Specific patterns of neural impulses regulate genes controlling nervous system development and plasticity, but it is not known how intracellular signaling cascades and transcriptional activation mechanisms can regulate specific genes in response to specific patterns of action potentials. Studies using electrical stimulation of mouse dorsal root ganglion neurons in culture show that the temporal dynamics of intracellular signaling pathways are an important factor. Expression of c-fos varied inversely with the interval between repeated bursts of action potentials. Transcription was not dependent on a large or sustained increase in intracellular Ca2+, and high Ca2+ levels separated by long interburst intervals (5 min) produced minimal increases in c-fos expression. Levels of the transcription factor cAMP-responsive element binding protein (CREB), phosphorylated at Ser-133, increased rapidly in response to brief action potential stimulation but remained at high levels several minutes after an action potential burst. These kinetics limited the fidelity with which P-CREB could follow different patterns of action potentials, and P-CREB levels were not well correlated withc-fos expression. The extracellular-regulated kinase (ERK) mitogen-activated protein kinases (MAPK) also were stimulated by action potentials of appropriate temporal patterns. Bursts of action potentials separated by long intervals (5 min) did not activate MAPK effectively, but they did increase CREB phosphorylation. This was a consequence of the more rapid dephosphorylation of MAPK in comparison to CREB. High expression of c-fos was dependent on the combined activation of the MAPK pathway and phosphorylation of CREB. These observations show that temporal features of action potentials (and associated Ca2+transients) regulate expression of neuronal genes by activating specific intracellular signaling pathways with appropriate temporal dynamics.
How specificity is maintained between stimulus and transcription of specific genes is a fundamental problem in cell biology. In neurobiology an additional complexity arises, because information in the nervous system is coded in the temporal pattern of impulse activity. Although there is considerable information on the multiple signal transduction pathways leading from membrane depolarization to gene transcription, it is not fully understood how these reactions operate as a system to extract and transmit information from temporally varying stimulation. This is an important question, considering the fundamental role of action potential-dependent gene regulation in brain function and nervous system development. Neuronal electrical activity has an important influence on development and postnatal shaping of neuronal connections (Shatz, 1990; Fields and Nelson, 1992; Goodman and Shatz, 1993; Gu and Spitzer, 1995). It is well established that specific patterns of action potentials can regulate the expression of specific genes, including those coding neural cell adhesion molecules (Itoh et al., 1995), ion channels (Li et al., 1966; Garcia et al., 1994), neurotransmitters (Hodaie et al., 1995), and immediate early (IE) genes (Sheng et al., 1993; Worley et al., 1993).
The concentration of second messengers and differences in activation thresholds for calcium-dependent signaling reactions are important factors in controlling cellular responses to stimulation (Clapham, 1995; Ghosh and Greenberg, 1995). In addition, subcellular heterogeneity in the concentration dynamics of second messengers can contribute to stimulus–response specificity (Bading and Greenberg, 1993; Bootman and Berridge, 1995). The objective of the present study was to investigate whether differences in the temporal dynamics of the second messenger Ca2+ and integration by downstream signaling elements could contribute to specific activation of genes in response to specific patterns of neuronal firing.
Patterns of stimulation were delivered that resembled normal bursting activity in utero (Meister et al., 1991; Fitzgerald and Fulton, 1992). Stimulus parameters were developed that would produce similar or countervailing differences in the magnitude of calcium increase (residual, peak, and time-integrated Ca2+) but that differed in specific temporal features (e.g., the interval of time between repeated bursts). In contrast to chemical stimulation of neurons in vitro, which drives cellular signaling systems to steady-state equilibrium, electrical stimulation in multicompartment cell cultures can model the important dynamics of intracellular signaling systems driven by impulse activity. Dorsal root ganglion (DRG) neurons do not form synapses and lack spontaneous impulse activity in culture. They respond to brief electrical stimulation with a single action potential, enabling precise regulation of the pattern of impulse activity (Fields et al., 1992). Electrical stimulation is the most direct means of depolarizing the membrane to stimulate gene expression.
The relation among neural impulse pattern, intracellular calcium transients, activation of MAP kinase (mitogen-activated protein kinase; MAPK) and the transcription factor cAMP-responsive element binding protein (CREB), and induction of c-fos mRNA and Fos-β-galactosidase were studied in mouse primary DRG neurons. The results provide insight into how information encoded in the temporal pattern of action potentials is transmitted and integrated within the neuron to control the expression of a gene implicated in adaptive responses in neurons.
MATERIALS AND METHODS
Multicompartment cell culture and stimulation.Multicompartment chambers were made of Teflon and attached to collagen-coated 35 mm culture dishes as described (Fields et al., 1992). DRG neurons were dissociated from 13.5 d mouse fetuses, and 0.25 × 106 cells were plated into the two side compartments. Culture medium contained 5% horse serum and 50 ng/ml nerve growth factor, as described previously (Fields et al., 1990;Sheng et al., 1993). Mitosis of non-neuronal cells was inhibited by adding 13 μg/ml fluoro-2-deoxyuridine 1–2 d after the initial cell plating. Cultures were fed twice a week with half changes of fresh medium. Cultures were used for experiments 3–4 weeks after plating to allow time for most DRG neurons in the side compartments to extend axons under the barrier between the side and central compartments. Axons traversing the barrier were stimulated through platinum electrodes on opposite sides of this barrier. Twenty-four hours before stimulation, cultures were washed three times in serum-free medium lacking added NGF to minimize the serum and neurotrophin response of the c-fos gene.
Stimulation parameters and electrophysiological responses to stimulation have been reported previously for DRG neurons in these multicompartment chambers (Fields et al., 1990, 1992; Sheng et al., 1993). DRG neurons in this preparation respond to 5 V, 200 μsec biphasic pulsed stimulation with a single action potential. This permits precise control of the frequency and pattern of action potential activity. These cells follow stimulation reliably and indefinitely at rates up to 3 Hz, and they follow 10 Hz stimulation for several seconds or 30 Hz for short periods (Fields et al., 1990; Sheng et al., 1993). In addition to testing various stimulus frequencies from 0.1 Hz to 10 Hz and different durations of stimulation (up to 30 min), we applied four intermittent pulse train stimuli. These consisted of 1.8, 3.6, 5.4, and 9 sec duration 10 Hz bursts repeated at 1, 2, 3, and 5 min intervals, respectively. Pulse train stimulation was applied for 30 min to deliver a total of 540 action potentials for each different stimulus pattern.
RNA preparation and RT-PCR analysis. Total RNA from each culture dish was extracted with 400 μl of TRIzol reagent (Life Technologies, Gaithersburg, MD). RNA was reverse-transcribed into cDNA with Superscript II (Life Technologies) and c-fos-specific or random hexamers. Approximately 1/10 fraction was added to the PCR reaction and amplified in a solution containing 10 mmTris-HCl, pH 8.3, 50 mm KCl, 0.2 mm individual dNTPs, 1.5 mm MgCl2, 1 μmeach primer, and 1 U of Taq DNA polymerase (Perkin-Elmer Cetus, Norwalk, CT) in a final volume of 50 μl. PCR conditions included a 1 min 94°C denaturation, 1 min 60°C annealing, and 1 min 72° primer extension for 20–28 cycles.
The oligonucleotide sequence of the forward and reverse primers corresponds to nucleotides 1392–1414 and 2232–2254 of the mousec-fos gene (Van Beveren et al., 1983). The primers flank introns 2 and 3, yielding a 341 bp RT-PCR product. Neuron-specific enolase (NSE) was coamplified as an internal target for normalization of cell count, RNA extraction, and PCR amplification. NSE primers GGA CAG CAA GAA AGA GGC TCC (downstream) and CTC TAC CAG GAC TTT GTC CGC (upstream) were based on the published sequence from rat brain (Sakimura et al., 1985). Preliminary experiments were performed to determine the range of nonsaturated amplification, using serial dilution of cDNA and variable cycle numbers. NSE cDNA was amplified colinearly with that of c-fos under the conditions described above. NSE is specific for neurons, and it is not regulated by electrical stimulation. Constant expression of NSE in our system was confirmed by comparing NSE mRNA levels with those of GADPH. The PCR products were electrophoresed in 6% polyacrylamide or 3% NuSieve 3:1 (FMC Bioproducts, Rockland, ME) gels and stained with 0.5 μg/ml ethidium bromide. After electrophoresis a picture of the gel was taken, and the intensity of the bands was quantified by densitometry (Image 1, Universal Imaging, West Chester, PA). The results were normalized with respect to the intensity of the NSE band and expressed as a percentage change from nonstimulated cultures. The specificity of the PCR product was judged by the expected size of the band in the gel and by hybridization to a specific probe for c-fos (1919–1942) under stringent conditions (Sheng et al., 1993).
Intracellular calcium. The intracellular calcium concentration was measured by ratiometric fluorescence video microscopy (Image 1, Universal Imaging), using the calcium-sensitive dye fura-2/AM as described previously (Fields et al., 1993). Neurons were washed in a balanced salt solution containing 2% bovine serum albumin and 5 μm fura-2/AM (Molecular Probes, Eugene, OR) for 10 min at 37°C. Cells were incubated at room temperature for 15 min and then washed and maintained at room temperature for 30 min to allow for hydrolysis of the acetoxymethyl ester from the dye. Electrically evoked changes in calcium concentration were measured in the cell bodies of neurons plated in the side compartments of the multicompartment chambers. Images were obtained with a 40× Zeiss lens on an inverted Zeiss microscope (Oberkochen, Germany) equipped with a quartz collector, using an image intensifier and a video camera. Image acquisition was triggered from the electrical stimulus delivered to the cell cultures to synchronize measurement of responses to electrical stimuli of different patterns. Excitation was switched between 340 and 380 nm with a filter wheel (Sutter Instruments, Novato, CA), and the fluorescent images were acquired and ratioed at rates of ∼1/sec. In experiments requiring the highest possible acquisition speed with this system, fluorescence intensity was measured at video rate in response to excitation at one wavelength, and an identical electrical stimulus was repeated after a 5 min rest to allow measurements to be acquired by using the other excitation wavelength. Then the fluorescent intensities from the two wavelengths were ratioed.
Confocal microscopy in single line and X-Y scan mode were used to study subcellular heterogeneity of electrically evoked calcium transients in DRG neurons stimulated in multicompartment chambers. Measurements were performed on a Bio-Rad 1024 visible/UV confocal microscope (Hercules, CA) and a Nikon 40× oil immersion objective on a Nikon inverted microscope (Tokyo, Japan). Quantitative calcium measurements were made with ratiometric measurements at a fluorescence intensity at 480 and 405 nm in DRG neurons loaded with Indo-1/AM and excited by an argon ion laser. These were confirmed by single wavelength measurements with Fluo-3/AM excited by a krypton/argon laser.
CREB phosphorylation. Cultures of DRG neurons were stimulated electrically or with 60 mm KCl and fixed immediately after stimulation with 1% paraformaldehyde for 20 min at room temperature. Cells were permeabilized with 0.2% Triton X-100 for 2 min, and endogenous peroxidase was blocked with 3% H2O2 for 5 min, followed by 3% normal goat serum (NGS) for 30 min. Cultures were incubated for 4 hr with antibody against unphosphorylated CREB or CREB phosphorylated at serine-133 (Ginty et al., 1993) (Upstate Biotechnology, Lake Placid, NY, and New England Biolabs, Beverly, MA) at a concentration of 0.25 μg/ml in PBS containing 1% NGS. After washing, the cultures were incubated for 30 min with biotinylated goat anti-rabbit IgG and localized by the ABC method (Vector Laboratories, Burlingame, CA) according to the manufacturer’s protocol. The relative intensity of the nuclear stain was quantified by using imaging densitometry on a video microscope (Image 1, Universal Imaging).
CREB phosphorylation also was analyzed by Western immunoblotting as described previously (Ginty et al., 1993). Cells lysates made in boiling sample buffer (125 mm Tris-HCl, pH 6.8, 4% SDS, 10% β-mercaptoethanol, 20% glycerol, and 0.04% bromphenol blue), were boiled for 5 min and resolved in 10% SDS polyacrylamide minigels and electrotransferred to polyvinylidene fluoride membranes (Immobilon-P, Millipore, Bedford, MA). Membranes were blocked with 5% dry milk in TTBS (10 mm Tris-HCl, pH 7.5, 150 mm NaCl, 0.1% Tween-20) for 1 hr at room temperature, washed, and incubated in the CREB or P-CREB antibodies (1:2500) overnight at 4°C. Immunocomplexes were visualized by enhanced chemiluminescence reagent (Amersham, Arlington Heights, IL) according to the manufacturer’s protocol.
MAP kinase activation. The in vivo MAP kinase activation was measured in cultured DRG neurons after four different action potential patterns were elicited. The activation of the extracellular-regulated kinase (ERK) MAP kinases was determined by Western immunoblotting as described above, except that polyclonal antibodies raised against dual phosphorylated/active epitope in ERK1 and ERK2 (12.5 ng/ml) (Anti-ACTIVE MAPK, Promega, Madison, WI) and total ERK1/ERK2 MAP kinases (5 ng/ml) (New England Biolabs) were used. The immunocomplexes were detected with either DAB (Vector Laboratories) or enhanced chemiluminescence reagent (Amersham) according to the manufacturers’ protocols. Digitized images of the immunoblots or autoradiograms were used for densitometric measurements with the Intelligent Quantifier software (Bio Image, Ann Arbor, MI). Relative enzyme activation was determined by normalization of the density of images from phosphorylated enzyme with that of the total ERK1 MAP kinase from parallel experiments in the same sample.
Data analysis and experimental design. Calcium responses to electrical stimulation were measured in multiple cells in several dishes from multiple dissections (sample size given with Results). The calcium responses over time for all cells at each stimulus frequency were pooled and plotted as the mean response at every time point ± the SEM. Statistical analysis was performed by ANOVA with data analysis computer software (Minitab, State College, PA). Calcium responses were measured in response to different frequencies of stimulation and in response to bursts of 10 Hz stimulation for durations up to 9 sec. Long-term calcium recordings also were performed to monitor changes in intracellular calcium during the 30 min stimulation period with the four different pulse train stimulus patterns used in studies of c-fos expression. The sampling rate was varied systematically by computer program to provide higher temporal resolution during stimulus bursts than during the intervals between bursts. The activation and recovery kinetics of calcium influx in response to different durations of 10 Hz stimulation were modeled with nonlinear regression software (Table Curve, Jandel Scientific, San Rafael, CA). The equation of best fit was selected on the basis of high regression coefficient and most uniform distribution of residuals.
The relative intensity of nuclear staining after immunocytochemistry, using antibodies against CREB or P-CREB, was compared by imaging densitometry in multiple cells from multiple dishes and multiple dissections (sample size reported with Results). Statistical comparisons were made by ANOVA on differences in nuclear staining intensity after 30 min of stimulation with the four different stimulus patterns used in experiments on c-fos expression or after 10 min of depolarization with 60 mm KCl. Differences in staining were also compared after 30 min of stimulation at 1 and 10 Hz and after fixation after 10 sec and 1, 5, and 10 min of 10 Hz stimulation. Fixation was begun immediately after stimulation was stopped. Dephosphorylation of CREB at Ser-133 was studied by stimulating cultures at 10 Hz for 5 min and fixing cultures immediately or 1, 5, 10, and 25 min after terminating the stimulus. Images were acquired from at least 15 randomly chosen fields in each culture with a Nuvicon video camera and digitized by computer for storage and display. The mean intensity of staining was quantified in the nucleus of every neuron in each field. All cultures from a given experiment were analyzed together to maintain uniformity. All values were normalized to the mean intensity of nuclear staining in unstimulated cultures to allow for pooling replicate experiments [arbitrary optical units = (((nuclear staining density/average staining density of control nuclei) × 100) −100)]. The results were presented as mean ± SEM, and statistical comparisons were evaluated by ANOVA or two-samplet test. The images were digitized by an eight-bit digitizer, which yields an intensity scale from 0–255 (pure white to pure black). The average intensity of unstimulated nuclei was ∼100; therefore, the maximal possible increase in staining intensity = 155%. All experimental designs were balanced to include each stimulus condition and unstimulated cultures to allow normalizing intensities relative to unstimulated controls.
Calcium transients evoked by trains of action potentials
Calcium is a critical second messenger mediating intracellular signaling and regulating gene expression in neurons (Ghosh and Greenberg, 1995; Fields, 1996; Ginty, 1997), but there is little quantitative data on the relation between calcium transients and gene expression. It is not yet clear how changes in resting calcium concentration, amplitude of calcium transient, and dynamics of calcium increase and recovery after stimulation relate to transcriptional activation of c-fos and other IE genes. Intracellular calcium concentration was measured in response to specific patterns of action potentials to test the hypothesis that the temporal dynamics of intracellular calcium transients could be an important factor in regulating c-fos mRNA levels, apart from differences in the concentration of intracellular calcium (peak levels, residual calcium, and total integral calcium).
Calcium imaging, using the fluorescent calcium probe fura-2, showed that calcium transients in response to action potentials exhibited rapid on rates, temporal summation, and relatively slower recovery (Fig. 1). Quantitative analysis of electrically evoked calcium transients was performed to allow prediction of changes in intracellular calcium in response to a wide range of possible stimulus patterns. The immediate objective was to define a set of stimulus patterns that would produce similar or countervailing increases in the amount of cytoplasmic calcium but that differed in temporal parameters, such as the interval between repeated bursts. The kinetics of increase in calcium produced by 1–90 action potentials (delivered at a frequency of 10 Hz) were well fit by nonlinear regression to an equation of the form: Equation 1where y = the calcium concentration,a = the prestimulus calcium concentration,b = the amplitude of the calcium increase, andt = the duration of stimulation at 10 Hz; cis the constant ranging from 1.02 for a single action potential to 0.459 for 90 action potentials (Fig. 1 H–N, Table1). Even a single action potential produced a small (∼20 nm) but measurable increase in intracellular calcium concentration (Fig. 1 G,N).
Next it was necessary to determine the rate at which the calcium concentration returns to normal after a stimulus burst, to predict how calcium levels are affected by repeated bursts of impulses. The recovery function followed nonlinear kinetics that differed for long and short stimulus bursts. For stimulus bursts of <54 impulses (Fig.1 Q–U), the kinetics of recovery were well fit by a single exponential equation of the form: Equation 2
with a time constant in the range of 1.7–3.6 sec.
For longer duration bursts, the kinetics of recovery deviated from the single exponential decay function by having a more sustained increase than expected at ∼10 sec after the stimulus was terminated (Fig. 1,insets to O,P). This plateau response (Fig. 1 O,P, Table 1) was better fit by a double exponential equation of the form: Equation 3Similar plateau responses after intense stimulation have been reported previously (Thayer and Miller, 1990; Bowie et al., 1994).
High concentrations of intracellular calcium in these neurons would be reached within the first few seconds of stimulation at a frequency of 10 Hz, with relatively smaller differences in peak calcium concentration produced after longer duration stimuli. After the stimulus was terminated, calcium recovered to near prestimulus levels within seconds after 1–18 impulse bursts (at 10 Hz) (Fig.1 R–U), but tens of seconds were required to recover from 36 to 90 impulse bursts (Fig. 1 O–Q). These kinetics were confirmed by measuring mean peak calcium responses to electrical stimulation in a larger number of neurons (n= 42), which were stimulated with 10 Hz bursts of seven different durations ranging from 0.1 to 9 sec (Fig.2 A).
Previous research has suggested that, rather than correlating with the peak levels of calcium during a given stimulus burst, gene expression might be related more closely to the net increase in intracellular calcium produced during the entire period of stimulation (30 min; Sheng et al., 1993). The total calcium–time integral, representing the net increase in intracellular calcium during the period of calcium influx and recovery from stimulation, increased as a linear function of the duration of an individual stimulus burst of 1.8–9 sec duration (Fig.2 B).
Neural impulse activity and c-fos expression
c-fos is implicated in a wide range of nervous system processes, including the conversion of short-term stimuli into long-term changes in neurons (Sheng and Greenberg, 1990; Morgan and Curran, 1991; Hughes and Dragunow, 1995), response to brain injury, sensory stimulation, activation of neurotransmitter receptors, stress, circadian rhythms, and long-term changes in synaptic strength (e.g., long-term potentiation) (Hughes and Dragunow, 1995). Temporal features of action potential activity may be relevant in regulating certain functional response of c-fos activity.
Expression of c-fos differed in response to different patterns of action potential stimulation. The amount ofc-fos mRNA increased as an exponential function of the stimulus frequency over the range of 0.1–10 Hz, confirming previous research (Sheng et al., 1993). One action potential every 10 sec (0.1 Hz) was near threshold stimulation frequency for elevatingc-fos expression. Expression increased from 145% of unstimulated values in response to 0.1 Hz stimulation to ∼600% of control after stimulation for 30 min at frequencies of 3 Hz or higher (data not shown).
Previous work has shown that c-fosexpression increases with increasing duration of stimulation in DRG neurons (Sheng et al., 1993). However, natural firing patterns are modeled more closely by pulse train stimulation than by constant frequency stimulation (Meister et al., 1991; Fitzgerald and Fulton, 1992). To investigate effects of bursts of action potentials that produced similar changes in intracellular calcium, but which differed in the amount of time between bursts, we used bursts of 1.8 sec or longer (Fig. 2 A). To avoid saturating stimulus conditions, we delivered the number of action potentials that produced ∼50% maximal c-fos expression (540 action potentials) in 30 min of constant frequency stimulation but grouped into four different stimulus patterns. These four pulsed patterns consisted of repeated bursts of 1.8, 3.6, 5.4, and 9 sec duration (at 10 Hz) separated by 1, 2, 3, and 5 min intervals, respectively (Fig.3 A).
Levels of c-fos mRNA differed significantly after these different patterns of 540 action potentials (Fig. 3 B). An inverse correlation was evident between c-fos expression and the interval between successive bursts of action potentials. This correlation held despite countervailing differences in the duration of bursts. Maximal expression was produced by short (1.8 sec) bursts (10 Hz) repeated at 1 min intervals (18/1), but bursts repeated at 5 min intervals produced minimal increase in c-fos mRNA, although the individual bursts were five times longer (9 sec). Shorter bursts of stimulation repeated at 5 min intervals do not produce a paradoxical increase in c-fos expression (data not shown). Thus, the interval between bursts of action potentials is a critical parameter in regulating c-fos mRNA levels. The same pattern of expression in response to these different stimulus patterns was observed at the protein level (Fig. 3 C) in DRG neurons cultured from transgenic mice containing the human fos/lacZ fusion gene (Schilling et al., 1991).
Calcium signaling and c-fos expression
The concentration dynamics of intracellular Ca2+ might encode differences in action potential firing pattern to produce the differences in c-fosexpression seen after the various stimuli. This was not the case for the set of pulsed stimuli tested in this study. Instead,c-fos expression correlated better with the temporal dynamics, i.e., expression declined with increasing intervals of time between periods of calcium influx (Fig.4). Peak calcium level, total calcium–time integral, and residual calcium produced by the various pulse trains of action potentials were not correlated withc-fos expression. For example, long duration bursts (9 sec) were not effective in stimulating c-fos expression when separated by 5 min intervals (Fig. 4 L) although the stimulus raised calcium to the highest level (735 nm) (Figs. 2 A, 4 J). Shorter stimulus bursts, e.g., 1.8 sec, were effective in activatingc-fos gene expression if repeated at more frequent intervals (Fig. 4 C), although short bursts raised calcium the least (485 nm) (Figs. 2 A,4 A). It is clear that a sustained elevation in intracellular calcium is not necessary for stimulating c-fosexpression, because the kinetics of recovery are sufficiently rapid to restore calcium to near prestimulus levels well within the 1–5 min period separating the bursts (Fig. 4 A,D,G,J). Differences in total calcium load to the neuron (calcium–time integral) were relatively minor (not statistically significant;p = 0.69) for the four stimulus patterns (Fig.2 C). For example, six bursts of 90 action potentials would produce 38.3 μm calcium/min, and 30 bursts of 18 action potentials would produce 35.9 μm calcium/min over the 30 min period of stimulation. Thus, this variable is not likely to have a controlling influence on differences in gene expression produced by the four different stimulus patterns.
The possibility that subcellular heterogeneity in calcium signaling in the nucleus versus the cytoplasm may account for the differences in c-fos expression (Hardingham et al., 1997) is not supported by experimental evidence in DRG neurons. Ratiometric fluorescence confocal microscopy with the calcium indicator Indo-1 was used to compare calcium responses in the nucleus and cytoplasm of DRG neurons. Cytoplasmic and nuclear calcium transients were nearly identical in response to membrane depolarization by trains of action potentials (Fig. 5). These results agree with previous studies of this question in DRG neurons (O’Malley, 1994).
CREB phosphorylation in response to action potentials
Nuclear transcription factors are key substrates for calcium-activated protein kinases. Phosphorylation of the nuclear transcription factor CREB protein at Ser-133 is critical in regulating transcription of the c-fos gene in response to cAMP or calcium (Gonzales and Montminy, 1989; Lamph et al., 1990; Dash et al., 1991; Sheng et al., 1991) and in consolidating long-term memory inDrosophila (Tully et al., 1994; Yin et al., 1994),Aplysia (Dash et al., 1990; Kaang et al., 1993; Alberini et al., 1994), and mice (Bourtchuladze et al., 1994; Frank and Greenberg, 1994; Stevens, 1994). The best relation between calcium transients andc-fos expression for the present set of stimuli was not the concentration of the second messenger calcium but the interval of time between the bursts of calcium influx (Fig.4 B,E,H,K). This suggests that, for these stimuli, integration by a signaling element downstream from calcium has a time constant of decay that outlasts the period of calcium increase and persists for 1–2 min but fails to propagate signals regulating gene expression if interrupted by 5 min intervals. The dynamics of CREB phosphorylation could account for this behavior, and this hypothesis was tested directly.
Using an antibody specific for CREB phosphorylated at Ser-133 (P-CREB) (Ginty et al., 1993), we studied the kinetics of phosphorylation of this transcription factor in response to action potential stimulation of different patterns and correlated this with c-fosexpression and the changes in intracellular calcium produced by the stimulus (Fig. 6). The specificity of immunocytochemical staining to CREB and P-CREB was confirmed by immunoblotting. This antibody showed an increase in staining intensity of a protein with a molecular weight of ∼41 kDa, after 10 Hz electrical stimulation (Fig. 7), consistent with the increase in nuclear staining in neurons stimulated electrically (Fig. 6). No change in total CREB staining was evident in the nucleus (Fig. 6) or in immunoblots (Fig. 7) after chemical or electrical stimulation. The results show that (1) phosphorylation of CREB at Ser-133 has a low activation threshold, (2) levels of phosphorylated CREB increase rapidly in the nucleus, and (3) phosphorylated CREB in the nucleus decreases slowly after stimulation is stopped.
Phosphorylation of CREB at Ser-133 increased in proportion to stimulus frequency between frequencies of 1 and 10 Hz (Fig.8 A), but high-frequency stimulation clearly was not required to phosphorylate CREB at Ser-133. A significant increase in levels of phosphorylated CREB was produced after 30 min of 1 Hz stimulation (p < 0.001). Phosphorylation of CREB in response to action potentials at a frequency of 10 Hz was rapid, reaching significantly elevated levels (p < 0.001) in <10 sec of stimulation (Fig.8 B). Stimulation for 10 min produced significantly higher levels of phosphorylation (p < 0.001), but stimulus durations from 10 sec to 5 min produced relatively similar effects (Fig. 8 B).
Dephosphorylation of CREB at the critical site of Ser-133, however, proceeded with a much slower time course, declining by one-half within ∼5 min after a brief stimulus burst, but remaining elevated even 30 min after stimulation was stopped (Fig. 8 C). The slow kinetics of CREB dephosphorylation suggest that activation of this transcription factor could serve to sustain the signaling reaction during the interval between stimulus bursts. However, the sustained response makes CREB phosphorylation a relatively poor indicator of stimulus pattern, and therefore phosphorylated CREB would not be expected to account for the differences in c-fos expression produced by the different patterns of action potentials investigated in this study. This was tested by measuring CREB phosphorylation in response to pulse train stimulation.
The four stimulus patterns used in experiments of c-fosexpression produced significantly different levels of CREB phosphorylation at Ser-133 after 30 min of stimulation (p < 0.001) (Figs. 6,9 B). Differences were also evident by the rightward shift in frequency histograms of the staining intensity of all DRG nuclei stimulated with KCl or stimulated electrically (Fig. 6) and in terms of the mean staining intensity of nuclei (Fig. 9 B). All four patterns of electrical stimulation increased phosphorylation of CREB at Ser-133 significantly, with KCl stimulation producing the greatest increase (Figs. 6,9 B). Levels of phosphorylated CREB in response to pulse train stimulation were not as would have been predicted simply from the kinetics of CREB phosphorylation and dephosphorylation in response to constant frequency stimulation. Either brief stimuli repeated frequently (18 impulses at 10 Hz every minute) or long stimuli repeated infrequently (90 impulses at 10 Hz every 5 min) produced comparable large increases in CREB phosphorylation after 30 min (p = 0.21 comparing these two patterns), and these levels were higher than the phosphorylation produced by the intermediate patterns (36/2 and 54/3; p < 0.001) (Fig.9 B).
The total amount of P-CREB integrated over the 30 min stimulus period did not differ substantially for the two extreme stimulus patterns (18/1 vs 90/5), suggesting that differences in time-integrated P-CREB do not account for the large difference in c-fos expression induced by these two stimulus patterns. Measurement of P-CREB after a single 1.8 sec burst is beyond the limits of the method (highly variable results; our unpublished observations), but stimulation for 10 sec elevated P-CREB levels significantly (Fig.10 A). P-CREB reached high levels in <14 min in response either to brief pulses (1.8 sec) repeated at 1 min intervals (p < 0.001 relative to control) or long pulses (9 sec) repeated at 5 min intervals (p < 0.001 relative to control) (Fig.10 A). By the end of the 30 min stimulus period with either a 9 or 1.8 sec pulses, P-CREB had reached maximal levels, because a subsequent stimulus pulse produced no further increase in P-CREB (no significant difference).
The basis for the nonlinear relation between P-CREB staining in the nucleus and stimulus pattern is not evident in the associated calcium transients. Ca2+ levels correlated directly with stimulus burst duration, and the peak increase in [Ca2+]i was not limiting for CREB phosphorylation in these experiments. The shortest stimulus burst (1.8 sec) produced the least increase in Ca2+concentration (383 ± 47 nm) but the highest level of CREB phosphorylation (Fig. 9 B). The high level of CREB phosphorylation at Ser-133 produced by 9 sec duration bursts repeated at 5 min intervals might be explained by sustained activation of a kinase (Hanson and Schulman, 1992) or by inhibition of phosphatase activity (Bito et al., 1996), perhaps in response to the high Ca2+ concentration produced by the long-duration stimulus bursts. Phosphatase activity (Bito et al., 1996) and calcium-dependent protein kinase activity (Hanson and Schulman, 1992) are known to exhibit nonlinearities with respect to the concentration of calcium.
MAP kinase activation in response to different action potential firing patterns
The MAP kinase cascade is involved in neurotrophin-inducedc-fos transcription through the phosphorylation of CREB at Ser-133 (Ginty et al., 1994; Xing et al., 1996) or phosphorylation of ternary complex factors, which, in association with serum response factor (SRF), bind to the SRE element of the c-fos promoter (Triesman, 1996; Xia et al., 1966). The MAP kinase cascade has been associated more closely with growth factor and serum stimulation, but recent work indicates activation of MAP kinase in response to increased [Ca2+] by way of Ras activation (Finkbeiner and Greenberg, 1996; Rosen and Greenberg, 1996). We tested whether (1) the MAP kinase pathway was activated by action potential firing in DRG neurons, (2) whether c-fos expression in response to action potential firing involved MAP kinase activation, and (3) whether the relative activation of CREB and the ERK MAP kinases differed depending on the pattern of action potential firing.
Selective inhibition with pharmacological agents revealed that both the calcium–calmodulin-dependent protein kinase (CaM kinase) and MAP kinase pathways were activated by action potential stimulation in DRG neurons to stimulate c-fos expression (Fig.11). Inhibition of the CaM kinase-dependent pathway was indicated by inhibition ofc-fos mRNA levels in response to a 1.8 sec electrical stimulation repeated at 1 min intervals in neurons pretreated with 30 μm KN-62 (Fig. 11). Similar results were obtained by preincubation with 50 μm PD098059, a MEK1 (ERK-activating kinase) inhibitor (Dudley et al., 1995), or a combination of the two inhibitors (Fig. 11).
The amount of MAPK activation differed significantly in response to the four different pulse train patterns of action potentials (Figs.9 C, 12; p < 0.001). Maximal activation of the ERK MAP kinases was measured after stimulation with 1.8 sec pulses repeated at 1 min intervals (p < 0.009 vs control). Increasing interburst interval was associated with less MAPK activation. No activation of MAPK was detected relative to unstimulated controls in response to 9 sec pulses repeated at 5 min intervals for 30 min (Figs. 9 C,12; p = 0.63). In contrast to CREB phosphorylation, MAP kinase activation increased during the 30 min stimulus period in response to short bursts of action potentials (1.8 sec) repeated at 1 min intervals, but not in response to longer duration bursts (9 sec) repeated at 5 min intervals (Fig. 10 B). This indicates a threshold of <5 min interpulse interval for integration by the MAP kinase enzyme. The difference between CREB and MAPK in responding to stimulus patterns with different interburst intervals is a consequence of the relatively faster dephosphorylation rate of MAPK (Fig. 8 C). No significant decrease in CREB phosphorylation was detected 1 min after stopping a 5 min 10 Hz stimulus, but MAPK phosphorylation levels had declined by ∼50% at this point (Fig.8 C).
We conclude that the temporal dynamics of intracellular signaling pathways are critical in controlling the expression of neuronal genes in response to specific patterns of action potentials. The most critical determinant of c-fos expression was the interval of time between bursts of action potentials. Brief but frequently repeated bursts can induce the coordinated activation of MAP kinase and CREB to induce expression of c-fos. Bursts of action potentials separated by long intervals (5 min) did not effectively activate the MAP kinase cascade, but this stimulus did increase CREB phosphorylation. Maximal induction of c-fos expression required the combined activation of both the MAP kinase pathway and CREB phosphorylation.
CREB phosphorylation and intracellular Ca2+signals display different temporal dynamics in response to action potentials. Phosphorylation of CREB at Ser-133 paralleled the rapid rate of increase in [Ca2+]i, but [Ca2+]i recovered to normal levels within several seconds. This recovery was much faster than CREB dephosphorylation, which remained elevated even 25 min after a 5 min, 10 Hz stimulus was stopped. The persistence in signaling between bursts of action potentials (and calcium transients) is mediated in part by the slow dephosphorylation of CREB at Ser-133; however, the slow kinetics of CREB dephosphorylation at Ser-133 make P-CREB a poor or at least nonlinear transducer of temporal features of action potential stimulation. The high P-CREB levels in the nucleus under stimulus conditions that failed to induce c-fos expression show that CREB phosphorylation at Ser-133 need not be the limiting factor inc-fos expression. This was an unexpected result, given that phosphorylation of CREB at Ser-133 is critical in activating transcription of Ca2+/CRE containing genes, and dephosphorylation of CREB at this site mediates transcriptional shut-off in some experiments (Hagiwara et al., 1992;Hunter, 1995; Thompson et al., 1995). However, multiple regulatory elements in the c-fos promoter and multiple signaling pathways cooperate in the transcriptional activation ofc-fos (Gilman et al., 1986; Hill and Treisman, 1995). The lack of correlation between CREB phosphorylation at Ser-133 produced by long but infrequent stimulus bursts suggests that other elements in the signaling system or transcription process must be limitingc-fos expression. MAP kinase is a convergence point for many signaling pathways in neurons (Seger and Krebs, 1995). Inhibitor studies and immunoblots showed that the MAPK cascade is activated in DRG neurons by membrane depolarization induced by antidromic action potential firing in cultures lacking synapses and that activation of this pathway is dependent on the pattern of action potential firing. In contrast to P-CREB, which shows sustained activation over long interburst intervals, MAPK is unable to integrate bursts of action potentials separated by intervals of ∼5 min. Thus, the inability of MAP kinase to integrate bursts of action potentials separated by intervals of 5 min can partly explain the weak c-fos gene response for this pattern of firing. Differences in nuclear versus cytoplasmic Ca2+ (Hardingham et al., 1997) do not explain the differences in MAP kinase and CREB activation, because measurements using confocal microscopy show no difference in peak calcium response in the nucleus and cytoplasm of DRG neurons. The results suggest that the different rates of activation and inactivation of CREB and MAPK impose different temporal limitations on these two regulatory molecules in integrating repeated bursts of stimuli.
The diminished MAP kinase response did not influence CREB phosphorylation, which was high after stimuli repeated at 5 min intervals. Thus, phosphorylation of CREB in response to this stimulus can be attributed to other Ca2+-dependent signaling enzymes (Hanson and Schulman, 1992; Enslen et al., 1994; Bito et al., 1996; Liu and Graybiel, 1996). SRF and the ternary complex factor elk1 and SAP-1 are important regulators of c-fos transcription, and these are important targets of MAP kinases (Triesman, 1996;Janknecht and Hunter, 1997). The lack of c-fos expression in response to 9 sec bursts at 5 min intervals may result from insufficient phosphorylation of the ternary transcriptional complex associated with SRE. This is consistent with work showing the dependence of c-fos transcription on the combined action of all DNA binding proteins in the promoter into a three-dimensional transcriptional complex (Robertson et al., 1995).
Relevance to nervous system development and synaptic plasticity
The importance of the interval between bursts in regulating gene expression may be relevant to activity-dependent regulation of genes during development of the nervous system. Prenatal spontaneous electrical activity is often low frequency, with long intervals between bursts (Fitzgerald, 1987). In fetal ferrets, for instance, 2–4 sec bursts of activity sweep across the unstimulated retina at intervals of 1–2 min (Meister et al., 1991). This type of stimulus pattern was highly effective in activating c-fos expression in the present study. The highly coordinated spatiotemporally patterned electrical activity is thought to be critical for fine-tuning the connections between neurons in the visual system (Shatz, 1990). Similarly, calcium currents and the number of L-type calcium channels in DRG neurons are downregulated after 1–2 d of electrical stimulation in culture (Li et al., 1996). The efficacy of stimulation in reducing calcium currents varies inversely with the interval between bursts. In developing Xenopus spinal neurons, the frequency of spontaneous calcium transients correlates with discrete effects on development (Gu and Spitzer, 1995). Low-frequency calcium transients regulate neurite outgrowth, and high-frequency calcium transients regulate maturation of potassium channels. No relation was evident between these developmental processes and the number of calcium transients, peak calcium concentration, duration, or time integral of the calcium transient. Recent work on Drosophila shows that multiple training sessions with a rest interval between them (spaced training) produces stronger, longer-lasting memory than the same number of training sessions with no rest interval (massed training) (Yin et al., 1995). Remarkably low frequencies of stimuli, which would cause minimal changes in concentration of intracellular signaling molecules, can be effective in regulating gene expression in the nervous system. Expression of the T-type calcium currents in DRG neurons can be decreased by stimulus frequencies as low as 0.5 Hz. Expression ofc-fos and the cell adhesion molecule L1 genes is regulated by single action potential stimulation delivered at 10 sec intervals, but not 1–3 sec intervals (Itoh et al., 1995).
These results may have relevance to activity-dependent synaptic plasticity, but, in contrast to hippocampal neurons, there is no evidence that stimulus–transcription coupling in DRG neurons is restricted to postsynaptic neurons or dependent on a specialized subsynaptic calcium sensor to activate CREB phosphorylation (Deisseroth et al., 1996). Phosphorylation of CREB in hippocampal neurons has been reported to require synaptic activity, but rapid action potential firing fails to increase CREB phosphorylation (Deisseroth et al., 1996). This is not the case for DRG neurons, because stimulation was provided by electrical depolarization of axons in cultures that lack synapses and that were not spontaneously active or able to interact in an excitatory network to produce more complex firing patterns.
Resonant signal transduction: a dynamical perspective
Differences in MAP kinase activity and CREB phosphorylation are critical in regulating c-fos expression in response to different patterns of stimulation in the present experiment, but it would be reasonable to expect that other patterns of stimulation might produce different correlations between signaling molecules and gene expression. Rather than attributing the input–output function between stimulus pattern and c-fos expression to a response of any one element in the signaling cascade, these results might be better viewed as a dynamic property of the system of signaling reactions transducing patterned membrane depolarization to the transcriptional activation of genes. Nonlinearities and complex behavior could result from a convolution of many different signaling and transcriptional processes under dynamic conditions that may be saturated or not activated in steady-state stimulus conditions. This could include parallel protein kinases, phosphatases, other transcription factors, coactivators or transcription factor binding proteins, or differential effects of Ca2+ on initiation of transcription versus elongation of mRNA transcripts (Lee and Gilman, 1994). Different temporal patterns of stimulation also might regulate CREB binding via dimerization with related bZIP repressor proteins (Meyer and Habener, 1993) or via interdependencies between multiple DNA binding proteins (Robertson et al., 1995; Thompson et al., 1995; Nakajima et al., 1996). Each of these processes may have kinetic features or threshold effects that limit their involvement to certain patterns of stimulation.
Concentration thresholds and subcellular spatial heterogeneity of intracellular signaling components are two important means for providing stimulus–transcription specificity. The present results suggest that differences in temporal dynamics of intracellular signaling pathways can be responsible for selective activation of genes by specific action potential firing patterns. Given the importance of temporal coding in the nervous system, the dynamic aspects of intracellular signaling could be particularly important in regulating intracellular responses to action potentials (Fields and Nelson, 1994;Fields, 1996).
We thank J. Morgan, T. Curran, and L. Robertson for generously providing the fos/LacZ transgenic mouse strain and D. Abebe for establishing and maintaining the colony of transgenic mice.
Correspondence should be addressed to Dr. R. Douglas Fields, Head, Neurocytology and Physiology Unit, National Institutes of Health, National Institute of Child Health and Human Development, Laboratory of Developmental Neurobiology, Building 49, Room 5A38, Bethesda, MD 20892.
Dr. Itoh’s present address: Tokyo Metropolitan Institute of Medical Science, Tokyo 113, Japan.