Abstract
The phases at which network neurons fire in rhythmic motor outputs are critically important for the proper generation of motor behaviors. The pyloric network in the crustacean stomatogastric ganglion generates a rhythmic motor output wherein neuronal phase relationships are remarkably invariant across individuals and throughout lifetimes. The mechanisms for maintaining these robust phase relationships over the long-term are not well described. Here we show that tonic nanomolar dopamine (DA) acts at type 1 DA receptors (D1Rs) to enable an activity-dependent mechanism that can contribute to phase maintenance in the lateral pyloric (LP) neuron. The LP displays continuous rhythmic bursting. The activity-dependent mechanism was triggered by a prolonged decrease in LP burst duration, and it generated a persistent increase in the maximal conductance (Gmax) of the LP hyperpolarization-activated current (Ih), but only in the presence of steady-state DA. Interestingly, micromolar DA produces an LP phase advance accompanied by a decrease in LP burst duration that abolishes normal LP network function. During a 1 h application of micromolar DA, LP phase recovered over tens of minutes because, the activity-dependent mechanism enabled by steady-state DA was triggered by the micromolar DA-induced decrease in LP burst duration. Presumably, this mechanism restored normal LP network function. These data suggest steady-state DA may enable homeostatic mechanisms that maintain motor network output during protracted neuromodulation. This DA-enabled, activity-dependent mechanism to preserve phase may be broadly relevant, as diminished dopaminergic tone has recently been shown to reduce Ih in rhythmically active neurons in the mammalian brain.
Introduction
The timing of neuronal firing is critical for many processes. An animal's location is coded by the phase of hippocampal neuron firing relative to the theta oscillation (Burgess et al., 2007). Human memory formation may depend upon coordinating neuronal spiking with local oscillations (Rutishauser et al., 2010). Changing the phase of neuronal firing in a motor circuit can change the ensuing motor behavior. In Aplysia the same feeding circuit can generate opposing ingestive versus egestive behaviors by switching the phase of the B8 motorneuron relative to other circuit neurons (Jing and Weiss, 2002). In the leech, the phasing among network neurons determines whether segmental contractions of the heart tubes are synchronous or peristaltic (Weaver et al., 2010).
A large literature suggests that neuronal firing phase can be modulated on multiple time scales. Modulatory systems use volume transmission that comprises both phasic micromolar and tonic nanomolar components (Zoli et al., 1998; Fuxe et al., 2010; Oginsky et al., 2010). Phasic and tonic dopamine (DA) transmissions have distinct functions (Schultz, 2007): phasic release encodes temporally relevant information and modifies neuronal function on a moment-to-moment basis, while steady-state or tonic DA in the extracellular space is thought to enable motor and cognitive processes. Tonic nanomolar DA can act through translation-dependent mechanisms to persistently regulate the voltage-gated transient potassium current (IA) in an activity-independent fashion (Rodgers et al., 2011). A recent study showed that diminished dopaminergic tone in mouse models of Parkinson's disease caused a loss of pacemaker activity and a reduction in Ih in globus pallidus neurons (Chan et al., 2011). Here we used a small motor circuit with well defined neuronal phase relationships underpinned in part, by IA and Ih, to examine the long-term effects of steady-state DA on neuronal firing phase.
The pyloric circuit in the stomatogastric nervous system of the spiny lobster, Panulirus interruptus, is a 14 neuron recurrent inhibitory circuit that produces a robust triphasic motor output (Fig. 1; Stein, 2009). All six pyloric cell types show rhythmic, ∼20 mV oscillations in membrane potential accompanied by bursts of action potentials on the depolarized plateau (Fig. 1B). Each pyloric cell type bursts at a specific phase in the motor output.
Neuronal firing phase is critical to a neuron's function within the pyloric network. For example, the lateral pyloric (LP) neuron normally functions to set the upper limit on pyloric cycle frequency by directly inhibiting the pacemaker (Weaver and Hooper, 2003). However, the effect of LP inhibition is phase dependent. LP can advance or delay the pacemaker, depending upon when LP inhibition occurs during the pacemaker oscillation (Thirumalai et al., 2006). Thus, LP firing phase determines its function within the network. It was previously shown that a 10 min exposure to micromolar DA disrupted normal LP function by distorting the phase relationship between LP and the pacemaker (Johnson et al., 2011). In contrast, population studies demonstrated that phase relationships were largely invariant between individuals and over their lifetimes (Bucher et al., 2005). How are these disparate findings reconciled? The data presented here suggest that dopaminergic tone enables homeostatic mechanisms that act over tens of minutes to preserve motor network output during protracted exposure to micromolar DA.
Materials and Methods
Animals.
California spiny lobsters, Panulirus interruptus, were purchased from Don Tomlinson Commercial Fishing, Catalina Offshore Products, and Marinus Scientific and housed in saltwater aquaria at Georgia State University (Atlanta, GA). Animals of both sexes were used in these experiments.
Stomatogastric nervous system (STNS) dissection, cell identification, extracellular recordings and motor pattern analyses.
Lobsters were anesthetized on ice for at least 30 min, followed by STNS dissection, as previously described (Selverston et al., 1976). The STNS was pinned in a Sylgard-lined dish (Fig. 1). The stomatogastric ganglion (STG) was desheathed, and during this process a portion of the juxtaposed stomatogastric nerve (stn) was also desheathed. A Vaseline well was constructed around the STG and the juxtaposed stn and dorsal ventricular nerve (dvn). The well was continuously superfused for the remainder of the experiment. Using a Dynamax peristaltic pump (Rainin), the STG was superfused with Panulirus saline [containing, in mm: 479 NaCl, 12.8 KCl, 13.7 CaCl2, 39 Na2SO4, 10 MgSO4, 2 glucose, 4.99 HEPES, 5 TES (N-[tris(hydroxymethyl)methyl]-2-aminoethanesulfonic acid, 2-[(2-hydroxy-1,1-bis(hydroxymethyl)ethyl)amino]ethanesulfonic acid), pH 7.4]. Temperature was continuously monitored with a miniature probe inside the well. Temperatures changed by <1°C throughout the course of the day (the change ranged from 0.1 to 0.9°C on any given day), and by only 3°C across all experiments (19−22°C).
Cells were identified by combining standard intracellular and extracellular recording techniques. Intracellular somatic LP recordings were obtained using 20–40 MΩ glass microelectrodes filled with 3 m KCl connected to Axoclamp 2B or 900A amplifiers (Molecular Devices). Extracellular recordings of identified motor neurons were obtained using a model 1700 differential AC amplifier (A-M Systems) and stainless steel pin electrodes. LP neurons were identified by their distinct waveforms, the timing of their voltage oscillations, and correlation of spikes on the extracellular and intracellular recordings.
Pyloric activity was recorded with Axoscope v8.2 software (Molecular Devices) via extracellular electrodes on the lateral ventricular nerve (lvn) and pyloric dilator nerve (pdn). Extracellular recordings were analyzed using DataView v6.3.2 (Heitler, 2009) to determine cycle period and frequency, spikes per burst, burst duration, duty cycle, intraburst spike frequency, LP-on delay, and LP phase. Reported values for all parameters represent a 10 cycle average. By convention the first spike in the pyloric dilator neuron (PD) defines the beginning of a pyloric cycle, and the phasing of pyloric neurons is determined relative to the start of the cycle. Since we were interested in recovery mechanisms intrinsic to LP, we broke with convention and used the last spike in PD to define the beginning of the cycle, because tonic DA can cause persistent changes in PD ionic conductances (Rodgers et al., 2011). One cycle period was therefore defined as the last spike in one PD burst to the last spike in the next PD burst (Fig. 1B). The LP-on phase, which defined the point in the cycle where LP began firing, was measured as the time between the last spike in PD and the first spike in LP (termed LP-on delay; Fig. 1B) divided by cycle period. LP duty cycle was calculated as LP burst duration divided by cycle period. Intraburst spike frequency was calculated as LP spikes per burst divided by LP burst duration.
Experimental design.
After STNS dissection and cell identification, we performed the following experimental protocol: DA (5 nm or 5 μm) was applied to the STG for 1 h (Fig. 1D). Control preparations received saline during this time. In both DA-treated and control conditions, this was followed by a 3 h saline wash, unless otherwise indicated. After 4 h, descending input was removed with a sucrose block on the stn and the ganglion was prepared for two-electrode voltage clamp (TEVC) to measure Ih, as described below.
Manipulations to study activity dependence.
We used two methods to study the role of LP activity in phase recovery. First, 100 nm tetrodotoxin (TTX) was added to the perfusate from t = −10 min to 1 h (Fig. 1D). This blocked all action potentials and slow membrane oscillations. A 1 h TTX (or TTX + DA) administration was followed by a 3 h wash with Panulirus saline. Preparations that did not recover a pyloric rhythm by the end of the 3 h wash were excluded from the analysis. In 19 of 21 cases, rhythmic activity resumed after a 1 h wash, and 3 experiments were excluded. Second, to either prevent decreases in LP burst duration in 5 μm DA, or to induce decreases in burst duration in control and 5 nm DA, positive (0.7–2.5 nA) or negative (0.5–2 nA) DC current was injected into the LP cell, respectively, for 1 h. We used the minimum amount of depolarizing current required to prevent a change in burst duration and the minimum amount of hyperpolarizing current required to induce a 20–30% decrease in burst duration. The necessary amount of current was determined empirically just before the start of the experiment (i.e., from t = −15–0 min; Fig. 1D).
TEVC.
A portion of the stomatogastric nerve was desheathed and isolated in a second Vaseline well. Descending inputs to the STG were removed with a sucrose block applied into this well for 1 h. During the same time the STG was continuously superfused with Panulirus saline containing picrotoxin (1 μm) to block glutamatergic synaptic inputs. Voltage-dependent ion channel blockers were also added to the perfusate: TTX (100 nm, INa), tetraethylammonium (20 mm, IK(V) and IK(Ca)), and cadmium chloride (CdCl2, 200 μm, ICa). TTX was from Tocris Bioscience, all other reagents were from Sigma. For TEVC the LP neuron was impaled with two micropipettes (7–10 MΩ when filled with 3 m KCl) connected to Axoclamp 2B or 900A amplifiers (Molecular Devices). Using Clampex software (Molecular Devices, v.8.2) LP was clamped to a −50 mV holding potential. Ih was then measured using a series of 4 s hyperpolarizing voltage steps, from −60 to −120 mV in 10 mV increments. Ih was measured by subtracting the initial fast leak current step from the slowly developing peak of Ih at the end of each negative voltage step. Current peaks were converted to conductance (G = Ipeak/(Vm − Vrev) and fit to a first-order Boltzmann equation. Vrev Ih = −35 mV (Kiehn and Harris-Warrick, 1992).
Statistical analysis.
Data were checked for normal distribution and analyzed using parametric statistical tests with Prism software package v5.01 (GraphPad). Significance was set at p < 0.05 in all cases. Means are followed by SDs. Individual samples that were >2 SDs from the mean were excluded from the analyses after determining the mean. ANOVAs are followed by post hoc tests that make all possible comparisons between columns (Tukey's) or that compare all columns to t = 0 (Dunnett's).
Results
A slow phase recovery mechanism
The pyloric network in the crustacean STG is a small central pattern generator that controls rhythmic movements of the foregut (Fig. 1). This 14 neuron network produces a triphasic motor output (Fig. 1B). The electrically coupled pacemaker kernel, composed of one anterior burster and two PD neurons, rhythmically inhibits follower neurons, including the single LP and the eight pyloric constrictor neurons (PYs). Using the conventional definition of phase (Bucher et al., 2005), PD firing represents phase 1 of the triphasic output. Upon removal of pacemaker inhibition, LP and PY neurons rebound to fire a burst of action potentials on a depolarized plateau. Relative to PY, LP has a faster rate of postinhibitory rebound (PIR) and fires first to strongly inhibit PD and weakly inhibit PY (phase 2). PY subsequently fires on rebound from pacemaker inhibition and inhibits LP (phase 3), thereby terminating the LP burst and LP inhibition of the pacemaker, causing the cycle to begin again. Thus, the order of neuronal firing is consistently PD-LP-PY, and the delay in firing between PD-LP and PD-PY scales with cycle frequency.
Experimental preparation and protocol. A, Diagram of the dissected STNS pinned in a dish with a Vaseline well (gray, open rectangle) around the STG. All descending inputs project to the STG through the stn. There are ∼30 neurons in the STG. The single LP is diagrammed. The LP sends its axon through the dvn and lvn as it projects to its target muscles (small, stippled rectangles). B, The spontaneous motor output of the pyloric network. Top trace is a representative LP intracellular recording showing spontaneous, rhythmic oscillations in membrane potential and spikes on a depolarized plateau. Lower traces are extracellular recordings from the pdn and the lvn in the same experiment. The pdn exclusively contains the two axons from the two PDs in the STG. The lvn contains several axons including those from the LP, PD and PY neurons, whose spikes are seen on the traces. Parameters measured from the traces: a, LP burst duration and intraburst spike frequency; b, cycle period; c, LP-on delay. C, Pyloric circuit. Open circles represent pyloric neurons: pacemaker kernel = anterior burster (AB) with two PD cells; follower cells = one LP neuron and 8 PY cells. Filled circles represent inhibitory chemical synapses (PD, acetylcholine; all others, glutamate). Resistors represent electrical coupling. It should be noted that this is a reduced pyloric circuit and only 4 of 6 cell types are represented. D, Experimental timeline. The STNS was dissected and cells were identified. This process typically took 3–5 h. The end of cell identification marked t = 0. At this point, Panulirus saline (control) or DA (DA-treated) was superfused for 1 h, followed by a 3 h wash with Panulirus saline. At t = 4 h a sucrose block was applied to the stn and the STG was superfused with blocking saline for 1 h, after which TEVC was used to measure LP currents. Both lvn and pdn recordings were maintained from t = 0–4 h.
Within the network, LP functions to slow cycle frequency by directly inhibiting the pacemaker kernel (Weaver and Hooper, 2003). A 10 min bath application of 100 μm DA phase advanced LP, reduced LP burst duration and increased pacemaker cycle frequency (Flamm and Harris-Warrick, 1986a). Under these conditions the phase of LP was shifted relative to pacemaker kernel, and LP synaptic inhibition was ineffective because it occurred when the pacemaker kernel was in the refractory period of its oscillation (Johnson et al., 2011). LP lost its functionality in this situation; it could no longer slow cycle frequency, so the upper limit on network cycle frequency vanished. Given that this network is built to maintain phase (Bucher et al., 2005), we were curious as to what would happen if the experiment continued beyond 10 min.
To induce a persistent perturbation in LP phase, we challenged the pyloric circuit with tonic DA applications. The experimental design is shown in Figure 1 and described in Materials and Methods: The STNS was dissected, the STG was superfused with DA (5 nm or 5 μm, DA-treated) or Panulirus saline (control) for 1 h followed by a 3 h wash with Panulirus saline. Extracellular recordings from the pdn and lvn were maintained throughout this portion of the experiment. After the 4 h treatment, descending input was blocked and the STG was superfused with blocking saline for 1 h, and then LP currents were measured with TEVC. Using the extracellular recordings we measured the phase of LP every 10 min during the first hour and every 30 min thereafter throughout the 3 h wash. Changes in LP-on phase over the course of the experiments are shown in Figure 2. A two-way ANOVA indicated there were no significant differences in LP-on phase between treatment groups at any time point examined (Fig. 2A); however, there were significant changes over time within DA treatment groups (Figs. 2B–D). There was not a significant change in LP phase in the control treatment group; however, there was an overall trend toward phase advancement (Fig. 2B). The 5 nm DA preparations displayed a significant, sustained phase advance throughout the experiment (Fig. 2C). The 5 μm DA treatment group also showed a significant phase advance by 10 min, but phase slowly recovered over the next 60 min, despite the continued presence of 5 μm DA (Fig. 2D). Upon DA washout average LP-on phase became delayed, suggesting recovery induced a persistent change and was not simply due to receptor desensitization. In the 5 μm DA treatment group, the LP phase advance at 10 min must have been at least partially due to previously demonstrated DA-induced shifts in LP IA and Ih voltage dependencies causing IA to be reduced and Ih to be increased (Harris-Warrick et al., 1995b). The change in LP IA is largely responsible for the phase advance (Harris-Warrick et al., 1995b), but the threshold for this change is ∼1 μm DA (Zhang et al., 2010); thus, it is not clear what caused the significant, persistent phase advance in the 5 nm DA treatment group. It may not even have been due to DA given the trend in the control preparations. Here we concentrate on the mechanism for slow phase recovery, as observed in the 5 μm DA treatment group. It should be noted that LP-off phase was similarly phase-advanced in the 5 μm DA, but not control or 5 nm DA treatment groups (repeated-measures ANOVAs: control, F(5,11) = 0.7413, p = 0.6946; 5 nm DA, F(4,11) = 1.929, p = 0.0613; 5 μm DA, F(5,11) = 4.650, p < 0.0001).
Tonic application of 5 μm DA triggered a slow phase recovery mechanism. The experiments in Figure 1D were performed for control, 5 nm, and 5 μm DA treatment groups; and LP-on phase was measured every 10 min for the first hour and every 30 min thereafter. Some 5 μm DA preparations also received 0.5 mm Cs from t = 0–4 h. A, There are no significant differences in LP-on phase between treatment groups. LP-on phase (mean + SD) is plotted over time for the four treatment groups. Vertical slashes on the x-axis indicate a change in scale. There were no significant differences between treatment groups at any time point examined (two-way ANOVA: treatment, F(3,253) = 0.2340, p = 0.8718; time, F(11,253) = 5.020, p < 0.0001 (see B–E); interaction, F(33,253) = 3.531, p < 0.0001). B–D, There are significant changes in LP-on phase over time within treatment groups. Fold changes in LP-on phase are plotted over time for a given treatment group. Each thin line represents the fold change (t/t = 0) in LP-on phase throughout one experiment. The thick black line in each graph represents the average fold change for the treatment group. Highlighted time points indicate the period of DA application. Asterisks indicate significantly different from t = 0, as determined using repeated-measures ANOVAs with Dunnett's post hoc tests that compares all time points to t = 0. (B, control: F(11,6) = 1.926, p = 0.0515; C, 5 nm DA: F(11,8) = 4.253, p < 0.0001; D, 5 μm DA: F(11,5) = 9.72, p < 0.0001). E, A Cs dose–response curve showing the fraction of Ih remaining as increasing concentrations of Cs were sequentially superfused into the Vaseline well surrounding the STG. Ih Gmax was measured using TEVC 10 min after a given Cs concentration entered the bath. Fraction of Ih remaining was determined to be: (Gmax in x mm Cs) ÷ (Gmax in 0 Cs). Each data point represents the mean and SD for 3 experiments. F, Cs prevents LP-on phase recovery. Fold changes are plotted over time for the 5 μm DA treatment group that also received 0.5 mm Cs from t = 0–4 h (black line under x-axis). *Significantly different from t = 0 using a repeated-measures ANOVA with a Dunnett's post hoc test (F(11,4) = 4.2, p = 0.0003).
Slow phase recovery requires persistent changes in Ih
Ih is a resonating conductance with demonstrated plasticity (Hutcheon and Yarom, 2000; Wahl-Schott and Biel, 2009). We tested Ih involvement in LP-on phase recovery by repeating the experimental paradigm shown in Figure 1D for control and 5 μm DA treatment groups with the inclusion of 0.5 mm CsCl (Cs) in the perfusate from t = 0–4 h to block ∼30% of Ih (Fig. 2E). Adding Cs to the control treatment group had little effect on LP-on phase: Average changes in LP-on phase were <10% and not statistically significant at any time point examined (repeated-measures ANOVA, F(11,5) = 0.5579, p = 0.8539). On the other hand, adding Cs to the 5 μm DA treatment group had a significant effect. While partially blocking Ih did not prevent the DA-induced LP-on phase advance, it did prevent phase recovery during DA administration (Fig. 2F). These results implied that Ih was critical to the slow phase recovery mechanism.
We next examined whether or not LP Ih was persistently altered in neurons that showed phase recovery (5 μm DA) versus those that did not (5 nm DA and control). LP Ih was measured with TEVC at the end of the experiments diagrammed in Figure 1D. In addition, to obtain a baseline for comparison, in a separate series of experiments LP Ih was also measured acutely (i.e., using the diagram in Fig. 1D, block from t = 0–1 h, followed by TEVC). We found that peak LP Ih was significantly increased in 5 μm DA relative to acute, control and 5 nm DA treatment groups (Fig. 3A). The increased peak current was not due to a change in Ih voltage dependence of activation (Fig. 3B), but to a significant, persistent increase in LP Ih Gmax (Fig. 3C). Average Ih Gmax in the 5 μm DA treatment group (0.162 ± 0.03 μS) was increased 35%, 50% and 44% relative to acute (0.119 ± 0.017 μS), control (0.108 ± 0.031 μS), and 5 nm DA (0.113 ± 0.027 μS) treatment groups, respectively.
5 μm DA caused a persistent increase in LP Ih Gmax by the end of the experiment diagrammed in Figure 1D. A, Typical LP Ih current traces elicited by a series of hyperpolarizing steps using TEVC. Acute measures were obtained by blocking from t = 0–1 h followed by TEVC. Control and DA treatment groups were obtained as diagrammed in Figure 1D. B, Plots of normalized voltage dependence of activation for LP Ih are not significantly different across the four treatment groups. Using TEVC, the LP was held at −50 mV and subjected to a series of hyperpolarizing test pulses. Conductance was obtained for each test pulse and normalized by the conductance for the −120 mV test pulse. Each data point represents the mean ± SD for n ≥ 6. C, Plots of LP Ih Gmax for the four experimental preparations. Each symbol represents the LP Ih Gmax measured with TEVC at the end of one experiment. Horizontal line represents the mean for the group. *Significantly different from acute, control and 5 nm DA as determined using a one-way ANOVA with Tukey's post hoc test: F(3,25) = 5.619, p = 0.0051. Acute, control, and 5 nm DA treatment groups were not significantly different from one another (Tukey's, p > 0.05).
A decrease in burst duration is necessary for the increase in LP IhGmax and phase recovery
We conceptualized the slow phase recovery mechanism in terms of an activity-dependent homeostat (Davis, 2006). A homeostat maintains a target, in this case phase. A homeostat triggers a target recovery mechanism(s), in this case an increase in LP Ih Gmax. An activity-dependent homeostat also contains a sensor that detects a change in some measure of neuronal or network activity.
To identify changes in activity that were unique to the 5 μm DA treatment group during the time period for triggering phase recovery, we measured the following parameters for all treatment groups throughout the experiments diagrammed in Figure 1D: LP burst duration (Fig. 4), LP duty cycle, which defines the fraction of the cycle throughout which LP is active (Fig. 5), LP intraburst spike frequency (Fig. 6) and pyloric cycle frequency (Fig. 7). Two-way ANOVAs indicated that there were no significant differences between treatment groups at any time points examined for any of these parameters (Figs. 4A, 5A, 6A, 7A). However, within group analyses revealed two distinct features that were exclusive to 5 μm DA treatment during phase recovery: a significant decrease in burst duration (Fig. 4B–D) and a significant increase in pyloric cycle frequency (Fig. 7B–D). There were no significant differences in LP duty cycles (Fig. 5B–D) or intraburst spike frequencies (Fig. 6B–D) during the initiation of phase recovery (∼10–40 min) for any treatment group. However, intraburst spike frequency was significantly and persistently increased by 50 min in the control treatment group, while duty cycle was significantly decreased from 60 min through the end of the experiment in the 5 μm DA treatment group.
LP burst duration significantly decreased during the first hour of the experiment diagrammed in Figure 1D for 5 μm, but not control or 5 nm DA preparations. A, Plot of LP burst duration (mean + SD) for the 3 treatment groups at 10 min intervals during the first hour of the experiment, and every 30 min thereafter. Vertical slashes on the x-axis indicate a change in scale. There were no significant differences between treatment groups at any time point (two-way ANOVA: treatment, F(2,253) = 0.008879, p = 0.9912; time, F(11,253) = 3.888, p < 0.0001 (see B–D); interaction, F(22,253) = 1.794, p = 0.0178). B–D, Fold changes in burst duration over time within a treatment group. Each thin line represents one experiment and is a plot of the fold change (t/t = 0) in LP burst duration throughout the experiment diagrammed in Figure 1D. Thick black lines represent average fold changes. Highlighted data points represent the period of DA application. *Significantly different from t = 0, as determined using repeated-measures ANOVAs with Dunnett's post hoc tests that compare all time points to t = 0 (B, control: F(11,8) = 3.723, p = 0.0002; C, 5 nm DA: F(11,8) = 1.017, p = 0.4386; D, 5 μm DA: F(11,7) = 3.235, p = 0.0011).
Fold changes in LP duty cycle during the experiments illustrated in Figure 1D. A, LP duty cycle (mean + SD) is plotted every 10 min for the first hour of the experiment and every 30 min thereafter. Vertical slashes on the x-axis indicate a change in scale. There were no significant differences in duty cycle between the three treatment groups at any time point examined (two-way ANOVA: treatment, F(2,231) = 0.09067, p = 0.9137; time, F(11,231) = 2.826, p = 0.0018; interaction, F(22,231) = 1.554, p = 0.0587). B–D, Fold changes in duty cycle over time within each treatment group. Each thin line represents one experiment and is a plot of the fold change (t/t = 0) in LP duty cycle throughout the experiment diagrammed in Figure 1D. Thick black lines represent average fold changes. Highlighted data points represent the period of DA application. *Significantly different from control as determined using repeated-measures ANOVAs with Dunnett's post hoc tests that compare all time points to t = 0 (B, control: F(11,7) = 0.8325, p = 0.6084; C, 5 nm DA: F(11,9) = 1.918, p = 0.0458; D, 5 μm DA: F(11,7) = 3.265, p = 0.0011).
LP intraburst spike frequency significantly increased during the first hour of the experiment illustrated in Figure 1D for control but not 5 nm or 5 μm DA preparations. A, LP intraburst spike frequency (mean + SD) is plotted for the three treatment groups every 10 min for the first hour of the experiment and every 30 min thereafter. Vertical slashes on the x-axis indicate a change in scale. There were no significant differences between the three treatment groups at any given time point (two-way ANOVA: treatment, F(2,220) = 1.574, p = 0.2319; time, F(11,220) = 2.924, p = 0.0013 (see B–D); interaction, F(22,220) = 1.039, p = 0.4184). B–D, Fold changes in intraburst spike frequency over time within a treatment group. Each thin line represents a plot of the fold change (t/t = 0) in LP intraburst spike frequency during one experiment. Thick black lines represent average fold changes. Highlighted data points represent period of DA application. *Significantly different from t = 0 as determined using repeated-measures ANOVAs with Dunnett's post hoc tests that compare all time points to t = 0 (B, control: F(11,9) = 3.690, p = 0.0002; C, 5 nm DA: F(11,6) = 1.401, p = 0.1935; D, 5 μm DA: F(11,6) = 0.5352, p = 0.8724).
Pyloric cycle frequency is significantly increased during the first hour of the experiment diagrammed in Figure 1D for 5 μm, but not control or 5 nm DA preparations. A, Plot of pyloric cycle frequency (mean + SD) for the three preparations at 10 min intervals during the first hour of the experiment and every 30 min thereafter. Vertical slashes on the x-axis represent a change in scale. There were no significant differences between treatment groups at any time point (two-way ANOVA: treatment, F(2,385) = 1.283, p = 0.2900; time, F(11,385) = 2.285, p = 0.0103 (see B–D); interaction, F(22,385) = 3.829, p < 0.0001). B–D, Fold changes in pyloric cycle frequency over time within a treatment group. Each thin line represents a plot of the fold change (t/t = 0) in cycle frequency throughout one experiment. Thick black lines represent the average fold changes. Highlighted data points represent the period of DA application. *Significantly different from t = 0, as determined using repeated-measures ANOVAs with Dunnett's post hoc tests that compare all time points to t = 0 (B, control: F(11,9) = 4.288, p < 0.0001; C, 5 nm DA: F(11,14) = 0.9297, p = 0.5136; D, 5 μm DA: F(11,12) = 2.931, p = 0.0017).
Based on these data, we hypothesized that a decrease in LP burst duration was necessary for the increase in LP Ih Gmax and phase recovery. To test this hypothesis, we repeated the experiment shown in Figure 1D for the 5 μm DA treatment group, except we continuously injected depolarizing current into the LP from t = 0–60 min to prevent the decrease in LP burst duration without altering the other indices of activity (Fig. 8). It is likely that DA normally decreases LP burst duration by phase advancing PY synaptic inhibition to terminate the LP burst (Selverston and Moulins, 1987; Harris-Warrick et al., 1995a). Depolarizing LP may not prevent PY phase advancement, but nevertheless prevented the decrease in LP burst duration in 5 μm DA (compare Fig. 8A vs 4D). A comparison between 5 μm DA treatment groups that did versus did not receive current injection showed that the depolarization did not appreciably alter LP duty cycle (Fig. 8B vs 5D) or LP intraburst spike frequency (Fig. 8C vs 6D) but may have reduced or prevented the DA-induced increase in pyloric cycle frequency (Fig. 8D vs 7D). LP was similarly phase advanced in the 5 μm DA treatment groups with or without current injection (Fig. 8E vs 2D); however, abolishing the decrease in LP burst duration prevented phase recovery in 5 μm DA (Fig. 8E) and the increase in LP Ih Gmax (Fig. 8F). Together these data suggested that a decrease in LP burst duration was necessary for the increase in LP Ih Gmax and slow phase recovery.
A reduction in LP burst duration is necessary to elicit slow phase recovery. The experiment illustrated in Figure 1D was performed for the 5 μm DA treatment group with depolarizing current injections from t = 0–1 h. A–E, The fold changes (t/t = 0) in LP burst duration (A), LP duty cycle (B), LP intraburst spike frequency (C) pyloric cycle frequency (D) and LP-on phase (E) were plotted every 10 min for the first hour of the experiment and every 30 min thereafter. Vertical slashes on the x-axis represent a change in scale. Each thin line represents one experiment. Thick black lines represent the average fold changes. Highlighted data points represent the period of DA application. *Significantly different from t = 0 as determined using repeated-measures ANOVAs with Dunnett's post hoc tests that compared all time points up to t = 60 min with t = 0 (A, LP burst duration: F(11,5) = 0.7345, p = 0.6257; B, LP duty cycle: F(11,5) = 2.453, p = 0.0413; C, LP intraburst spike frequency: F(11,5) = 0.7780, p = 0.5936; D, pyloric cycle frequency: F(11,5) = 1.758, p = 0.1419; E, LP-on phase: F(11,5) = 5.188, p = 0.0009). F, Depolarizing current injection prevented the increase in LP Ih Gmax in 5 μm DA. LP Ih Gmax was plotted for control, 5 μm DA and 5 μm DA with depolarizing current injection (depol I). Each symbol represents one experiment. *Significantly different from control and 5 μm DA with depol I, as determined with a one-way ANOVA with Tukey post hoc test, F(2,18) = 8.976, p = 0.0024.
Tonic 5 nm DA permits decreases in burst duration to evoke persistent increases in LP IhGmax
The phase recovery mechanism was activity dependent, but was it also DA dependent, or was DA necessary only insofar as it caused the change in activity? To answer this question, we mimicked an extreme decrease in burst duration with TTX. The experiments shown in Figure 1D were repeated for all three treatment groups, but 100 nm TTX was included from t = −10 min− 1 h. If a decrease in activity alone was sufficient to increase LP Ih Gmax, then TTX should significantly increase LP Ih Gmax in control and both DA-treated preparations. If both a reduction in activity and DA were absolutely required for the increase in LP Ih Gmax, then TTX should have no effect on control, but should significantly increase LP Ih Gmax in 5 nm and/or 5 μm DA preparations, depending upon the DA concentration required.
TTX blocked spontaneous activity in all three treatment groups during its application and for at least 1 h into the wash period (Fig. 9A). Blocking activity in the control preparations did not significantly increase average LP Ih Gmax (Fig. 9B). On the other hand, blocking activity in the 5 nm DA treatment group produced a significant increase in average LP Ih Gmax relative to control and 5 nm DA preparations lacking TTX (Fig. 9B). Blocking activity in the 5 μm DA treatment group did not increase the average LP Ih Gmax beyond that observed for 5 μm DA alone (Fig. 9B).
DA and decreased activity are both necessary to produce a significant increase in LP Ih Gmax. A, A typical LP intracellular recording showing voltage trace over time and lvn and pdn extracellular traces before and after TTX application for experiments illustrated in Figure 1D. Note that TTX was applied from t = −10 min− 1 h. Calibrations: 10 mV, 500 ms. B, Plots of LP Ih Gmax at the end of the experiments diagrammed in Figure 1D under the indicated conditions. All groups were compared against each other by post hoc analysis, A denotes significant differences from control group, and B denotes significant differences from 5 nm alone as determined using a one-way ANOVA with a Tukey's post hoc test, F(5,36) = 5.832, p = 0.0007.
We interpreted the data to mean that a significant, persistent increase in LP Ih Gmax required at least two events: a decrease in LP burst duration and an additional unknown activity-independent event(s) evoked by tonic nanomolar DA. A 1 h bath application of 5 μm DA triggered both events and a further decrease in activity did not produce a larger increase in LP Ih Gmax (Fig. 9B). On the other hand, a 1 h bath application of 5 nm DA alone had no effect, but when combined with a decrease in activity, it produced a significant increase in LP Ih Gmax (Fig. 9B). Interestingly, LP Ih Gmax in TTX alone was not significantly different from LP Ih Gmax for any other treatment group (Fig. 9B), suggesting that a complete ∼2 h cessation of activity incrementally increased LP Ih Gmax without requiring DA.
One potential problem with these experiments is that TTX not only modifies burst duration, but also removes most modulatory input (due to action potential blockage in the terminals), and it also removes membrane potential oscillations, so the conclusion that the persistent increase in LP Ih Gmax requires a decrease in LP burst duration may not be correct. To address this issue, the experiments shown in Figure 1D were repeated for control and 5 nm DA treatment groups, but burst duration was significantly reduced with a continuous hyperpolarizing current injection into LP from t = 0–60 min. If our interpretation of the TTX data was correct, then LP Ih Gmax should be increased in the 5 nm DA but not control treatment groups.
Hyperpolarizing current injection significantly reduced burst duration and duty cycle in both control and 5 nm DA treatment groups, but had no effect on their cycle or intraburst spike frequencies (Fig. 10A–D). The changes in burst duration induced by hyperpolarizing current injection were within the range observed for the 5 μm DA treatment group (compare Fig. 10A to Fig. 4D). The results were consistent with our interpretation of the TTX data. Under these conditions LP Ih Gmax was significantly increased in the 5 nm DA relative to the control treatment group (Fig. 10E), despite the fact that LP activity was the same in both treatment groups (Figs. 10A–D). Notably, hyperpolarizing current injection induced a phase delay in LP (Fig. 10F), rather than the phase advance induced by 5 μm DA (Fig. 2D). The phase delay was sustained in both treatment groups (Fig. 10F), perhaps because other network neurons did not undergo changes in activity; for example, cycle period was not significantly altered in these experiments. Together with our previous findings (Figs. 2, 8, 9), these data suggested that tonic 5 nm DA enabled a mechanism whereby a decrease in LP burst duration triggered an increase in LP Ih Gmax, and contributed to phase recovery in 5 μm DA.
5 nm DA permits a decrease in LP burst duration to trigger an increase in LP Ih Gmax. The experiments in Figure 1D were performed for control and 5 nm DA treatment groups, but hyperpolarizing current (hyperpol I) was injected into the LP from t = 0–1 h. A–D, Current injection reduced LP burst duration (A) and duty cycle (B) to the same extent in control and 5 nm preparations, but had no effect on cycle frequency (C) or intraburst spike frequency (D). Solid lines represent fold changes for the 5 nm treatment groups (mean ± SD, n = 5). Dashed lines represent fold changes for the control treatment groups (mean ± SD, n = 5). E, The decrease in burst duration produced an increase in LP Ih Gmax in 5 nm preparations relative to control. *Significant difference, Student's t test, p = 0.002. F, Current injection produced a sustained average LP phase delay in both treatment groups. Solid line represents fold changes for the 5 nm treatment group (mean ± SD, n = 5). Dashed line represents fold changes for the control treatment group (mean ± SD, n = 5).
Discussion
The pyloric network generates a rhythmic motor output with tightly maintained phase relationships (Marder and Bucher, 2007). We focused on one network neuron, the LP, which exclusively expresses type 1 DA receptors (D1Rs) (Zhang et al., 2010) and normally functions to set the upper limit on pyloric cycle frequency (Weaver and Hooper, 2003). We demonstrate that steady-state nanomolar DA enables an activity-dependent mechanism that can contribute to LP phase maintenance by persistently increasing LP Ih Gmax. This mechanism (1) was triggered when there was a prolonged loss of LP network function, (2) may be evolutionarily conserved as loss of dopaminergic tone decreases Ih in rhythmically active globus pallidus neurons (Chan et al., 2011), and (3) may represent an adaptation to preserve motor network output with increasing dopaminergic neuronal activity.
The slow phase recovery mechanism
Application of tonic micromolar DA generated an LP phase advance. Phase recovered within 1 h despite the continued presence of DA. Two statistically significant changes in network activity occurred during the time window for phase recovery: an increase in pyloric cycle frequency and a decrease in LP burst duration. We did not investigate additional alterations in other network neurons, though all are directly modulated by micromolar DA (Flamm and Harris-Warrick, 1986b). The prolonged change in LP burst duration was necessary for phase recovery. We did not test whether the change in cycle frequency was also necessary. It is noteworthy that cycle period scales with LP burst duration across individuals (Goaillard et al., 2009), suggesting that a change in this ratio may trigger phase recovery. A decrease in LP burst duration elicited a persistent increase in LP Ih Gmax, but only in the presence of nanomolar DA. Phase recovery was always accompanied by a persistent increase in LP Ih Gmax, and changes in Ih were necessary for phase recovery. The most parsimonious interpretation of the data is that tonic nanomolar DA permitted a change in LP burst duration to trigger a persistent increase in LP Ih Gmax that was necessary for LP phase recovery in 5 μm DA.
It is possible that nanomolar DA coordinately regulates multiple currents in a given neuron to bring about phase recovery. Experimental and computational studies suggest that distinct conductance ratios are maintained in identified pyloric neurons, including the ratio of LP IA to Ih (Schulz et al., 2006, 2007; Hudson and Prinz, 2010), and modulators are necessary for their maintenance (Khorkova and Golowasch, 2007). Interestingly, steady-state DA also regulates LP IA Gmax (Rodgers et al., 2011).
Dopaminergic systems are organized to prevent disruptions in the circuits they modulate
The phase recovery mechanism may represent an adaptation to preserve motor network output with increasing dopaminergic neuronal activity. Dopaminergic systems use volume transmission, whereby phasically released DA diffuses out of open synapses (Descarries and Mechawar, 2000) and reuptake mechanisms are distantly distributed (Zoli et al., 1998), resulting in tonic DA in the extracellular space. Dopaminergic neuronal activity contributes to steady-state DA levels in conjunction with hormonal DA and DA transporter-mediated efflux (Borland and Michael, 2004; Moquin and Michael, 2011). As dopaminergic neuronal bursting increases, steady-state DA concentration should also increase, at least locally, thereby enabling the phase recovery mechanism. Continued dopaminergic neuron activity may then trigger the phase recovery mechanism by creating prolonged changes in network neuron activity. Thus, the 5 nm DA treatment group shows a persistent LP phase advance that does not recover over 4 h; whereas, prolonged micromolar DA triggers phase recovery over tens of minutes.
Phase maintenance in the pyloric network
Pyloric neuron phase relationships are partially determined by differential rates of PIR. Endogenous membrane currents strongly influence PIR (Rabbah and Nadim, 2005). IA and Ih decrease and increase the rate of PIR, respectively (Tierney and Harris-Warrick, 1992, 1995a). An unknown slow conductance or combination of conductances shapes PIR according to activity averaged over a period of seconds (Hooper et al., 2009; Goaillard et al., 2010). In addition to endogenous currents, PIR is also influenced by the strength and timing of synaptic inhibition onto follower neurons (Eisen and Marder, 1984; Rabbah and Nadim, 2005, 2007).
Phase is well maintained in the pyloric network. Injecting current into the pacemaker to change cycle period does not alter neuronal phase (Hooper, 1997a,b), largely due to the dynamics of intrinsic conductances and synaptic depression (Hooper, 1998; Manor et al., 2003; Bose et al., 2004; Greenberg and Manor, 2005). Phase is also maintained across temperatures, due, in part, to the Q10 values of the channels mediating Ih and IA (Tang et al., 2010). Our studies revealed an additional activity-dependent mechanism for phase maintenance that operated on a slow time scale (tens of minutes) and required dopaminergic tone.
Phase homeostasis?
A previous study demonstrated that neuronal phase was maintained in pyloric network output across individuals and throughout their lifetimes, prompting the authors to speculate that phase might be the target of a homeostat (Bucher et al., 2005). We did not prove the existence of a DA-enabled homeostat because bidirectional perturbations did not produce phase recovery (Davis, 2006). Perhaps the error signal detected by the homeostat sensor represents a complex set of changes that were not reproduced in all experiments. Alternatively, tonic DA may not enable a true homeostat, because phase is a network property whose maintenance requires coordinated changes across neurons. DA could provide this coordination by enabling multiple cell-specific compensatory mechanisms to explicitly redress the alterations induced by prolonged phasic DA.
An emergent idea in the field of activity-dependent homeostasis is that homeostatic mechanisms are active only at specific developmental stages and/or in certain cells (Turrigiano, 2011). Our findings might suggest that all cells possess a similar menu of recovery mechanisms that are differentially enabled by distinct modulators. Modulator enabling imparts flexibility; it allows a given neuron to maintain multiple states depending upon the circumstances. This would be beneficial in several instances, such as network switching (Hooper and Moulins, 1989; Dickinson et al., 1990) or when a motor circuit can generate two stable outputs.
Recovery mechanisms engender preparation-to-preparation variability and raise the question of how differentially compensated circuits maintain similar outputs when conditions change (Marder, 2011). Our data suggest that the correct extracellular milieu might permit further changes in synaptic strength (Grashow et al., 2010) and/or ionic conductances to restore network function within tens of minutes.
Ih function and plasticity
Hyperpolarization activated, cyclic nucleotide gated (HCN) cation channels mediate Ih (Wahl-Schott and Biel, 2009). They operate at subthreshold levels and contribute to membrane potential, dendritic integration and temporal summation (Magee, 1999; Poolos et al., 2002; Angelo et al., 2007). HCN channels are also critical determinants of membrane resonance properties in several neurons, including pyloric (Narayanan and Johnston, 2007, 2008; Tohidi and Nadim, 2009; Zemankovics et al., 2010). Neurons have frequency preferences that allow them to oscillate at, and best respond to, a particular frequency, termed the resonance frequency (Hutcheon and Yarom, 2000). Resonance frequency depends upon the low- and high-pass filtering properties of the membrane. High pass filters are generated by slowly activating currents that oppose changes in membrane potential, such as Ih (Hutcheon and Yarom, 2000).
Ih shows a high degree of plasticity. Chronic changes in activity, seizures and sensory deprivation can persistently alter Ih expression (Brewster et al., 2002; Gibson et al., 2006; Arimitsu et al., 2009; Hassfurth et al., 2009). Changes in Ih accompany LTP and LTD (van Welie et al., 2004; Fan et al., 2005; Brager and Johnston, 2007; Campanac et al., 2008), and it has been suggested that Ih is an effector for activity-dependent homeostatic plasticity (Narayanan and Johnston, 2010). Resonance frequency is variable in CA1 pyramidal neurons such that the increase in Ih that accompanied LTP increased resonance frequency, suggesting that neurons may tune their resonance frequencies to the inputs they receive (Narayanan and Johnston, 2007).
We hypothesize that the increase in LP Ih Gmax could retune LP resonance frequency to match an increase in pacemaker cycle frequency. The LP neuron receives rhythmic inhibition from the pacemaker, and provides the pacemaker with reciprocal feedback inhibition (Fig. 1C). Pyloric cycle frequency, which correlates with pacemaker resonance frequency (Tohidi and Nadim, 2009), is significantly increased in micromolar DA. LP may retune its resonance frequency to match the pacemaker to maintain network functionality, as the efficacy of LP feedback inhibition depends on pacemaker phase (Thirumalai et al., 2006). Indeed, LP cannot establish an upper limit on network cycle frequency in prolonged micromolar DA, because the phase relationship between LP and the pacemaker is distorted, causing LP synaptic inhibition onto the pacemaker to occur when the pacemaker is less sensitive to inhibition (Johnson et al., 2011). Alternatively changes in LP Ih may alter the strength of the PD synapse onto LP. Previous work in the pyloric circuit suggests Ih may act postsynaptically to control synaptic strength (Goeritz et al., 2011). Additionally, LP activity may be mostly synaptically driven, as it has a much slower burst frequency when isolated from pyloric pacemaker input (Bal et al., 1988).
Activity-dependent regulation of HCN channels occurs at multiple levels including transcription (Brewster et al., 2002; Fan et al., 2005), trafficking (Bender et al., 2007; Shin and Chetkovich, 2007; Hardel et al., 2008; Noam et al., 2010), and heteromerization (Zha et al., 2008). In dendrites, activity-dependent trafficking of HCN channels involves Ca2+ and CaMKII (Shin and Chetkovich, 2007; Noam et al., 2010). Depletion of Ca2+ stores and a concomitant increase in cytosolic Ca2+ can also increase HCN surface expression perisomatically (Narayanan et al., 2010). We have previously shown that steady-state nanomolar DA can regulate LP IA Gmax through a translation-dependent, activity-independent mechanism mediated by target of rapamycin (TOR) (Rodgers et al., 2011). Reductions in LP burst duration may modulate this pathway. Many changes in protein expression accompanying LTP and LTD rely on TOR and are triggered by DA and changes in activity (Hoeffer and Klann, 2010). Whether changes in LP Ih are local or global, and the molecular mechanisms involved, are important topics for future studies.
Footnotes
-
This work was supported by NIH Grant DA024039 to D.J.B. We thank Tim Dever for excellent technical assistance.
- Correspondence should be addressed to Deborah J. Baro at the above address. dbaro{at}gsu.edu
