Recent evidence demonstrates that low-frequency oscillations of intracellular calcium on timescales of seconds to minutes drive distinct aspects of neuronal development, but the mechanisms by which these calcium transients are coupled to signaling cascades are not well understood. Here we test the hypothesis that spontaneous electrical activity activates protein kinase A (PKA). We use live-cell indicators to observe spontaneous and evoked changes in cAMP levels and PKA activity in developing retinal neurons. Expression of cAMP and PKA indicators in neonatal rat retinal explants reveals spontaneous oscillations in PKA activity that are temporally correlated with spontaneous depolarizations associated with retinal waves. In response to short applications of forskolin, dopamine, or high-potassium concentration, we image an increase in cAMP levels and PKA activity, indicating that this second-messenger pathway can be activated quickly by neural activity. Depolarization-evoked increases in PKA activity were blocked by the removal of extracellular calcium, indicating that they are mediated by a calcium-dependent mechanism. These findings demonstrate for the first time that spontaneous activity in developing circuits is correlated with activation of the cAMP/PKA pathway and that PKA activity is turned on and off on the timescale of tens of seconds. These results show a link between neural activity and an intracellular biochemical cascade associated with plasticity, axon guidance, and neural differentiation.
- activity-dependent development
- cAMP-dependent protein kinase
- retinal ganglion cell
- spontaneous activity
Neuronal activity is linked to long-term changes in cells, such as activity-dependent expression of genes, by activation of second-messenger cascades (West et al., 2002). Although several such cascades have been identified, there is a short list of second messengers, such as calcium, cAMP, cGMP, DAG, and IP3, that mediate these effects. Therefore, the activation of specific cascades is not likely to be determined by overall concentrations of second messengers but rather by subcellular spatial localization and dynamic changes in their concentrations. Oscillations have been detected in both calcium and cAMP concentrations, and there is growing evidence that these oscillations are “tuned” to drive particular cellular events (Zaccolo and Pozzan, 2003; Spitzer et al., 2004; Fields et al., 2005; Dyachok et al., 2006).
The establishment and refinement of retinal projection patterns occurs at an early stage of development, before the development of vision (for review, see Grubb and Thompson, 2004; Ruthazer and Cline, 2004; Torborg and Feller, 2005). During this time, retinal ganglion cells (RGCs) exhibit spontaneous bursts of action potentials that periodically propagate across the developing ganglion cell layer, an activity pattern termed retinal waves (Meister et al., 1991; Wong et al., 1993). The cAMP/protein kinase A (PKA) pathway has been implicated in the establishment and normal refinement of RGC axonal projections. Knock-out mice lacking calcium-activated adenylate cyclase 1, an enzyme that produces cAMP, exhibit defects in retinotopic and eye-specific refinement (Ravary et al., 2003; Plas et al., 2004). Knockdown of the transcription factor cAMP response element-binding protein, a known target of PKA, leads to a reduction in eye-specific segregation (Pham et al., 2001). These effects of the cAMP/PKA pathway on refinement of retinal projections may be mediated by activation of this pathway in RGCs, perhaps by regulating the response of RGC growth cones to guidance molecules (Nicol et al., 2006) or modulating release probability at RGC axon terminals (Lu et al., 2006). Although spontaneous patterned activity and the cAMP/PKA pathway are important for the establishment of retinal projections, how retinal waves affect the cAMP/PKA second-messenger cascades either in RGCs or their targets remains unknown.
Here we used three genetically encoded fluorescence-based indicators to monitor the cAMP/PKA pathway in both dissociated retinal neurons and the intact retina. The indicators are as follows: (1) a PKA-based cAMP sensor (CS) (Zaccolo et al., 2002), (2) ICUE2 (indicator of cAMP using Epac), which is modified from ICUE1 (DiPilato et al., 2004), and (3) A-kinase activity reporter AKAR2.2 (Zhang et al., 2001, 2005), which undergoes changes in its fluorescent properties in response to phosphorylation by PKA. To demonstrate the utility of CS, ICUE2, and AKAR2.2, we compared their responses to direct activation of adenylate cyclase with activation of G-protein-coupled receptors by dopamine and to depolarization. We also detected spontaneous oscillations in cAMP levels and PKA activity in retinal explants. These oscillations were blocked by agents that block retinal waves and exhibited a temporal correlation to spontaneous depolarizations associated with retinal waves. Together, these findings indicate that retinal waves induce oscillations in cAMP levels and PKA activity in RGCs.
Materials and Methods
Dissociated neurons and glia from postnatal day 1 (P1) rat retinas were plated on coverslips, with a subset of cultures placed on superior colliculus cultures from littermates plated the day before as described previously (Colicos et al., 2004). Cells were plated on surfaces treated with 10 μg/ml poly-d-lysine (70 kDa; Sigma, St. Louis, MO) overnight, followed by 2 μg/ml laminin for 2 h at 37°C. Retinal neurons were plated at a density of 2 × 104 cells/cm2 in serum-free Neurobasal medium [serum free culture media (SFCM)] supplemented with B27 (Invitrogen, Carlsbad, CA), 0.6% glucose, 2 mm glutamine, 10 mm HEPES, 1 mm Na-pyruvate, 50 μg/ml penicillin G, 50 U/ml streptomycin, 2.5 μg/ml insulin, 6 μm forskolin, 10 ng/ml CNTF, and 50 ng/ml BDNF. Forskolin was added to the medium to ensure survival of RGCs (Meyer-Franke et al., 1995). It is not known whether forskolin enhances cell survival by elevating baseline levels of activity or by influencing periodic activation of the cAMP/PKA pathway. To avoid complications with the interpretation of cAMP imaging experiments, cultures were removed from forskolin-containing media for 10 h before experiments. Control experiments indicated stable baselines of cAMP after this treatment.
Transfection of dissociated neurons was performed by electroporation using an exponentially decaying pulse. Acutely isolated whole retinas from P0–P2 rats were mounted on filter paper precoated with 100 μg/μl poly-d-lysine and 50 μg/μl laminin. They were incubated in dissection solution (Colicos et al., 2004) containing plasmid DNA (concentration of ∼0.2 μg/μl) for 10 min, electroporated with two square wave pulses with an electric field strength of 6.75 V/mm, allowed to recover for 10 min in dissection solution, and cultured in SFCM-A (same media as above, except using Neurobasal-A). Overnight before imaging, electroporated retinas were transferred to SFCM-A without forskolin. For recordings, explants are placed ganglion cell layer up in a temperature-controlled chamber (30°C; Warner Instruments, Hamden, CT) mounted on the stage of an upright microscope (Axioskop; Zeiss, Oberkochen, Germany) and perfused continuously with artificial CSF (ACSF) (in mm): 119 NaCl, 2.5 KCl, 1.3 MgCl2, 1.0 K2HPO4, 2.5 CaCl2, 26.2 NaHCO3, and 11 d-glucose (bubbled with 95% O2/5% CO2). Explant health was assayed using whole-cell recording to measure resting potentials, action potentials, and spontaneous depolarizations (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). Imaging experiments without independent evidence of retinal waves in explants were discarded from analysis.
Live imaging of fluorescence resonance energy transfer (FRET) indicators was performed on an upright Zeiss Axioskop 2, using a 60× LUMPLFLW water immersion objective (Olympus Optical, Tokyo, Japan). Individual FRET channel detection was accomplished by using a dual-view image splitter with appropriate yellow and cyan channel filters. Images were captured on a PhotoMetrics (Huntington Beach, CA) CoolSnap HQ CCD camera and analyzed in MetaMorph version 6.3 (Universal Imaging Corporation, West Chester, PA). Background fluorescence was subtracted from both channels, and cyan fluorescent protein (CFP) bleed-through into the yellow fluorescent protein (YFP) channel was corrected (FYFP = FFRET − 0.51 × FCFP) (Gordon et al., 1998; van Rheenen et al., 2004).
We conducted acceptor photobleaching experiments to measure FRET efficiency. FRET efficiency is defined as E = 1 − (FDA/FD), where FDA is the intensity of the donor (in our case, CFP) in the presence of the acceptor (YFP), and FD is the intensity of the donor in the absence of the acceptor. In our experiments, YFP was only partially photobleached, so to compute FD, we used the equation FD = (F − FDA)/x + FDA, where F is the CFP intensity after partial photobleaching of YFP, and x = (FA-pre − FA-post)/FA-pre, the fraction of the acceptor that is photobleached.
FRET ratios are calculated as R = (FYFP/FCFP). The magnitude of drug or depolarization-induced change was represented by the change in FRET ratio. Changes in FRET ratio, ΔR, were computed by subtracting the value of the FRET ratio averaged over five images before manipulation from the value of the FRET ratio averaged over five images around the maximum response within 40 s after the drug application or depolarization. We observed oscillations in baseline lasting one to three frames; therefore, an average of five frames (corresponding to 10 s) gave an accurate measure of the baseline without decrementing the amplitude of the induced changes. Note, for the cAMP sensor or ICUE2, increases in cAMP levels lead to a decrease in FRET efficiency. For ease of comparison, ratio changes for all three indicators are represented as the absolute value of ΔR = (FYFP/FCFP)control − (FYFP/FCFP)manipulation. Live Ca2+ imaging was performed with the same equipment as for FRET imaging. Fura-2 AM loading was performed as described previously (Bansal et al., 2000).
All pharmacological agents were purchased from Sigma. External solution for dissociated neurons contained the following (in mm): 5 KCl, 123 NaCl, 3 CaCl2, 2 MgCl2, 10 glucose, and 10 HEPES, pH adjusted with NaOH to 7.3. High-potassium solution was similar, with 100 mm KCl substituted for an equivalent amount of NaCl. Short applications of dopamine or high-K+ external solution were delivered through a glass electrophysiology electrode (∼2 μm tip size, located <20 μm from transfected neuron) using a PV830 Pneumatic PicoPump at 2–10 psi, 50 ms pulse duration at 4 Hz to RGCs in retinal explants. The duration of dopamine and potassium applications was altered by increasing the number of pulses. For experiments with 0 external calcium, ACSF was modified to replace 2.5 mm CaCl2 with 2 mm EGTA and 1.7 mm MgCl2.
Whole-cell voltage-clamp recordings from retinal explants were made using an Axopatch 200B amplifier and pClamp6 software (Molecular Devices, Palo Alto, CA). The intracellular solution consisted of the following (in mm): 98.3 potassium-gluconate, 1.7 KCl, 0.6 EGTA, 5 MgCl2, 2 Na2-ATP, 0.3 GTP, and 40 HEPES, pH 7.25 with KOH. The calculated ECl was −60 mV in ACSF. Whole-cell current-clamp recordings were used to compare spontaneous depolarizations recorded from explants and acutely isolated retinas (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). The duration of spontaneous depolarization was defined as the time between membrane potential deviations from baseline that were more than two times the root mean square (which was ∼0.5 mV at 5 kHz sampling).
Cell-attached recordings were used to depolarize RGCs transfected with AKAR2.2. We were unable to use whole-cell current-clamp recordings because dialyzing the contents of the cell in whole-cell configuration led to significantly decreased sensitivity of the indicator. Gigaohm seals were obtained with a series resistance of ∼500 MΩ. Depolarizing pulses of 150 pA, lasting 1–3 s, caused local increases in membrane potential of ∼150 mV.
Expression of FRET indicators in dissociated retinal neurons
To monitor cAMP levels and PKA activity in living mammalian neurons, we used three different genetically encoded indicators dependent on FRET. The first, which we will refer to as the CS, uses an intermolecular FRET paradigm in which CFP is fused to a regulatory subunit of PKA and YFP is fused to the catalytic subunit of PKA (Zaccolo et al., 2002). An increase in cAMP concentration results in the binding of two cAMP molecules to the regulatory subunit of this fusion protein and a consequent release of the catalytic subunit. This results in a measurable decrease in FRET efficiency. The second indicator, ICUE2, consists of a fusion of CFP, a truncated form of Epac1 including a cAMP binding site, and YFP [modified from ICUE1 (described by DiPilato et al., 2004)]. Binding of cAMP to the site on Epac1 leads to a disruption of basal FRET interactions. The third indicator, AKAR2.2 (Zhang et al., 2005), is derived from AKAR2 (Zhang et al., 2001). Both AKAR2 and AKAR2.2 consist of a fusion of CFP, a phosphothreonine-binding domain (FHA1), a PKA consensus phosphorylation target, and the YFP variant citrine (Griesbeck et al., 2001). When PKA is activated, the catalytic subunit of PKA phosphorylates the target in AKAR2.2, causing a conformational change that increases FRET efficiency. AKAR2.2 differs from AKAR2 (Zhang et al., 2005) in that it includes A206K substitutions within both the CFP and YFP moieties to eliminate their tendency to dimerize (Zacharias et al., 2002). These mutations also greatly accelerate the phosphatase-mediated decrease in FRET efficiency that follows a PKA-induced increase in FRET efficiency (K. Kain, J. Zhang, L. Goldfinger, R. Tsien, M. Ginsberg, unpublished observations).
To determine whether the indicators were forming functional FRET pairs, a partial acceptor photobleach assay was performed for CS and AKAR2.2. For both indicators, FCFP significantly increased after YFP was selectively photobleached using 515 nm light, indicating that, before stimulation, there was a baseline FRET interaction for both indicators. We measured FRET efficiencies of 0.11± 0.09 (n = 9) and 0.18 ± 0.06 (average ± SD; n = 10) for CS and AKAR2.2, respectively (see Materials and Methods).
To characterize the physiological responses of CS, ICUE2, and AKAR2.2, the indicators were expressed in dissociated retinal neurons. Retinas from P0–P1 rats were dissociated and electroporated with plasmids encoding one of the three indicators. The neurons were plated and cultured in vitro for 2–4 d using conditions that promote survival of all classes of retinal neurons (Meyer-Franke et al., 1995; Goldberg et al., 2002a; Colicos et al., 2004). The indicators were strongly expressed throughout the cell (Fig. 1A–C). For comparisons across indicators, all ratio changes are represented as the absolute value of ΔR = (FYFP/FCFP)control − (FYFP/FCFP)manipulation.
We next determined the maximal response of the indicators to direct elevation of intracellular levels of cAMP. Neurons expressing CS, ICUE2, or AKAR2.2 exhibited large responses to bath application of both the adenylate cyclase activator forskolin (10 μm) and the phosphodiesterase inhibitor IBMX (100 μm), a manipulation that strongly stimulates the cAMP/PKA pathway (Fig. 1D–G) (ΔRCS = 0.099 ± 0.048, n = 10; ΔRICUE = 0.346 ± 0.065, n = 8; ΔRAKAR = 0.118 ± 0.040, n = 10). The response to forskolin alone was significantly smaller than the response to the combination of forskolin and IBMX for both AKAR2.2 (ΔRAKAR = 0.044 ± 0.014; n = 8; p < 0.01) and CS (ΔRCS = 0.012 ± 0.005; n = 8; p < 0.01), demonstrating that the indicator response was correlated with the strength of activation of the cAMP/PKA pathway. We found no correlation between the starting ratio and the size of the FRET ratio change, indicating a lack of saturation of the indicator while in the basal state or with forskolin alone (data not shown).
Evoked changes in cAMP levels and PKA activity in retinal explants
To determine the effectiveness of indicators in monitoring activation of the cAMP/PKA pathway in the intact retina, we introduced the plasmids into neonatal rat retinal explants using electroporation (see Materials and Methods). After 20 h, indicators were strongly expressed in RGCs, readily identified by the appearance of axons at the inner retinal surface (an example of an AKAR2.2-expressing RGC is shown in Fig. 2A).
First, we assessed the changes in PKA activity in response to brief application of forskolin (50 μm). Short applications of forskolin caused a transient increase in the AKAR2.2 FRET ratio (Fig. 2C) (ΔR = 0.037 ± 0.012; n = 6 neurons). The amplitude of the AKAR2.2 FRET ratio change did not change in calcium-free solutions (Fig. 2C) (ΔR = 0.036 ± 0.019; n = 6 neurons; p = 0.929). Hence, forskolin induced PKA activity in a calcium-independent manner.
Second, we investigated whether AKAR2.2 could detect changes in PKA activity attributable to activation of G-protein-coupled receptors. Dopamine D1 receptor immunolabeling is present throughout the inner plexiform layer in neonate rats (Koulen, 1999), and D1 receptor, but not D2 receptor, mRNA is present in the ganglion cell layer of late prenatal rats (Schambra et al., 1994). D1-like receptors are positively coupled to adenylate cyclase; therefore, dopamine application in neonate rats is predicted to increase PKA activity. Indeed, brief application of dopamine (150 μm) caused an increase in the AKAR2.2 FRET ratio in a majority of RGCs (Fig. 2B) (ΔR = 0.034 ± 0.027; n = 6 of 8 neurons).
Third, we investigated whether the cAMP/PKA pathway in RGCs is activated by depolarization in retinal explants. RGCs express calcium-dependent adenylate cyclases (Abdel-Majid et al., 2002; Nicol et al., 2005), and therefore depolarization is likely to elevate cAMP/PKA levels (Ferguson and Storm, 2004). Short applications of external solution containing high concentrations of KCl (105 mm) to neurons expressing ICUE2 or AKAR2.2 induced changes in their FRET ratios (Fig. 2C) (ΔRICUE = 0.011 ± 0.003, n = 6 neurons; ΔRAKAR = 0.010 ± 0.006, n = 12 neurons), approximately one-third the amplitude of those seen in response to forskolin or dopamine applications (Fig. 2C). Depolarization-induced increases in AKAR2.2 FRET ratio were blocked in the presence of calcium-free external solution, implying that calcium influx is necessary for depolarization-induced activation of the cAMP/PKA pathway in RGCs. Depolarization-induced increases in the AKAR2.2 FRET ratio were significantly inhibited by the presence of H89 (N-[2-(p-bromocinnamylamino)ethyl]-5-isoquinolinesulfonamide 2HCl) (25 μm), a specific PKA inhibitor (Zhang et al., 2001) (Fig. 2D) (ΔRH89/ΔRcontrol = 0.25 ± 0.10; n = 7; p < 0.0001), confirming that AKAR2.2 was reporting increases in PKA activity.
Spontaneous oscillations in cAMP concentration and PKA activity are induced by and temporally correlated with retinal waves
We demonstrated that depolarization transiently activates the cAMP/PKA pathway. We next investigated the kinetics of the second-messenger pathway in response to spontaneous depolarizations associated with retinal waves. As described above, we introduced ICUE2 and AKAR2.2 plasmids into neonatal retinal explants using electroporation. Calcium imaging and whole-cell current-clamp recordings from individual RGCs in retinal explants revealed that RGCs exhibited correlated increases in intracellular calcium and periodic spontaneous depolarizations, consistent with the persistence of retinal waves after 48 h in culture (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). However, there was a decrease in the duration of depolarization in explants compared with acutely isolated retina (explant, duration of 1.7 ± 0.6 s, n = 8; acutely isolated retina, duration of 3.9 ± 0.5 s, n = 8; p < 0.001). These shorter depolarizations were reflected in reduced amplitude calcium transients. Although >95% of cells in the ganglion cell layer exhibited spontaneous calcium transients in both acutely isolated and explanted retinas, there was a stark difference in amplitude in most cells (supplemental Fig. 1C, available at www.jneurosci.org as supplemental material). Sixty percent of cells in explant retinas exhibited average fractional changes in fluorescence of <4%, whereas only 5% of cells in acutely isolated retinas had calcium transients that small.
Despite this decrease in depolarization and calcium influx, we detected spontaneous increases in FRET ratio in a subset of RGCs expressing either ICUE2 or AKAR2.2. [ΔRICUE = 0.015 ± 0.011, n = 7 of 31 RGCs (Fig. 3A); ΔRAKAR = 0.025 ± 0.015, n = 12 of 96 RGCs (Fig. 3B)]. The magnitude of FRET ratio changes for both ICUE2 and AKAR2.2 were similar in amplitude and kinetics to those induced by acute application of potassium in RGCs (rise time measured trough to peak: ICUE2, τspontaneous = 13.6 ± 3.7 s, n = 7; AKAR2.2: τspontaneous = 20.5 ± 4.6 s, n = 12; τevoked = 18.2 ± 10.3 s, n = 14; p > 0.5) (Fig. 3D). These similarities suggest that retinal waves may lead to cAMP and PKA activity transients via depolarization.
Several observations indicate that spontaneous transients in PKA activity were induced by retinal waves. First, the spontaneous AKAR2.2 FRET ratio changes occurred with a periodicity similar to that of spontaneous depolarizations associated with retinal waves (peak-to-peak interval 2.05 ± 0.94 min; n = 12; p = 0.173). Second, spontaneous AKAR2.2 FRET ratio changes were blocked by bath application of the nicotinic acetylcholine receptor antagonist dihydro-β-erythroidine (10–20 μm), which reliably blocks retinal waves (n = 5) (Fig. 3C). Third, in one instance in which we observed two transfected RGCs in the same field of view, we observed simultaneous AKAR2.2 FRET ratio changes (data not shown).
To determine whether the spontaneous PKA transients exhibit a consistent temporal relationship with retinal waves, we conducted simultaneous AKAR2.2 imaging and whole-cell voltage-clamp recordings from a nearby cell to detect the compound synaptic currents associated with retinal waves (n = 38 waves from 4 cells) (Fig. 4A). We computed the “wave-triggered” AKAR2.2 FRET ratio change by averaging the segment of the FRET ratio for AKAR2.2 from 5 s preceding to 40 s after the compound PSC. We found that, on average, the AKAR2.2 FRET ratio increased after a barrage of synaptic current (Fig. 4B). For comparison, we computed the wave-triggered fura-2 response using simultaneous whole-cell recording and calcium imaging. In contrast to the AKAR2.2 FRET ratio change, the calcium increase in nearby cells occurred during the compound synaptic event (Fig. 4C,D) (n = 21 waves from 5 cells). This is consistent with previous measurements that indicate that increases in intracellular calcium associated with retinal waves are driven by influx through voltage-gated channels during spontaneous depolarizations (Penn et al., 1998). Combined with the observation that blockade of retinal waves also blocks AKAR2.2 FRET ratio changes, the increase in FRET ratio after a retinal wave is consistent with the hypothesis that retinal waves induce PKA activity transients.
Depolarization-induced activation of cAMP/PKA pathway requires 3-s-long depolarizations and calcium influx
Why do we detect spontaneous PKA transients in only a subset of RGCs? In transfected RGCs that did not exhibit spontaneous oscillations in the FRET ratio, we measured substantial ratio changes in response to bath application of forskolin (10 μm; ΔR = 0.091 ± 0.058; n = 5 of 5 neurons), demonstrating that the absence of oscillations is not attributable to a lack of functional AKAR2.2. In addition, we found that non-oscillating cells had detectable FRET ratio changes in response to strong depolarization either via short applications of high-potassium solutions or cell-attached recordings (Fig. 5). Based on these observations, we hypothesized that AKAR2.2 detected oscillations only in the RGCs that exhibited the strongest depolarizations during waves.
To quantify the magnitude of depolarization necessary to induce a PKA activity transient, we varied the duration of depolarization induced by short applications of a high-potassium solution (Fig. 5A, inset). Evoked FRET ratio changes and rise times are similar to those attributable to spontaneous PKA oscillations. We found that longer-duration depolarizations more reliably induced detectable changes in FRET ratio of AKAR2.2 (n = 11 RGCs) (Fig. 5B). To verify that these changes in FRET ratio were attributable to direct depolarization of RGCs, we reproduced these findings using a cell-attached patch electrode (Fig. 5C,D) (n = 10 RGCs). Using both stimulation methods, we found a significant increase in detectable AKAR2.2 FRET ratio changes in response to depolarizations >2 s (Fig. 5E).
Based on these observations, we hypothesize that the low percentage of RGCs with spontaneous PKA activity transients in explants is attributable to the shorter spontaneous depolarization associated with retinal waves (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). Hence, we would predict that, in acutely isolated retina, or more importantly in vivo, in which recent experiments report spiking bursts lasting 3–4 s (Hanganu et al., 2006), a larger percentage of RGCs exhibit PKA transients.
Depolarization-induced PKA activity transients require Ca2+ influx
Depolarization-induced activation of second-messenger pathways are proposed to function via calcium-dependent adenylate cyclases (Ferguson and Storm, 2004). Calcium imaging of RGCs loaded with fura-2 AM revealed that longer-duration depolarizations led to larger increases in intracellular calcium concentration (Fig. 5F). Hence, longer depolarizations increase the likelihood of PKA activity transients because they lead to a larger increase in intracellular calcium. In addition, we found that removal of calcium from ACSF abolished the depolarization-induced AKAR2.2 FRET ratio changes observed in RGCs (Fig. 5D) (n = 17 events from 4 RGCs and 4 retinas). Together, these findings indicate that PKA activity transients in RGCs are induced by depolarization in a calcium-dependent manner.
We demonstrate that live indicators for cAMP levels and PKA activity can reliably measure activation of this second-messenger pathway in neurons in the intact circuitry of the developing retina. In addition, our findings for the first time link spontaneous neural activity in developing networks with temporal oscillations in kinase activity. This temporal coding of second-messenger pathway activation may be critical for the onset and timing of downstream developmental processes that are driven by spontaneous neural activity.
Comparison of cAMP and PKA indicators
We found that the genetically encoded indicators ICUE2 and AKAR2.2 reliably reported both evoked (Figs. 1, 2, 5) and spontaneous (Figs. 3, 4) changes in cAMP concentration and PKA activity in retinal neurons. These measurable changes in FRET efficiency occurred as responses to either direct [via stimulation of adenylate cyclase (Figs. 1B, 2A)] or indirect [via activation of a G-protein-coupled receptor or depolarization (Fig. 2B–D)] activation of the cAMP/PKA pathway. One notable difference between the three indicators was the kinetics of the response to strong stimulation (Fig. 1D–F). The CS ratio change was much slower than that of ICUE2 and AKAR2.2. One possible explanation is that cAMP may activate PKA kinase activity before completely dissociating the regulatory and catalytic subunits (Yang et al., 1995; Nikolaev et al., 2004). For this reason, the CS FRET ratio changes occur only after PKA dissociates, whereas AKAR2.2 ratio changes occur faster. Similarly, ICUE2 FRET ratio changes will occur as cAMP concentration increases independent of PKA dissociation. In summary, we conclude that, in our system, ICUE2 and AKAR2.2 function as more sensitive reporters for monitoring brief activation of the cAMP/PKA pathway. These findings indicate that ICUE2 and AKAR2.2 can be used to observe activity of the cAMP/PKA pathway in vivo, for example in developing neocortex, in which strong calcium transients have been observed in newborn mice (Adelsberger et al., 2005). In addition, immobilized AKAR2.2 can be used to monitor subcellular compartments of PKA activity, such as those observed in growth cones (Terman and Kolodkin, 2004; Kain, Zhang, Goldfinger, Tsien, Ginsberg, unpublished observations).
Potential function of periodic PKA activation in developing circuits
AKAR2.2 FRET ratio changes were induced in nearly all RGCs with sufficient depolarization (Fig. 5). In our explants that had shorter spontaneous depolarizations than observed in acutely isolated retina (supplemental Fig. 1, available at www.jneurosci.org as supplemental material), retinal waves induce transient activation of the cAMP/PKA pathway (Figs. 3, 4). This evidence suggests that most RGCs in vivo exhibit periodic activation of PKA.
There are several examples that demonstrate that temporal coding of second-messenger pathway activation is critical for influencing development (Berridge et al., 2000; Spitzer et al., 2004). For example, in developing Xenopus spinal neurons, the frequency of calcium transients is critical in establishing neurotransmitter phenotype (Gu and Spitzer, 1995; Borodinsky et al., 2004). Similarly, in the ventricular zone in neocortex (Owens et al., 2000) and retina (Pearson et al., 2002, 2004), calcium transients are important for proliferation of ventricular zone cells and the subsequent fate differentiation of postmitotic neurons. Later in development, after distinct neuronal classes are specified but before motoneurons innervate their targets, spontaneous activity in the developing spinal cord is frequency tuned to modulate the expression of proteins critical for pathfinding of motoneurons (Hanson and Landmesser, 2004). The frequency of calcium transients in growth cones has also been implicated in mediating interactions of the growth cone with the local environment (Henley and Poo, 2004). Periodic activation of isolated RGCs and the spontaneous activity in retinal explants profoundly increase the rate of axonal outgrowth in response to BDNF (Goldberg et al., 2002b). This enhancement of BDNF-induced outgrowth is blocked by adenylate cyclase inhibitors (Goldberg et al., 2002b), indicating that the endogenous pattern of retinal activity regulates axon outgrowth via this second-messenger cascade, possibly by increasing the surface expression of neurotrophin receptors (Meyer-Franke et al., 1998; Nagappan and Lu, 2005).
Whether periodic activation of RGCs is required for normal development of the visual system is not yet determined. Although it is well established that spontaneous retinal activity patterns are critical for the refinement of retinal projections to central targets, the precise role of correlated activity patterns is the subject of intense debate. In a knock-out mouse that has uncorrelated retinal ganglion cell firing instead of retinal waves (Bansal et al., 2000; McLaughlin et al., 2003), it has been shown recently that several features of both retinocollicular projections (McLaughlin et al., 2003; Chandrasekaran et al., 2005; Mrsic-Flogel et al., 2005) and retinogeniculate projections (Rossi et al., 2001; Muir-Robinson et al., 2002; Grubb et al., 2003) were altered. In contrast, both pharmacological manipulations of activity patterns in ferret and a different knock-out mouse with altered retinal patterns develop normal refinement of retinal projections (Stellwagen and Shatz, 2002; Huberman et al., 2003; Torborg et al., 2005). By comparing the retinal firing patterns that drive or permit segregation with those that do not, we identified several candidate activity features that may be essential for this refinement process (Torborg et al., 2005).
The primary hypothesis that arose from these studies is that RGCs must fire high-frequency bursts with a slow periodicity for normal eye-specific refinement. Such an activity pattern would strongly activate both calcium-dependent and cAMP-dependent second-messenger cascades. However, a critical component of the argument that retinal waves are instructive for driving eye-specific segregation or retinotopic refinement is that the plasticity process that reads out this instructive signal must be able to distinguish firing from neighboring cells versus firing from distant cells. Therefore, the elongated PKA transients we observed in RGCs are not likely to provide an instructive signal for driving map refinement but rather may function in a cell-autonomous manner. However, the plasticity mechanisms that convert this firing pattern into long-term changes in cellular morphology remain to be determined.
Another potential role for PKA oscillations may be in the generation of retinal waves. Retinal waves are generated by a network of starburst amacrine cells (Zheng et al., 2004, 2006), and the frequency of retinal waves is determined by changes in cAMP levels in starburst amacrine cells (Zheng et al., 2006). Indeed, bath application of the PKA inhibitor Rp-cAMP blocks retinal waves (Stellwagen et al., 1999). The data presented here do not address the role of PKA activity in the generation of retinal waves because we recorded PKA activity exclusively in RGCs, which “read out” the activity of the starburst network (Feller et al., 1997; Zheng et al., 2006) but are not thought to participate in the generation of retinal waves. Whether cAMP levels or PKA activity spontaneously oscillate in developing starburst amacrine cells has yet to be determined.
Models of the complex interplay between calcium and cAMP levels demonstrate that changes in either the level or kinetics of intracellular cAMP or calcium concentrations can lead to dramatic changes in the emergent behavior of the intracellular network (Gorbunova and Spitzer, 2002). The implications of these models is that neural activity patterns can activate different spatiotemporal patterns of second-messenger networks, which in turn may be uniquely tuned to drive specific long-term cellular changes. For example, expression of BDNF, a neurotrophic factor critical for normal arborization of retinal ganglion cell dendrites and axons as well synapse formation (Cohen-Cory and Lom, 2004), is upregulated by increased intracellular calcium and cAMP concentration but not by stimulation of the cAMP pathway alone (Fukuchi et al., 2005). Similarly, surface expression of TrkB may also be regulated by neuronal activity and cAMP levels (Nagappan and Lu, 2005). This suggests that the pattern of intracellular calcium and cAMP oscillations caused by periodic depolarizations may be critical for determining mature morphology in the developing visual system.
In conclusion, we report data that support the hypothesis that retinal waves induce oscillations in cAMP concentration and PKA activity of RGCs. Visualization of second-messenger cascade kinetics is an important step in elucidating the link between activity and the mechanisms by which it influences development of neural circuits.
This work was supported by National Institutes of Health Grants EY13528 and EY016417 (M.B.F.), NS27177 (R.Y.T.), and DK73368 (J.Z.), the McKnight Scholars Fund (M.B.F.), Telethon Italy Grants TCP00089 and GGP05113, the Fondazione Compagnia di San Paolo, and Human Frontier Science Program Organization Grant RGP0001/2005-C (M.Z.). L.M.D. was supported by National Research Service Award F31 DK074381 from National Institutes of Health. We thank Nick Spitzer and Aaron Blankenship for helpful comments on this manuscript.
- Correspondence should be addressed to Marla B. Feller, Neurobiology Section 0357, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0357.
- Abdel-Majid et al., 2002.↵
- Adelsberger et al., 2005.↵
- Bansal et al., 2000.↵
- Berridge et al., 2000.↵
- Borodinsky et al., 2004.↵
- Chandrasekaran et al., 2005.↵
- Cohen-Cory and Lom, 2004.↵
- Colicos et al., 2004.↵
- DiPilato et al., 2004.↵
- Dyachok et al., 2006.↵
- Feller et al., 1997.↵
- Ferguson and Storm, 2004.↵
- Fields et al., 2005.↵
- Fukuchi et al., 2005.↵
- Goldberg et al., 2002a.↵
- Goldberg et al., 2002b.↵
- Gorbunova and Spitzer, 2002.↵
- Gordon et al., 1998.↵
- Griesbeck et al., 2001.↵
- Grubb and Thompson, 2004.↵
- Grubb et al., 2003.↵
- Gu and Spitzer, 1995.↵
- Hanganu et al., 2006.↵
- Hanson and Landmesser, 2004.↵
- Henley and Poo, 2004.↵
- Huberman et al., 2003.↵
- Koulen, 1999.↵
- Lu et al., 2006.↵
- McLaughlin et al., 2003.↵
- Meister et al., 1991.↵
- Meyer-Franke et al., 1995.↵
- Meyer-Franke et al., 1998.↵
- Mrsic-Flogel et al., 2005.↵
- Muir-Robinson et al., 2002.↵
- Nagappan and Lu, 2005.↵
- Nicol et al., 2005.↵
- Nicol et al., 2006.↵
- Nikolaev et al., 2004.↵
- Owens et al., 2000.↵
- Pearson et al., 2002.↵
- Pearson et al., 2004.↵
- Penn et al., 1998.↵
- Pham et al., 2001.↵
- Plas et al., 2004.↵
- Ravary et al., 2003.↵
- Rossi et al., 2001.↵
- Ruthazer and Cline, 2004.↵
- Schambra et al., 1994.↵
- Spitzer et al., 2004.↵
- Stellwagen and Shatz, 2002.↵
- Stellwagen et al., 1999.↵
- Terman and Kolodkin, 2004.↵
- Torborg and Feller, 2005.↵
- Torborg et al., 2005.↵
- van Rheenen et al., 2004.↵
- West et al., 2002.↵
- Wong et al., 1993.↵
- Yang et al., 1995.↵
- Zaccolo and Pozzan, 2003.↵
- Zaccolo et al., 2002.↵
- Zacharias et al., 2002.↵
- Zhang et al., 2001.↵
- Zhang et al., 2005.↵
- Zheng et al., 2006.↵
- Zheng et al., 2004.↵