Perspective Perspectives on: Cyclic nucleotide microdomains and signaling specificity Can we decipher the information content contained within cyclic nucleotide signals? Thomas C. Rich,1,2,3,6 Kristal J. Webb,1,3 and Silas J. Leavesley3,4,5 1Center for Lung Biology, 2Department of Pharmacology, 3Basic Medical Sciences Graduate Program, and 4Department of Pharmacology, College of Medicine; and 5Department of Chemical and Biomolecular Engineering, 6College of Engineering, University of South Alabama, Mobile, AL 36688 Second messengers such as Ca2+, cGMP, and cAMP are approach for measuring cAMP signals (Adams et al., known to regulate diverse cellular functions includ- 1991). They labeled the catalytic and regulatory sub- ing excitability, contraction, movement, proliferation, units of PKA type I with a fluorescent donor (fluorescein) and gene expression. Our understanding of how Ca2+ and acceptor (rhodamine). When cAMP concentrations signals orchestrate such diverse cellular functions has were low, the subunits were in the holoenzyme com- increased dramatically over the last forty years, due in plex, and FRET occurred between fluorescein and rho- y g large part to the development of single-cell methods for damine. However, when cAMP levels were high, cAMP o measuring intracellular Ca2+ (Tsien, 1992). Our under- bound to the regulatory subunits, the catalytic subunits l o standing of the subcellular localization, kinetics, and dissociated, and FRET diminished. This ingenious i s y frequency of cyclic nucleotide signals has lagged far be- method has been described as a real-time cAMP sensor h hind. Thus, we are only beginning to understand how (Adams et al., 1991; Goaillard et al., 2001; Gorbunova P information is encoded within cyclic nucleotide signals. and Spitzer, 2002). However, there are limitations to l a r Until recently, investigators were limited by a lack of its use: e n real-time, single-cell sensors for cAMP and cGMP. How- (1) The reassociation of PKA subunits may be slow e ever, in the last decade several groups have developed (Rich and Karpen, 2002, and references therein). G various cAMP and cGMP sensors based on the binding (2) PKA is regulated by (high) physiological concen- f o domains of PKA, PKG, CNG channels, phosphodiester- trations of cGMP (Francis and Corbin, 1999). l ases (PDEs), and exchange factors activated by cAMP (3) Fluorescently labeled PKA is catalytically active a n (Epacs). Each of these sensors has inherent advantages (Adams et al., 1991; Goaillard et al., 2001). r u and disadvantages. In this Perspective we first outline (4) High concentrations of labeled PKA are required o J the strengths and limitations of several single-cell cyclic to overwhelm endogenous PKA (otherwise, binding of e nucleotide sensors. We then consider how information fluorescently labeled subunits to endogenous subunits h may be encoded within cyclic nucleotide signals and will distort FRET signals). PKA has a high affinity for T how current cyclic nucleotide sensors may be used to cAMP. High concentrations of high-affinity buffers will decipher the mechanisms that underlie signaling speci- severely blunt cAMP signals (Rich and Karpen, 2002). ficity. We believe that a better understanding of the These limitations hinder the utility of labeled PKA as strengths and limitations of these biosensors will pro- a cAMP sensor. However, this work sparked researchers mote a quantitative understanding of cyclic nucleotide from several groups to develop novel cAMP and cGMP signaling and help to direct the design of the next gen- probes, each with advantages and disadvantages. eration of probes. CNG channel-based cyclic nucleotide sensors. Two groups Single-cell sensors for cAMP and cGMP developed genetically modified CNG channels, which are directly opened by binding of cyclic nucleotides, as PKA-based sensors. More than twenty years ago Tsien cyclic nucleotide sensors (Trivedi and Kramer, 1998; and colleagues published the first report describing a Rich et al., 2000, 2001b). Unlike many other ion channels, novel, Förster resonance energy transfer (FRET)-based CNG channels do not desensitize in response to pro- longed cyclic nucleotide exposure (Dhallan et al., 1990; Correspondence to Thomas C. Rich: t r i c h @ s o u t h a l a b a m a . e d u Abbreviations used in this paper: AC, adenylyl cyclase; ANP, atrial na- © 2014 Rich et al. This article is distributed under the terms of an Attribution–Noncom- triuretic peptide; Epac, exchange factor activated by cAMP; FRET, Förster mercial–Share Alike–No Mirror Sites license for the first six months after the publication date (see http://www.rupress.org/terms). After six months it is available under a Creative resonance energy transfer; PDE, phosphodiesterase; pGC, particulate GC; Commons License (Attribution–Noncommercial–Share Alike 3.0 Unported license, as de- ROS, reactive oxygen species; sGC, soluble GC. scribed at http://creativecommons.org/licenses/by-nc-sa/3.0/). The Rockefeller University Press J. Gen. Physiol. Vol. 143 No. 1 17–27 www.jgp.org/cgi/doi/10.1085/jgp.201311095 17 Rich et al., 2000, 2001a), making them suitable for mon- These FRET-based probes have similar advantages itoring cyclic nucleotide levels. Open channels allow and disadvantages. cations (Na+, K+, Ca2+) to pass through the surface mem- (1) The probes can be readily targeted to different brane; thus, activation of CNG channels is readily detected intracellular locations (Terrin et al., 2006). with electrophysiological or Ca2+ imaging techniques. (2) Although the kinetics of Epac activation have not CNG channels have several other characteristics that both been carefully measured, experimental measurements lend themselves to specific experimental designs and suggest that Epac-based sensors are fast enough to faith- preclude them from others: fully reproduce most cAMP signals. (1) CNG channels have fast kinetics (90% rise time (3) In general, fluorescence and FRET measurements in <0.2 s), allowing measurement of rapid changes in are technically simple to perform (Börner et al., 2011). cyclic nucleotide levels near the plasma membrane However, calibration of these sensors in intact cells is (Rich et al., 2000). difficult because one can seldom demonstrate that both (2) CNG channels are targeted to the plasma mem- minimum and maximum fluorescence or FRET levels brane, allowing for membrane-localized cAMP measure- have been reached. ments. However, they cannot be readily used in other (4) Epac-based FRET probes have a limited range regions of the cell. and high background (Fig. 1 A), leading to a low signal- (3) CNG channel activity is readily detected at low to-noise ratio. This is further complicated by the lack expression levels. Thus, buffering of cyclic nucleotides of an appropriate comparison of commonly used FRET is low and in most cases will not substantively alter cyclic imaging and analysis approaches. Only recently have nucleotide levels (Rich and Karpen, 2002). investigators quantitatively assessed the signal-to-noise (4) CNG channels are regulated by other intracellu- ratio of different analysis approaches (Leavesley et al., lar signals including PIP3 and Ca2+ (Brady et al., 2006). 2013). And although substantive progress has been made Thus, careful controls are required to ensure that mea- in the development of FRET probes with increased range sured responses are indeed caused by changes in cyclic (Klarenbeek et al., 2011), the actual dynamic range of nucleotide levels. FRET probes has not been adequately assessed. Thus, it is (5) High kinetic resolution measurements require unclear what increments of cAMP or cGMP can actually electrophysiology; electrophysiological experiments are be discriminated. considered technically more difficult than imaging ex- (5) The (over)expression levels of Epac have not been periments. The Ca2+ permeability of CNG channels has measured; thus, the effects of cyclic nucleotide buffering been used to detect changes in cAMP by monitoring cannot be adequately assessed (discussed later). intracellular Ca2+ levels (Rich et al., 2001b). (6) Only recently have systematic, automated ap- (6) CNG channels conduct Ca2+. Thus, experiments proaches for analysis of FRET measurements been im- need to be conducted in the absence of extracellular plemented (Jalink and van Rheenen, 2009; Leavesley Ca2+ to avoid CNG channel–mediated Ca2+ regulation et al., 2013). Such approaches allow unbiased estimates of cyclic nucleotide synthesis and degradation. of FRET efficiency for all cells within a field of view, as This combination of strengths and limitations makes well as clear criteria for the exclusion of certain cells CNG channels well-suited for measuring the kinetics of from analysis. near-membrane signals (Rich and Karpen, 2002). How- ever, CNG channels are ill-suited for the study of feed- Artifacts associated with fluorescence- and FRET-based back interactions between Ca2+, phosphoinositide, and probes. Fluorescent proteins, and fluorophores in general, cyclic nucleotide signaling pathways. can be modified by a variety of biochemical processes. Some of the factors that can influence fluorescence Epac-based cyclic nucleotide sensors. Several groups emission are considered below. developed FRET-based cAMP sensors using the cAMP (1) Heterologously expressed proteins, especially binding protein Epac to measure cAMP signals in dif- targeted constructs, can accumulate at distinct subcel- ferent cellular domains (Nikolaev et al., 2004; Ponsioen lular localizations. Thus, both intermolecular and intra- et al., 2004). These sensors are comprised of catalyti- molecular FRET may occur. cally inactive Epac sandwiched between the fluorescent (2) Commonly used fluorescent proteins are differ- donor (CFP) and fluorescent acceptor (YFP; Ponsioen entially sensitive to environmental variables including et al., 2004). In the basal state (low cAMP), efficient temperature, pH, reactive oxygen species (ROS), and FRET occurs between CFP and YFP. When cAMP binds viscosity (Bernas et al., 2005; Ha and Tinnefeld, 2012). to Epac, there is a conformational change such that Both spectral and lifetime measurements are subject to FRET efficiency is reduced. There are several versions these artifacts (Suhling et al., 2002). To illustrate the of the Epac probe available (van der Krogt et al., 2008; potential problems associated with changes in the envi- Klarenbeek et al., 2011), as well as FRET-based probes ronment, we measured the effects of pH on the fluores- for the measurement of cGMP (Nikolaev et al., 2006). cence emission from CFP and YFP. We observed that 18 Real-time cyclic nucleotide sensors Figure 1. Measurements of apparent FRET. (A–C) Representative responses from cell lysates of HEK-293 cells expressing the Epac- based FRET probe (A), CFP (B) or YFP (C). Measurements were made as described previously (Leavesley et al., 2013; Rich et al., 2013). In brief, cell lysates were suspended in a stirred cuvette (4 × 106 cells in 4 ml) and excited at 430 nm, then emission intensity was mea- sured from 450 to 650 nm using a spectrofluorimeter. Data were normalized to the peak intensity. (A) The change in apparent FRET response measured at 0 and 10 µM cAMP (black and gray lines, respectively). Note the modest change in FRET associated with a saturat- ing change in cAMP. (B) Changing pH from 7 (blue line) to 6 (black line) had little effect on the measured fluorescence intensity of CFP. (C) Changing pH had a marked effect on the measured fluorescence intensity of YFP. (D and E) Simulations demonstrating the effects of changing pH on apparent FRET measurements and interpretations of the underlying cAMP signals. changing pH from 7 to 5 had little effect on the fluores- and if possible these FRET pairs should be used. How- cence emission from CFP (Fig. 1 B), but had a marked ever, the effects of other environmental variables on effect on the fluorescence emission from YFP (Fig. 1 C). these fluorophores have not been evaluated. Thus, sys- To illustrate the potential problems associated with tematic approaches for detecting and compensating for pH-mediated changes on apparent FRET measurements, environmental changes in fluorescence and apparent we modeled the effects of constant pH 7.0 or a pH FRET need to be developed. change from 7 to 6 on the apparent FRET response. In this scenario, the input cAMP signal rose from baseline Other real-time, single-cell cyclic nucleotide sensors. FRET- to a steady plateau in 30 s. We assumed that the pKa of based probes for cGMP have also been developed. CFP was 5 and the pKa of YFP was 7. The change in pH These probes use cGMP binding sites from PKG, PDE2, triggered an overall reduction in FRET (Fig. 1 D). If the PDE5, and CNG channels sandwiched between fluo- effects of pH were not appropriately accounted for, rescent donors and acceptors (Honda et al., 2005; then the interpreted cAMP signal would be transient Nikolaev et al., 2006). These probes have similar ad- rather than sustained (Fig. 1 E). Although this was a vantages and disadvantages to the FRET-based cAMP worst-case scenario for pH, other environmental factors probes described earlier. However, FRET-based cGMP may differentially alter emission from fluorescent pro- probes have not been as widely used, in part because teins. This may become particularly problematic in lo- Dostmann and colleagues have developed probes in calized regions of the cell such as the mitochondria. which the cGMP binding site of PKG was fused with Certain FRET pairs have increased photostability and circularly permutated enhanced green fluorescent pro- reduced sensitivity to pH (van der Krogt et al., 2008), tein (Nausch et al., 2008). Although these probes have Rich et al. 19 high sensitivity, high range, and likely a high dynamic (Corbin et al., 1977). The idea of localized pools of range, the experimental application of these probes is PKA activity was further supported by the discovery of A not always straightforward, and fluorescence emission kinase–anchoring proteins (AKAPs), the scaffolds that is subject to environmental changes as well as changes tether PKA to cellular targets (see Kapiloff et al. in this in cGMP. issue). However, spatial proximity of the proteins asso- Although there are limitations to the overall utility of ciated with this signaling pathway, e.g.,  -adrenergic 2 FRET-based cyclic nucleotide probes and data inter- receptors, adenylyl cyclase (AC), PKA, and L-type Ca2+ pretation is not always straightforward, these FRET channels is not enough to ensure localized responses probes are the most widely used cyclic nucleotide sen- (Feinstein et al., 2012; Saucerman et al., 2014, sors to date. The relative ease of use and the ability to and references therein). The signal itself must be spa- be targeted to different regions of a cell allow FRET- tially restricted. based probes to readily detect changes in localized cy- We and others have used compartmental models to clic nucleotide levels. describe the localization of cyclic nucleotide signals We next consider the strengths and limitations of the (Saucerman et al., 2014, and references therein). Here, different cyclic nucleotide sensors, with an emphasis on we present an updated compartmental model based upon the FRET-based sensors, in deciphering the informa- our recent estimates of near-membrane PDE activity tion content of cyclic nucleotide signals. How might information be encoded within second messenger signals? Cyclic nucleotide signals regulate dozens of cellular pro- cesses over timescales ranging from seconds to hours (Francis and Corbin, 1999, and references therein). However, it remains unclear how the information re- quired to differentially regulate multiple cellular func- tions is encoded. Several concepts have been proposed, including encoding information in the spatial distri- bution of signals (compartmentalization), in the fre- quency content of signals (frequency encoding), or in the combination of signals that are turned on or off, above or below thresholds, at a given time (digital encoding; Brooker, 1973; Hanson et al., 1994; Dolmetsch et al., 1998; Rich and Karpen, 2002; Ruf et al., 2006; Feinstein et al., 2012). These mechanisms are outlined in the following sections. Spatial encoding of information in cyclic nucleotide signals. Spatial encoding of cyclic nucleotide signals, or signal compartmentalization, requires that cAMP and cGMP levels be high enough to activate PKA, PKG, Epac, or CNG channels in one region of a cell, but not in others. The evidence for this phenomenon is straightforward: cAMP rises triggered by different hormones regulate distinct cellular processes. For example, in cardiac my- ocytes two agents that trigger similar rises in total cel- lular cAMP levels—isoproterenol and prostaglandin E 1 (PGE )—have markedly different downstream effects 1 (see Saucerman et al. in this issue). Treatment of car- diac myocytes with isoproterenol, a -adrenergic ago- nist, triggers cAMP-dependent activation of PKA and Figure 2. Schematic model of compartmentalized cAMP sig- subsequent phosphorylation of proteins associated with nals. (A) A two-compartment model with diffusional restrictions cardiac excitability; however, treatment with PGE trig- between a membrane-localized microdomain (compartment 1) 1 gers the cAMP-dependent activation of PKA, but not and the bulk cytosol (compartment 2). See text for details. (B) Cyclic AMP signals triggered by activation of AC. Cyclic phosphorylation of these proteins. This has been par- AMP levels in compartment 1 are markedly higher than those tially explained by the observation that these agents in compartment 2. (C) Total cellular cAMP accumulation over activate different pools of PKA: particulate and soluble the same time course. 20 Real-time cyclic nucleotide sensors (Fig. 2). In this model, cAMP is produced by AC in com- in detail elsewhere (Conti et al., 2014; Saucerman partment 1. The overall PDE activity is markedly lower in et al., 2014). compartment 1 than in compartment 2 (equations and Evidence presented from several groups suggests parameter values are provided in the supplemental text). that cyclic nucleotide compartmentalization is a critical Upon stepwise activation of AC, cAMP concentrations component of signaling specificity, as discussed in Conti within compartment 1 reach a plateau of >3 µM within et al. (2014), Kapiloff et al. (2014), and Saucerman 10 min. Diffusional restrictions that slow the flux of et al. (2014). The model presented here illustrates how cAMP between compartments and the greater PDE ac- diffusional barriers and variable PDE activity may allow tivity in compartment 2 blunt cytosolic cAMP accumula- distinct cyclic nucleotide signals in different subcellular tion to <0.5 µM. locations. Multiple compartments containing scaffold- The mechanisms by which this rapidly diffusible mes- localized signaling complexes likely contribute to signal- senger is localized to different regions of a cell are not ing specificity in several intracellular signaling pathways. well understood. Potential mechanisms of signal local- Localized signaling complexes may also facilitate ization have been discussed in detail elsewhere (Feinstein crosstalk between signaling pathways in a controlled et al., 2012; Conti et al. in this issue; Saucerman et al., manner, facilitating both frequency-dependent and 2014). Here we outline four mechanisms likely to con- digital signals. tribute to signal localization, and thus, specificity within Two basic approaches have been used to elucidate cAMP and cGMP signaling pathways: the mechanisms underlying the spatial segregation of (1) Colocalization of signaling proteins (e.g., re- cyclic nucleotide signals. The first approach relies on ceptors, G-proteins, AC, PDE, PKA, phosphatases, and sensors localized to a discrete region of the cell, such downstream targets) into signaling complexes is an as a CNG channel localized to the plasma membrane, essential component in ensuring signaling specificity. and on changing the location of the cyclic nucleotide However, the localization of signaling proteins by itself source. For example, Piggott et al. (2006) examined the is unlikely to account for spatial segregation of cAMP ability of cGMP produced by particulate GC (pGC) and signals because without restrictions on the spatial spread, soluble GC (sGC) to activate CNG channels in both cAMP and cGMP would rapidly diffuse throughout the HEK-293 cells and vascular smooth muscle cells. They cell (Piggott et al., 2006; Feinstein et al., 2012; Saucerman observed that in a 10-min timeframe, cGMP produced et al., 2014, and references therein). Rather, these as- by pGC readily activated CNG channels, whereas cGMP semblies ensure that signaling proteins experience the produced by sGC did not, even in the presence of PDE same signals and that PKA and PKG are poised to phos- inhibitors. In conceptually similar experiments, investi- phorylate specific downstream targets. gators have dialyzed known concentrations of cyclic (2) Physical barriers may slow the rate of cAMP diffu- nucleotides into cells and measured the activation of sion from one region of the cell to another. There is CNG channels to estimate the effective diffusion coeffi- strong evidence that ER or SR come into close apposition cient (Koutalos et al., 1995; Chen et al., 1999; Rich to the plasma membrane and limit the spatial spread of et al., 2000). We have recently used a similar approach Ca2+ and Na+ in specialized cells such as hair cells, neu- to provide evidence that PDE4 is lower in the near- rons, and cardiac myocytes. Such barriers may partially membrane compartment than in the bulk cytosol (un- restrict the spatial spread of cyclic nucleotides. In addi- published data). Although these basic approaches allow tion, the F-actin cortical rim may slow the movement of a quantitative estimate of the averaged effective diffu- cAMP due to steric hindrance and charge effects. In ad- sion coefficient throughout the cell, they cannot be used dition, F-actin networks may trigger gelation of cytosol, to further dissect the spatial spread of cyclic nucleotide dramatically lowering the local effective diffusion coef- signals because measurements of CNG channel activity ficient in the near-membrane space. For a more complete are localized to the plasma membrane. discussion see Feinstein et al. (2012) and Saucerman A second approach compares responses measured in et al. (2014) and references therein. different subcellular locations. Different studies have (3) Buffering may contribute to the slow spatial spread compared CNG channels with changes in total cellular of cAMP and cGMP signals. This may be particularly cyclic nucleotide levels (Rich et al., 2000, 2001a; Piggott important when high concentrations of PKA are local- et al., 2006), CNG channels and soluble FRET probes ized to discrete regions of the cell by AKAPs. (Rochais et al., 2006; Willoughby et al., 2006), FRET (4) PDE activity limits the spatial spread of cyclic probes based upon different cyclic nucleotide binding nucleotide signals. Mathematical models indicate that sites (Warrier et al., 2007), and FRET probes localized PDE activity is particularly effective in limiting the spa- to different subcellular compartments (Saucerman et al., tial spread of cAMP and cGMP when signals are par- 2006; Terrin et al., 2006; Blackman et al., 2011). As tially segregated by diffusional restrictions. The roles indicated earlier, the ability of fluorescence- and FRET- of PDE activity in regulating the kinetics and spatial based probes to be targeted to discrete subcellular do- spread of cyclic nucleotide signals have been discussed mains makes them well-suited for such studies. However, Rich et al. 21 data interpretation is not necessarily straightforward. small, especially compared with the abundant descrip- For example, changes in viscosity and ROS in the near- tion of Ca2+ oscillations. Thus, we are left with the ques- membrane space may differentially alter fluorescence tion: are experimental difficulties limiting our ability to emission of the donor and acceptor (e.g., CFP and detect cyclic nucleotide oscillations or are such oscilla- YFP). Similarly, dynamic changes in ROS and pH in the tions rare epiphenomena? mitochondria may confound interpretation of FRET We believe experimental limitations have impeded measurements. Also, as noted earlier, localized FRET our ability to detect cyclic nucleotide oscillations. The probes may reach concentrations in which both intra- PKA-based cAMP probes developed more than twenty molecular and intermolecular FRET occur. Thus, great years ago lacked the fast kinetics to detect rapid changes care must be taken when comparing the magnitude or in cAMP levels (Rich and Karpen, 2002). Buffering kinetics of fluorescence and FRET measurements from of cAMP signals was also problematic due to cellular different subcellular locations. We believe that until microinjection or overexpression of high probe con- more systematic approaches for monitoring the effects centrations. This was further complicated by the fact of changing cellular environments are developed, the that the fluorescently labeled PKA is an active kinase. In best approach is to use multiple probes for the same contrast, the rapid response time of CNG channels and measurement. For example, Blackman et al. (2011) low buffering capacity allow detection of high-frequency used targeted FRET probes, CNG channels, radioim- cAMP oscillations (Rich et al., 2000). However, CNG munoassays, and measurements of downstream effector channels are active proteins that allow Na+ and Ca2+ activity to elucidate the specific roles of PDE4 isoforms influx. Thus, kinetic experiments need to be conducted localized within discrete subcellular compartments. in Ca2+-free buffer to prevent CNG channel-mediated Ca2+ regulation of enzymes in the signaling system, pre- Frequency encoding of information in cyclic nucleotide venting the study of Ca2+-cAMP interactions. The re- signals. Although compartmentalization appears to make sponse time of FRET-based probes for cAMP and cGMP a major contribution to signaling specificity, other mech- is more than fast enough for the measurement of oscil- anisms, such as frequency-dependent signaling, are lations with a period (T) of 30 s, and they are not active likely to contribute to the information content in sec- enzymes. Yet few investigators have been able to measure ond messenger systems. Frequency-dependent signal- cyclic nucleotide oscillations with this probe. To better ing implies that downstream effectors respond to understand why, we have developed a realistic mathe- particular frequencies in signals. In other words, effec- matical description of the cGMP signaling pathway tors may respond differently to fast oscillations in cyclic (equations as well as parameter definitions and values nucleotide levels than to slow oscillations. Oscillations are provided in the supplemental text). in cAMP levels were proposed almost forty years ago In this model, activation of pGC increases intracel- (Brooker, 1973; Wollenberger et al., 1973). Mathematical lular cGMP accumulation. Dephosphorylation of pGC simulations demonstrated that feedback interactions causes receptor desensitization and a reduction in the between Ca2+-inhibitable ACs and Ca2+-permeable ion rate of cGMP synthesis. Cyclic GMP is hydrolyzed by channels could lead to stable oscillations in cAMP levels PDE type 5 (PDE5), which is regulated by cGMP bind- (Cooper et al., 1995). Similarly, simulations indicated ing to the noncatalytic site and phosphorylation. Inputs that Ca2+-mediated stimulation of PDE activity could to the model were analogous to sinusoidal oscillations also trigger stable oscillations in cAMP levels (Rich and in atrial natriuretic peptide (ANP) with periods (T) of Karpen, 2002). Experimental evidence that cAMP levels 30, 300, and 3,000 s. The levels of pGC activity were may oscillate was provided by Reisert and Matthews similar to those observed experimentally in response (2001), who observed oscillations in currents through to 50 nM ANP. These inputs triggered sinusoidal in- CNG channels of olfactory cilia in response to sustained tracellular cGMP accumulation with a lag that varied odorants. Later studies in  cells demonstrated cAMP with stimulus frequency (Fig. 3). The amplitude of oscillations and the importance of pulsatile cAMP cGMP oscillations increased with the period of sinusoi- production in maintaining these oscillations (Dyachok dal stimulation. A cGMP probe would require a high et al., 2006; Tian et al., 2012). Clever use of a Ca2+-stim- dynamic range (the ability to reproducibly measure ulated AC8 overexpression system elucidated the poten- small changes in cGMP) to detect the cGMP responses tial for crosstalk between cAMP and Ca2+ pathways and to ANP oscillations with periods of 30 or 300 s (Fig. 3, coordinated oscillations due to feedback interactions A and B). The cGMP FRET probes currently available between pathways (Willoughby and Cooper, 2006). lack the dynamic range to adequately sample either Zhang and colleagues provided convincing experimen- signal. Similarly, cAMP probes that use CFP and YFP as tal evidence that slow cAMP oscillations trigger PKA the fluorescent donor and acceptor lack the requisite oscillations (Ni et al., 2011). Although there is substan- sensitivity to measure analogous fluctuations in cAMP tial evidence that cyclic nucleotide oscillations occur, in intact cells. Because of their limited dynamic range, the number of studies describing these oscillations is these FRET-based sensors are only able to reproducibly 22 Real-time cyclic nucleotide sensors Figure 3. Simulations depicting intracellular cGMP accumulation in response to an oscillatory stimulus. Broken lines represent model inputs: sinusoidal stimuli with periods (T) of 30 s (A), 300 s (B), or 3,000 s (C). Solid lines represent simulations of intracel- lular cGMP accumulation. detect large cyclic nucleotide fluctuations, similar to As such, these next-generation sensors are better suited those modeled in response to slow ANP oscillations for detection of cyclic nucleotide oscillations because (Fig. 3 C). The ability of these probes to detect small they are likely able to detect changes of smaller incre- fluctuations in cyclic nucleotide levels is further limited ment in cyclic nucleotide levels, and can be expressed by their buffering capacity. To illustrate this we repeated with weaker promoters, allowing lower expression levels the simulations depicted in Fig. 3 with exogenous buf- and reduced buffering capacity. fer concentrations of 0.05, 0.5, and 5 µM (Fig. 4, black, In addition to difficulties in detecting cyclic nucleo- red, and green lines, respectively). These simulations tide oscillations inherent to the characteristics of the demonstrate that buffering by the probe may alter the first generation of genetically encoded cyclic nucleotide cyclic nucleotide signals being measured. At present, sensors, it may be difficult to detect cyclic nucleotide we cannot accurately estimate absolute fluorescent oscillations localized to discrete subcellular compart- probe concentrations in single cells; thus, the effects ments. To illustrate this, we modeled the effects of com- of cyclic nucleotide buffering capacity of these sensors partmentalization on oscillatory cGMP signals (Fig. 5). is unknown and likely varies from cell type to cell type pGC activity was constrained to a near-membrane com- (if not from cell to cell). Therefore the first generation partment (C1) that comprised 10% of the cell volume cGMP and cAMP sensors using CFP and YFP lack the (depicted in Fig. 2). Thus, in C1 enzymes, concentra- dynamic range to reproducibly discern small fluctua- tions were 10-fold higher than listed in Table S2. We tions in cyclic nucleotide levels. However, recently de- next added both PDE5 activity (maximal PDE5 activity veloped cGMP sensors by W. Dostmann and colleagues of 2 µM/min) and basal sGC activity (0.026 µM/min) in (Nausch et al., 2008) and cAMP sensors developed by the cytosolic compartment (C2). No stimulation of sGC K. Jalink and colleagues (Klarenbeek et al., 2011) have was considered. We modeled a sinusoidal stimulus to larger responses and likely increased dynamic range. trigger oscillatory cGMP fluctuations. Oscillatory cGMP Figure 4. Simulations depicting the effects of buffering by cyclic nucleotide probes. Simulations depict the cGMP responses in cells expressing 0.05 (black lines), 0.5 (red lines), and 5 µM (green lines) FRET probes (the parameter buf = 0.05, 0.5, and 5 µM, respectively). The K of the probe for cGMP was set to 1 µM. Input frequencies were 30, 300, and 3,000 s (as in Fig. 3). Note the different timescales 1/2 for each simulation. Equations as well as parameter definitions and values are given in the supplemental text. Rich et al. 23 responses occurred in C1 in response to all the stimuli signaling systems. Thus, careful titration is needed (T = 30, 300, and 3,000 s; Fig. 5, black lines). The am- to evaluate the information content within physiologi- plitude of cGMP oscillations in C1 was greater at the cal signals. faster input frequencies compared with the one-com- partment model (compare Figs. 3 and 5). Only small Frequency-dependent downstream effectors. While there cGMP responses occurred in C2 (Fig. 5, red lines). is growing evidence that cyclic nucleotide signals oscil- Assuming that soluble FRET probes would be uni- late, at least under certain experimental conditions, it is formly distributed throughout the cell (this may not be unclear how these oscillations are interpreted by down- the case, see Raymond et al., 2007), then 90% of the stream effectors. We have previously used simulations signal would be from C2, where cGMP levels are low. of CNG channels and PKA to examine how each ef- In addition, using membrane-localized cGMP probes fector would respond to cAMP oscillations (Rich and would increase the concentration of the probe in C1, Karpen, 2002). We observed that CNG channels would and in turn increase the cGMP buffering capacity of the faithfully reproduce fast cAMP oscillations (T = 2 s). probe in C1. The situation becomes more difficult if the Simulations also indicated that PKA activity would track cyclic nucleotide fluctuations occur near highly local- slow cAMP oscillations (T = 1 min). However, as the ized receptor complexes rather than in the sizeable sub- period of oscillations decreased, PKA activation would cellular compartment used in this simulation. Thus, the reach a plateau, largely because of the slow reassocia- choice of probe and experimental conditions required tion of catalytic and regulatory subunits. Zhang and for triggering and detecting cyclic nucleotide oscilla- colleagues used mathematical models of oscillatory cir- tions may be based upon trial, error, persistence, and a cuits and came to a similar conclusion (Ni et al., 2011). bit of luck. The frequency content of larger fluctuations Simulations provided by both groups indicated that the may be accurately measured, but the amplitude of fluc- level of PKA activation was proportional to the fre- tuations and the effects of cyclic nucleotide buffering quency and duty cycle of the oscillations. Thus, a par- by the probes are difficult to assess. ticular level of PKA activity could be sustained by A final thought on the measurement of cyclic nucleo- altering the frequency and duty cycle of cAMP fluctua- tide oscillations: investigators typically use high agonist tions rather than maintaining steady intracellular cAMP concentrations to trigger cyclic nucleotide production. concentrations. Similarly, it seems likely that positive This often allows for measurement of (relatively) large feedback provided by autophosphorylation of PKG would responses, and the responses are also typically more allow the frequency and duty cycle of cGMP oscillations reproducible because experiments are conducted well to dictate the level of PKG activity. above the EC for cyclic nucleotide production. Al- 50 though this may allow unraveling of potentially com- Digital encoding of information in cyclic nucleotide signals. plex interacting pathways, it may not be physiological Thus far we have discussed the strengths and weak- and, importantly, may trigger damped overshoot or nesses of current cyclic nucleotide probes when used to transient responses due to feedback regulation within measure localized and oscillating signals (spatial and Figure 5. Simulations depicting intracellular cGMP accumulation in a two-compartment model in response to an oscillatory stimulus. In this model cGMP is produced by pGC in a near-membrane compartment (C1). cGMP levels rapidly equilibrate in C1, but the flux of cGMP into the larger cytosolic compartment (C2) is slow. PDE5 activity is present in both C1 and C2. Basal sGC activity maintains a low baseline cGMP level in C2. Black lines represent cGMP concentration in C1 in response to sinusoidal stimuli oscillations with pe- riods (T) of 30 s (A), 300 s (B), or 3,000 s (C). Red lines represent cGMP accumulation in C2. Simulations assumed no cGMP sensors were present, thus the parameter buf was set to 0 µM. Note the different timescales for each simulation. Model details are given in the supplemental text. 24 Real-time cyclic nucleotide sensors frequency encoding). We briefly discuss a third mecha- acousto-optical tunable filters (AOTFs) attenuate light nism by which information can be encoded in second by 80%, and long acquisition times are required. Recent messenger signals: digital encoding. Digital encoding is advances in filter technology attenuate light by only 5%, simply the combination of a set of signals that are on or and thus allow for faster image acquisition (Favreau off. For example, consider the case with three signals: et al., 2013). Finally, it should be also noted that even cAMP, cGMP, and Ca2+. The information contained in with the advances in technology, choosing the micro- a signal in which cAMP is on (above a threshold), cGMP scope system best suited for hyperspectral imaging is is off, and Ca2+ is off is different from the information not always straightforward (Annamdevula et al., 2013), contained in the signal in which cAMP is on, cGMP is and spectral calibration may be required (Leavesley off, and Ca2+ is on. The information content is 2N, where et al., 2012). N is the number of signals (in this case, N = 3). Another aspect of interpreting cyclic nucleotide sig- Biology is seldom this simple. There may be multiple nals is unbiased data analysis. Traditional image analysis effectors, for example PKA and Epac. These effectors approaches require investigators to select regions of have different affinities for cAMP, and thus there would interest (ROIs) for automated analysis of emission inten- be three states: off (in which cAMP levels are not high sities. This has become a point of contention because enough to substantively activate PKA), low (in which manually selecting ROIs may inadvertently bias data in- cAMP is high enough to substantively activate PKA but terpretation, especially when examining signals from not Epac), and high (in which cAMP levels are high intact tissues rather than cell monolayers. Recent ad- enough to substantively activate both PKA and Epac). vances in automated image analysis overcome this limi- There may also be multiple inputs for one signaling tation, selecting ROIs based upon event detection above pathway, for example epinephrine-mediated activation noise thresholds or geometrical constraints (Francis of  -adrenergic receptors and PGE -mediated activa- et al., 2012; Leavesley et al., 2013). Thus, ROIs can be 2 2 tion of EP2 receptors. Both of these inputs trigger in- quantified in a defined manner, thresholds for signal creases in cAMP; however, the information contained response can be set based upon baseline noise levels, in each signal may be additive or just different. Thus, and signals above baseline noise can then be automati- there may be information encoded in the combination cally identified and analyzed. of signals that are on or off. Digital encoding of infor- mation is also likely to work in conjunction with spatial and frequency encoding. Thus, to decipher the infor- A bright future for decoding the information content mation encoded in cyclic nucleotide signals we must in cyclic nucleotide signals simultaneously monitor other intracellular signals such In the last fifteen years a great deal of progress has been as Ca2+. State-of-the-art measurements are primarily fluo- made in measurement of cyclic nucleotide signals. Here rescence- and FRET-based probes. How can we quanti- we have discussed several cyclic nucleotide sensors, high- tatively measure signals from several fluorescence- and lighted their strengths and weaknesses, and summa- FRET-based probes in cells and tissues? rized three mechanisms for encoding information within intracellular signals. And although all protein- Hyperspectral imaging and analysis of cyclic nucleotide based fluorescent probes have limitations in terms of signals. One promising approach to measure several intra- dynamic range and buffering of the signals, the latest cellular signals simultaneously is hyperspectral imaging generation of probes has increased range, and hyper- and analysis. Hyperspectral imaging approaches were spectral imaging approaches can be used to improve developed by the Department of Defense and NASA the signal-to-noise ratio of fluorescence and FRET mea- to solve remote sensing problems associated with sat- surements. These advances should allow detection and ellite imaging. These approaches have been used to quantification of the localized cyclic nucleotide dynam- study biological systems and second messenger signal- ics in both single cell and tissue preparations. Hyper- ing (Leavesley et al., 2013; Rich et al., 2013; Woehler, spectral imaging also allows measurement of multiple 2013). Hyperspectral approaches allow for measure- probes to simultaneously track fluctuations of several ments of the fluorescence emission spectrum of a sam- intracellular signals. Automated image analysis provides ple, allowing accurate detection and quantification of a set of tools that allow investigators to analyze data in the abundance of unique spectral signatures, e.g., the an unbiased fashion. These data can be used to validate emission spectra of CFP, YFP, and Hoechst 33342 and test mathematical descriptions of signaling systems (a nuclear stain), within cells and tissue. As such, hyper- in cells and tissues. In turn, mathematical models can spectral approaches are well suited for simultaneously be used to direct our experimental design. In the last monitoring fluorescence emission from several probes decade, the technologies required to decode the infor- (Leavesley et al., 2013; Rich et al., 2013). The tradeoff mation content of intracellular signals have developed when using commercially available spectral implemen- to the point that we can begin to decipher the flux of tations is reduced temporal resolution. For example, information between cells and their environment. Rich et al. 25 This Perspectives series includes articles by Karpen, channel from olfactory neurons. Nature. 347:184–187. http://dx Kapiloff et al., Conti et al., and Saucerman et al. .doi.org/10.1038/347184a0 Dolmetsch, R.E., K. Xu, and R.S. Lewis. 1998. Calcium oscillations Equations describing compartmentalized and oscillatory cyclic increase the efficiency and specificity of gene expression. Nature. nucleotide signals are provided. Table S1 provides the parameter 392:933–936. http://dx.doi.org/10.1038/31960 values, descriptions, initial conditions, and references from which Dyachok, O., Y. Isakov, J. Sågetorp, and A. Tengholm. 2006. parameter values were obtained or estimated for compartmen- Oscillations of cyclic AMP in hormone-stimulated insulin-secreting tal models describing subcellular cyclic nucleotide distribution. beta-cells. Nature. 439:349–352. http://dx.doi.org/10.1038/ Table S2 shows provides parameter values, descriptions, initial nature04410 conditions, and references from which the parameter values were Favreau, P., C. Hernandez, A.S. Lindsey, D.F. Alvarez, T. Rich, P. obtained or estimated for models describing oscillatory cGMP Prabhat, and S.J. Leavesley. 2013. Thin-film tunable optical filters signals. Online supplemental material is available at http://www for hyperspectral microscopy. J. Biomed. Opt. 19:011017. http:// .jgp.org/cgi/content/full/jgp.201311095/DC1. dx.doi.org/10.1117/1.JBO.19.1.011017 Feinstein, W.P., B. Zhu, S.J. Leavesley, S.L. Sayner, and T.C. Rich. This work was supported by National Institutes of Health grant 2012. Assessment of cellular mechanisms contributing to cAMP P01H06629. compartmentalization in pulmonary microvascular endothelial cells. Am. J. Physiol. Cell Physiol. 302:C839–C852. http://dx.doi.org/ Olaf S. Andersen served as editor. 10.1152/ajpcell.00361.2011 Francis, S.H., and J.D. Corbin. 1999. Cyclic nucleotide-dependent REFERENCES protein kinases: intracellular receptors for cAMP and cGMP action. Crit. Rev. Clin. Lab. Sci. 36:275–328. http://dx.doi.org/10.1080/ Adams, S.R., A.T. Harootunian, Y.J. Buechler, S.S. Taylor, and R.Y. 10408369991239213 Tsien. 1991. Fluorescence ratio imaging of cyclic AMP in single Francis, M., X. Qian, C. Charbel, J. Ledoux, J.C. Parker, and M.S. cells. Nature. 349:694–697. http://dx.doi.org/10.1038/349694a0 Taylor. 2012. Automated region of interest analysis of dynamic Annamdevula, N.S., B. Sweat, P. Favreau, A.S. Lindsey, D.F. Alvarez, Ca²+ signals in image sequences. Am. J. Physiol. Cell Physiol. T.C. Rich, and S.J. Leavesley. 2013. An approach for character- izing and comparing hyperspectral microscopy systems. Sensors 303:C236–C243. http://dx.doi.org/10.1152/ajpcell.00016.2012 (Basel). 13:9267–9293. http://dx.doi.org/10.3390/s130709267 Goaillard, J.M., P.V. Vincent, and R. Fischmeister. 2001. Simul- Bernas, T., J.P. Robinson, E.K. Asem, and B. Rajwa. 2005. Loss taneous measurements of intracellular cAMP and L-type Ca2+ of image quality in photobleaching during microscopic imaging current in single frog ventricular myocytes. J. Physiol. 530:79–91. of fluorescent probes bound to chromatin. J. Biomed. Opt. 10:064015. http://dx.doi.org/10.1111/j.1469-7793.2001.0079m.x http://dx.doi.org/10.1117/1.2136313 Gorbunova, Y.V., and N.C. Spitzer. 2002. Dynamic interactions of Blackman, B.E., K. Horner, J. Heidmann, D. Wang, W. Richter, T.C. cyclic AMP transients and spontaneous Ca(2+) spikes. Nature. Rich, and M. Conti. 2011. PDE4D and PDE4B function in dis- 418:93–96. http://dx.doi.org/10.1038/nature00835 tinct subcellular compartments in mouse embryonic fibroblasts. Ha, T., and P. Tinnefeld. 2012. Photophysics of fluorescent probes J. Biol. Chem. 286:12590–12601. http://dx.doi.org/10.1074/jbc for single-molecule biophysics and super-resolution imaging. .M110.203604 Annu. Rev. Phys. Chem. 63:595–617. http://dx.doi.org/10.1146/ Börner, S., F. Schwede, A. Schlipp, F. Berisha, D. Calebiro, M.J. annurev-physchem-032210-103340 Lohse, and V.O. Nikolaev. 2011. FRET measurements of intra- Hanson, P.I., T. Meyer, L. Stryer, and H. Schulman. 1994. Dual role cellular cAMP concentrations and cAMP analog permeability in of calmodulin in autophosphorylation of multifunctional CaM intact cells. Nat. Protoc. 6:427–438. http://dx.doi.org/10.1038/ kinase may underlie decoding of calcium signals. Neuron. 12:943– nprot.2010.198 956. http://dx.doi.org/10.1016/0896-6273(94)90306-9 Brady, J.D., E.D. Rich, J.R. Martens, J.W. Karpen, M.D. Varnum, and Honda, A., C.L. Sawyer, S.M. Cawley, and W.R. Dostmann. 2005. R.L. Brown. 2006. Interplay between PIP3 and calmodulin regu- Cygnets: in vivo characterization of novel cGMP indicators and lation of olfactory cyclic nucleotide-gated channels. Proc. Natl. in vivo imaging of intracellular cGMP. Methods Mol. Biol. 307: Acad. Sci. USA. 103:15635–15640. http://dx.doi.org/10.1073/pnas 27–43. .0603344103 Jalink, K., and J. van Rheenen. 2009. Filter FRET: Quantitative Brooker, G. 1973. Oscillation of cyclic adenosine monophosphate imaging of sensitized emission. In Laboratory Techniques in concentration during the myocardial contraction cycle. Science. Biochemistry and Molecular Biology, Vol. 33. T.W.J. Gadella, edi- 182:933–934. http://dx.doi.org/10.1126/science.182.4115.933 tor. Elsevier, Oxford. 289–349. Chen, C.H., T. Nakamura, and Y. Koutalos. 1999. Cyclic AMP dif- Kapiloff, M.S., M. Rigatti, and K.L. Dodge-Kafka. 2014. Perspectives fusion coefficient in frog olfactory cilia. Biophys. J. 76:2861–2867. on: Cyclic nucleotide microdomains and signaling specificity: http://dx.doi.org/10.1016/S0006-3495(99)77440-0 Architectural and functional roles of A kinase–anchoring pro- Conti, M., D. Mika, and W. Richter. 2014. Perspectives on: Cyclic teins in cAMP microdomains. J. Gen. Physiol. 143:9–15. nucleotide microdomains and signaling specificity: Cyclic AMP Klarenbeek, J.B., J. Goedhart, M.A. Hink, T.W. Gadella, and K. compartments and signaling specificity: Role of cyclic nucleotide Jalink. 2011. A mTurquoise-based cAMP sensor for both FLIM phosphodiesterases. J. Gen. Physiol. 143:29–38. and ratiometric read-out has improved dynamic range. PLoS ONE. Cooper, D.M.F., N. Mons, and J.W. Karpen. 1995. Adenylyl cyclases 6:e19170. http://dx.doi.org/10.1371/journal.pone.0019170 and the interaction between calcium and cAMP signalling. Nature. Koutalos, Y., K. Nakatani, and K.-W. Yau. 1995. Cyclic GMP diffu- 374:421–424. http://dx.doi.org/10.1038/374421a0 sion coefficient in rod photoreceptor outer segments. Biophys. J. Corbin, J.D., P.H. Sugden, T.M. Lincoln, and S.L. Keely. 1977. 68:373–382. http://dx.doi.org/10.1016/S0006-3495(95)80198-0 Compartmentalization of adenosine 3:5-monophosphate and Leavesley, S.J., N. Annamdevula, J. Boni, S. Stocker, K. Grant, B. adenosine 3:5-monophosphate-dependent protein kinase in Troyanovsky, T.C. Rich, and D.F. Alvarez. 2012. Hyperspectral heart tissue. J. Biol. Chem. 252:3854–3861. imaging microscopy for identification and quantitative analysis of Dhallan, R.S., K.-W. Yau, K.A. Schrader, and R.R. Reed. 1990. Primary fluorescently-labeled cells in highly autofluorescent tissue. J. Bio­ structure and functional expression of a cyclic nucleotide-activated photonics. 5:67–84. http://dx.doi.org/10.1002/jbio.201100066 26 Real-time cyclic nucleotide sensors