemaker neurons

To test cAMPr in an intact brain, we chose to study Drosophila circadian pacemaker neurons. These neurons respond to the Pigment-dispersing factor (PDF) neuropeptide by increasing the concentration of cAMP, and they were previously used to validate Epac-camps in transgenic flies (28). We generated transgenic flies by integrating UAS-cAMPr into a defined genomic locus using the attB/attP targeting system.

We first focused on the lateral ventral circadian pacemaker neurons (LNvs). There are two types of LNvs in adult flies: The small LNvs (s-LNvs) are the principal pacemaker cells during constant darkness (29), whereas the large LNvs (l-LNvs) are important in sleep but are not critical for maintaining circadian rhythms (30, 31). Using the Pdf-Gal4 promoter, we expressed cAMPr in all adult LNvs. To test whether cAMPr detected changes in cAMP abundance induced by forskolin in intact brains, we dissected brains from adult flies, placed the brains in a glass-bottom dish, and used two-photon imaging to measure fluorescence in the LNvs in response to either 40 μM forskolin or solvent (DMSO). We quantified the change in fluorescence from both sets of LNvs, which exhibited a ~40% increase over baseline in cAMPr fluorescence in response to forskolin and a small decrease in fluorescence in response to DMSO (Fig. 5A). Thus, cAMPr responded to forskolin in an intact brain and detected changes in cAMP abundance in a nonmammalian system.

Fig. 5 Testing cAMPr in Drosophila pacemaker neurons.

(A) Top: Diagram showing an adult Drosophila brain and the approximate location of adult lateral ventral circadian pacemaker neurons (LNvs) expressing cAMPr in Pdf-Gal4/UAS-cAMPr; Pdf-Gal4/+ adult flies (green). Bottom: Response of LNvs expressing cAMPr to either 40 μM forskolin (red) or DMSO (blue). The data are an average of 12 LNvs from three brains. (B) Top: Diagram showing an adult Drosophila brain and the approximate location of adult LNvs expressing Epac-camps in Pdf-Gal4/UAS-Epac-camps 50A; Pdf-Gal4/+ adult flies (light blue). Bottom: Response of LNvs expressing Epac-camps to either 40 μM forskolin (red; n = 5 neurons from two brains) or DMSO (blue; n = 8 neurons from three brains). (C) Diagram illustrating communication between larval LNv and DN (dorsal neuron) clock neurons. cAMPr (green) was expressed in all nine clock neurons per hemisphere using tim(UAS)-Gal4 (five LNvs and four DNs). The four pigment-dispersing factor (PDF)–expressing LNvs were also programmed to express the mammalian purinergic receptor P2X2 (orange circle). Adenosine 5′-triphosphate (ATP) stimulates the four LNvs to fire and release the neuropeptide PDF (black dots), which bind to and activate the PDF receptor (PDFR; black circles) present on all larval clock neurons. Flies of the following genotypes were tested using two-photon imaging: tim(UAS)-Gal4/UAS-cAMPr; Pdf-LexA, LexAOP-P2X2/+ (Pdf > P2X2); tim(UAS)-Gal4/UAS-cAMPr; +/TM6B] (Control). (D) Left: Images of cAMPr responses in larval DNs before (Pre-ATP) and after (Post-ATP) the application of ATP. Images are in grayscale (top) and pseudocolor (bottom). DNs were from larvae in which P2X2 was expressed only in LNvs, and cAMPr was expressed in all clock neurons. Right: Quantification of the cAMPr responses in larval DNs. Red: DNs with P2X2 in LNvs and cAMPr in all clock neurons. Data are from four neurons from three brains. Blue: DNs from control larvae with no Pdf-LexA or LexA-P2X2 transgenes. Data are from 12 neurons from three brains. ATP (2.5 mM) was added during the time represented by the gray box. Data are plotted as described in Fig. 2B. (E) Left: Images of cAMPr responses in larval LNvs before (Pre-ATP) and after (Post-ATP) the application of ATP. Images are in grayscale (top) and pseudocolor (bottom). LNvs were from larvae in which P2X2 was expressed in LNvs and cAMPr was expressed in all clock neurons. Right: Quantification of cAMPr responses in DNs. Red: LNvs from larvae with P2X2 in LNvs and cAMPr in all clock neurons. Data are from 22 neurons from three brains. Blue: LNvs from control larvae with no Pdf-LexA or LexA-P2X2 transgenes. Data are from 23 neurons from three brains. ATP (2.5 mM) was added during the time represented by the gray box. Data are plotted as described in Fig. 2B.

We compared cAMPr to the FRET-based cAMP sensor Epac-camps, which we expressed in LNvs using the Pdf-Gal4 promoter. Forskolin stimulated a ~35% decrease in the YFP/CFP ratio owing to the decrease in FRET (Fig. 5B; data are shown inverted for easy comparison to Fig. 5A) in the dissected adult brains, which is similar to previous reports (28); DMSO had no effect. The data showed that cAMPr and Epac-camps responded to forskolin with similar fold changes and kinetics. Thus, cAMPr resolves the limited dynamic range and slow kinetics of the previous PKA-based sensors and does not have the limitations of FRET-based sensors with two fluorescent proteins (11, 14).

Because cAMPr contains an active kinase and because genetically encoded sensors can disrupt cellular physiology, we tested whether LNvs that expressed cAMPr functioned normally. This is particularly important because altering PKA activity or cAMP abundance can change the period length and strength of Drosophila locomotor activity rhythms in constant darkness (3234). However, flies expressing UAS-cAMPr controlled by Pdf-Gal4 in LNvs had strong rhythms and a period length of 23.8 hours in constant darkness, which was not statistically significantly different from those of heterozygous UAS-cAMPr flies (23.7 hours) or Pdf-Gal4 control flies (23.9 hours; table S2). These data contrast with those obtained from LNvs expressing UAS-Epac-camps, which exhibit a period lengthened by ~1 hour (28), suggesting that Epac-camps alters either the amount of cAMP or cAMP signaling pathways in LNvs. Thus, cAMPr seems to interfere less with cAMP signaling pathways in LNvs than Epac-camps.

We next tested whether cAMPr was sensitive enough to detect signaling in a neuronal circuit within the brain by stimulating LNvs to fire and then measuring the cAMP responses of downstream neurons. We used the larval circadian neural network, which is simpler than the adult clock network and has only nine clock neurons per hemisphere (35). These nine neurons can be divided into subgroups: Five LNvs project to the dorsal brain and communicate with four dorsal neurons (DNs); four of the five LNvs release the neuropeptide PDF, which increases intracellular cAMP concentration by binding to and activating the Gαs-coupled PDF receptor (PDFR), which is likely found in all larval clock neurons (28, 36). We expressed cAMPr in all larval clock neurons using tim-Gal4. To induce LNv firing and thus PDF release, we also expressed the mammalian purinergic receptor P2X2 in LNvs using a LexAOp-P2X2 transgene (Fig. 5C). Adenosine triphosphate (ATP) activates P2X2 receptors, which depolarizes neurons, causing them to release neurotransmitters and neuropeptides. Because there is no P2X2 ortholog in Drosophila, adding ATP to brains with Pdf-LexA and LexAOp-P2X2 specifically activates LNvs.

We performed two-photon imaging of larval brains with P2X2 expressed only in LNvs and cAMPr in all clock neurons and exposed the brains to 2.5 mM ATP. We found that applying ATP increased cAMPr fluorescence in DNs (Fig. 5D and movie S7). These responses required P2X2 expression in LNvs because cAMPr fluorescence in DNs from control larvae without a P2X2 transgene decreased slightly after adding ATP (Fig. 5D). However, all neurons imaged from these control brains responded to forskolin, indicating that cAMPr was functional in DNs (fig. S4). Analogous experiments with GCaMP in adult s-LNvs also showed slightly decreased fluorescence after the addition of ATP to brains lacking P2X2, suggesting that this may be a general response of GFP derivatives to ATP (37, 38).

Larval LNvs expressing both cAMPr and P2X2 also responded to ATP with increased cAMPr fluorescence (Fig. 5E and movie S8). This was expected because adult s-LNvs express PDFR and respond to exogenously applied PDF (28). However, equivalent experiments in adult s-LNvs using Epac-camps did not detect changes in cAMP after s-LNv depolarization (37). This difference probably results from the increased sensitivity of cAMPr at detecting small changes in cAMP concentration compared to Epac-camps. Thus, our data from experiments with Drosophila circadian pacemaker neurons show that cAMPr can detect endogenous signals in the whole brain as assessed by two-photon imaging and suggest that cAMPr is more sensitive than Epac-camps for detecting cAMP signaling in vivo.

Simultaneously measuring Ca2+ and cAMP in whole brains

Ca2+ is another key signaling mediator of neuronal signaling; thus, it would be helpful to be able to simultaneously measure changes in cAMP and Ca2+ concentrations. As a single-wavelength reporter, cAMPr should make this possible when used in conjunction with RCaMP1h, a red-shifted, genetically encoded Ca2+ sensor (21). We used Pdf-Gal4 to simultaneously express transgenes encoding cAMPr and RCaMP1h in larval LNvs. Larval brains were dissected and then treated with carbachol, a general cholinergic agonist. Carbachol depolarizes PDF-expressing LNvs (34, 39), which increases intracellular Ca2+ concentrations through voltage-gated Ca2+ channels and leads to the release of PDF from the LNvs. PDF then activates PDFR on LNvs in an autocrine or paracrine manner, which increases the concentration of cAMP in LNvs. We monitored RCaMP1h and cAMPr fluorescence in LNv cell bodies and projections simultaneously (Fig. 6A). We noted that RCaMP1h fluorescence increased in LNv projections before it increased in cell bodies (Fig. 6B and movie S9). The fold change in cAMPr fluorescence was greater in LNv cell bodies than in the projections, which may indicate differences in PDF release, PDFR density, or signaling strength (Fig. 6C). We also noted that the increase in Ca2+ preceded the increase in cAMP in cell bodies and projections (Fig. 6, D and E).

Fig. 6 Simultaneous measurement of Ca2+ and cAMP in whole brains.

(A) Images of PDF-expressing LNv cell bodies or projections before (0 s) or during (150 and 300 s) the treatment of larval brains with carbachol. Top: cAMPr fluorescence in green and pseudocolor to measure intracellular cAMP. Middle: RCaMP1h in red and pseudocolor to measure intracellular Ca2+. Bottom: Merged cAMPr (green) and RCaMP1h (red) fluorescence images to simultaneously show cAMP and Ca2+. (B) RCaMP1h fluorescence from 12 PDF-expressing LNv cell bodies and projections from four brains. Shading indicates SEM. (C) cAMPr fluorescence from 12 PDF-expressing LNv cell bodies and projections from four brains. Shading indicates SEM. (D) Data replotted from (B) and (C) to compare the dynamics of changes in cAMPr and RCaMP1h fluorescence in PDF-expressing LNv cell bodies. (E) Data replotted from (B) and (C) to compare the dynamics of changes in cAMPr and RCaMP1h fluorescence in PDF-expressing LNv projections.

The previous experiment used single-photon imaging. We also tested whether cAMP and Ca2+ could be simultaneously monitored by two-photon imaging (fig. S5). Two-photon (fig. S5A) and single-photon (Fig. 6D) imaging of LNvs expressing the two sensors and exposed to carbachol produced similar changes in fluorescence patterns, although the fold changes were lower with two-photon excitation. We monitored the fluorescence of both sensors with two-photon excitation in isolated larval brains exposed to forskolin instead of carbachol (fig. S5B). As expected, forskolin stimulated an increase in cAMPr fluorescence without affecting RCaMP1h fluorescence. We conclude that cAMPr is a versatile genetically encoded cAMP sensor, which should be widely useful in neurobiology and beyond.


We have generated a new single-color cAMP reporter that is compatible with two-photon imaging and seems to minimally perturb cellular physiology. Tests in both mouse ES cells and neurons differentiated from ES cells showed ~50% increases in cAMPr fluorescence after appropriate stimulation and a return to baseline fluorescence in ~10 min, which is consistent with the kinetics of cAMP signaling. Testing cAMPr in circadian pacemaker neurons in Drosophila brains revealed that cAMPr exhibits a fold change similar to that of the existing Epac family of sensors and is compatible with nonmammalian systems. We also compared cAMPr to the only other available single-wavelength cAMP sensor, cADDis, and showed that cAMPr gave more quantitative readouts of cAMP amounts in both ES cells and neurons. Finally, we showed that cAMPr can be used to detect signaling within an endogenous neural circuit and to simultaneously measure Ca2+ and cAMP in Drosophila brains.

Although functional in its current version, we believe that cAMPr can be improved in the future. For example, the affinity of cAMPr for cAMP could be altered by incorporating previously studied point mutations in the PKA regulatory subunit (40). This could shift the dynamic range of cAMPr to monitor higher or lower cAMP concentrations, which might be necessary for using cAMPr in other cell types or in subcellular compartments with different cAMP amounts from those tested here. Our current version of cAMPr is cytoplasmic, but it could also be targeted to specific subcellular compartments.

In summary, we showed that cAMPr is functional and has minimal effects on cellular physiology in both mammalian cell culture and Drosophila brains. In addition, cAMPr is more sensitive to small changes in cAMP abundance than cADDis. cAMPr can also be easily combined with other single-wavelength sensors, unlike Epac-camps. Finally, cAMPr showed robust responses using either single- or two-photon imaging, with the latter being particularly important for experiments involving either deep tissue or live animals. We believe that cAMPr has many of the characteristics to make it the sensor of choice to study cAMP signaling in a range of different tissues and organisms.


Constructing C-G-R and R-G-C

The cDNAs encoding PKA-C, PKA-R (91-244), and cpGFP were amplified from the plasmids pET15b PKA Cat (41), pRSETB PKA R1α (42), and TorPE-G-GECO1.2 (Addgene plasmids 14921, 14922, and 32468, respectively) (43) using Phusion High-Fidelity DNA Polymerase (New England Biolabs). Purified polymerase chain reaction (PCR) products were mixed together in 1:2:3 ratios (PKA-C/cpGFP/PKA-R) and amplified for 10 cycles without primers. Primers were added, and the DNA was amplified for a further 20 cycles. The sequences encoding C-G-R and R-G-C were subcloned into the pTorPE vector (digested with Xba I and Hind III) using T4 DNA ligase (Promega), and the resulting constructs were used to transform bacteria. Clones were selected, and their sequences were verified. The primers used are listed in table S2.

Constructing and sorting C-G-R linker libraries in bacteria

C-G-R libraries were constructed by digesting pTorPE_C-G-R with Sph I and Xma I. We amplified cpGFP with the appropriate linker library primers for each pool. PCR products were purified, digested with Sph I and Xma I, and ligated into pTorPE_C-G-R that had been digested with the same enzymes. Bacteria were transformed with the contents of the ligation reaction, and five plates were plated per library, resulting in ~30,000 individual clones per library. All three libraries were scraped, mixed together, and grown overnight at 37°C with 0.01% w/v arabinose (Sigma-Aldrich) to induce expression. Cells were sorted for GFP expression on a FACSAria flow cytometer. Eight bright fluorescent outliers were sorted into LB containing ampicillin and grown overnight at 37°C before undergoing DNA extraction.

Generation of ES cell lines

The brightest C-G-R clones from bacteria were amplified and subcloned into Eco RI–digested p2Lox vector (22) using In-Fusion cloning (Clontech). Bacteria containing the C-G-R mini-library were scraped off a plate and grown in LB containing ampicillin before DNA extraction. A2Lox.cre ES cells (22) were then transformed with the purified DNA using a standard nucleofection protocol (Lonza).

C-G-R mini-library imaging and clone selection in ES cells

After transformation, ES cells containing C-G-R were cultured until they were ~80% confluent. Cells were trypsinized, plated at low density on a 35-mm dish, and grown for 5 to 6 days until colonies reached a sufficient size for imaging. Colonies were imaged on a Nikon Eclipse Ti microscope before and after the application of forskolin. Six different clones showing increased fluorescence after forskolin addition were picked and expanded. DNA was extracted for sequencing using a standard phenol/chloroform genomic DNA extraction. Each clone was reimaged before and after forskolin application to ensure a correct match between fluorescent phenotype and DNA sequence.

Differentiating ES cells into neurons

ES cells were differentiated using standard EB formation followed by treatment with either 1 μM RA or 500 nM SAG (24, 25). Differentiated neurons were maintained in a motor neuron buffer composed of Neurobasal-A, 2% fetal bovine serum, 1× B27, 500 nM l-glutamine, 10 nM β-mercaptoethanol, 4 mM 2′-deoxy-5-fluorouridine, and 4 mM uridine (44). Doxycycline (3 mg/ml) was added to induce cAMPr expression.

Imaging cAMPr in ES cells

ES cells were imaged in either 24-well plates or a flow chamber. For 24-well plates, cells were seeded in wells treated with 0.1% w/v gelatin and then cultured in medium containing doxycycline (3 mg/ml) until they reached 25 to 50% confluency. Cells were washed three times with 1× Hanks’ balanced salt solution and imaged on a Leica SP8 confocal microscope every 60 s. Either 40 μM forskolin or DMSO was applied at the times indicated in the figure legends for the initial characterization experiments. For cAMP and cGMP experiments, 1 μM digitonin was applied at the beginning of imaging, and either 1 mM cAMP or cGMP was added at the times indicated in the figure legends. For flow chamber experiments, cells were seeded as described earlier except that a glass coverslip was used. The glass coverslip was placed in a 100-mm cell culture dish, and the cells were cultured as described earlier. Cells were then imaged every 60 s with either single- or two-photon excitation on an Olympus FLV 1000, with drugs added at the times indicated by gray-shaded rectangles in the graphs. We used standard GFP filter sets for single-photon excitation. For two-photon imaging, laser excitation was set at 920 nm with standard filter sets used for monitoring emission.

Imaging cAMPr in differentiated neurons

Differentiated neurons were plated in 24-well plates precoated with 0.01% poly-l-lysine and imaged as described for ES cells, except that images were captured every 80 s. Neurons were treated with 40 μM forskolin, DMSO, 100 μM GABA, or 100 nM to 1 mM dopamine at the times indicated in the figure legends.

Generating UAS-cAMPr flies

The cDNA encoding cAMPr was inserted into Eco RI–digested pUASattB using In-Fusion cloning (Clontech). The pUASattB-cAMPr construct was injected into the VK37 landing site on chromosome 2 by Rainbow Transgenic Flies. Four independent UAS-cAMPr fly lines were tested for their expression of cAMPr in LNvs by crossing to Pdf-Gal4 flies. All lines exhibited the correct spatial expression, and one line was used for imaging.

Imaging in Drosophila brains

Pdf-Gal4 (45) and tim(UAS)-Gal4 (46) flies were crossed to either UAS-cAMPR; Pdf-LexA, LexAOP-P2X2 or UAS-Epac-camps; Pdf-LexA, LexAOp-P2X2 flies. Recombinant Pdf-LexA, LexAOp-P2X2 flies were provided by O. Shafer (University of Michigan). Adult flies were anesthetized on ice and dissected in hemolymph-like saline (HL3) media. Whole brains were imaged in glass-bottom dishes pretreated with 0.01% poly-l-lysine (Sigma-Aldrich) in HL3 medium at room temperature. An Olympus FLV 1000 microscope was used with 488-nm excitation and standard GFP emission settings for cAMPr or with 445-nm excitation and standard CFP/YFP FRET emission settings for Epac-camps. ATP (100 mM; Sigma-Aldrich) was dissolved in 20 mM tris-HCl, and the pH was adjusted to 7.2.

Simultaneously measuring Ca2+ and cAMP in whole brains

Pdf-Gal4 flies were crossed with UAS-cAMPr flies to generate a UAS-cAMPr; Pdf-Gal4 line. These flies were then crossed with UAS-RCaMP1h flies (21) to generate flies expressing cAMPr and RCaMP1h in the PDF-expressing LNvs. Flies were dissected and imaged as described earlier except that a Leica SP8 microscope was used. Excitation at 488 and 561 nm was used to image cAMPr and RCaMP1h, respectively. Two-photon imaging was performed with an Olympus FLV 1000 using 920-nm excitation.

Image quantification

All images were quantified with Fiji/ImageJ software. For ES cells, regions of interest (ROIs) generally included whole colonies, although only a portion of the colony was used for larger colonies. When imaging ES cells that had differentiated into neurons, we only quantified the soma. If two or more cell bodies were directly adjacent, they were quantified together using a single ROI. All values were background-subtracted. The fluorescence at a particular time was divided by the fluorescence at time zero to generate the change in fluorescence values plotted in the graphs.


Fig. S1. Comparison of cAMPr and C-G-R in ES cells.

Fig. S2. cAMPr excitation and emission spectra in ES cells.

Fig. S3. Responses of RA- and SAG-differentiated neurons to neurotransmitters.

Fig. S4. ATP-treated larval clock neurons respond to forskolin.

Fig. S5. Two-photon imaging of cAMPr and RCaMP1h in PDF-expressing LNvs.

Table S1. Locomotor activity rhythms of adult flies in constant darkness.

Table S2. Primers used for cloning.

Movie S1. Time lapse of ES cells expressing cAMPr exposed to forskolin.

Movie S2. Time lapse of ES cells expressing cAMPr exposed to DMSO.

Movie S3. Time lapse of RA-differentiated neurons exposed to forskolin.

Movie S4. Time lapse of SAG-differentiated neurons exposed to forskolin.

Movie S5. Time lapse of SAG-differentiated neurons exposed to dopamine.

Movie S6. Time lapse of RA-differentiated neurons exposed to dopamine.

Movie S7. Time lapse of the response of DNs expressing cAMPr to activation of LNvs by ATP shown in pseudocolor.

Movie S8. Time lapse of the response of LNvs expressing cAMPr to activation by ATP shown in pseudocolor.

Movie S9. Time lapse of the response of LNvs expressing cAMPr and RCaMP1h to carbachol.


Acknowledgments: We thank R. Campbell, J. Park, M. Rosbash, O. Shafer, S. Taylor, and the Bloomington Drosophila Stock Center for sharing DNA and flies. We thank C. Desplan for sharing the two-photon microscope and perfusion chamber, L. Christiaen and Z. Zhu for constructive ideas during the project, and Z. Zhu for comments on the manuscript. Funding: This investigation was conducted in facilities constructed with support from Research Facilities Improvement grant number C06 RR-15518-01 from the National Center for Research Resources, NIH. Flow cytometry and some imaging were performed in the New York University (NYU) Center for Genomics and Systems Biology GenCore facility. C.R.H. was partly supported by the NYU Graduate School of Arts and Science MacCracken Program. Work in the authors’ laboratories was supported by March of Dimes grant 5-FY14-99 and NIH grant R01HD079682 to E.O.M. and the NYU Research Challenge Fund, NYU Abu Dhabi affiliate faculty research funds, and NIH grant GM063911 to J.B. Author contributions: C.R.H. designed cAMPr and planned and performed all experiments. E.O.M. suggested the use of ES cells and provided expertise for these experiments. J.B. advised on experiments. All authors wrote and edited the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The cAMPr sensor construct is available from Addgene (deposit no. 99143). Additional cAMPr variants are also available through Addgene (deposit nos. 84932, 84933, 102437, 102438, 102439, and 102618). All other data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article