Abstract
Elevated extracellular [K+] is associated with many disorders including epilepsy, traumatic brain injury, ischemia and kidney failure. Experimentally, elevated [K+] is used to increase excitability in neurons and networks, by shifting the potassium equilibrium potential (EK) and consequently, the resting membrane potential. We studied the effects of increased extracellular [K+] on the well-described pyloric circuit of the crab, Cancer borealis, while recording pyloric network activity extracellularly and the activity of Pyloric Dilator neuron (PD) intracellularly. A 2.5-fold increase in extracellular [K+] (2.5x[K+]) depolarized PD neurons and resulted in an unexpected short-term loss of their normal bursting activity. This period of silence was followed by the recovery of spiking and/or bursting activity during the continued superfusion of 2.5x[K+] saline. In contrast, when PD neurons were pharmacologically isolated from pyloric presynaptic inputs, they exhibited no loss of spiking activity in 2.5x[K+], suggesting the existence of an acute inhibitory effect mediated by circuit interactions. Action potential threshold in PD neurons decreased markedly over the course of exposure to 2.5x[K+] concurrent with the recovery of spiking and/or bursting activity. This study illustrates a case of rapid adaptation to a global perturbation that is influenced by local synaptic connections. Moreover, the complex response of pyloric neurons to elevated [K+] demonstrates that electrophysiological recordings are necessary to determine how neuronal and circuit activity are affected by altered K+ concentrations.
Significance Statement To characterize the sensitivity of a neuronal circuit to global perturbation, we tested the response of the well-described pyloric circuit of the crab stomatogastric ganglion to saline with elevated [K+]. Unexpectedly, a 2.5-fold increase in extracellular [K+] led to a temporary loss of activity in pyloric neurons that is not due to depolarization block. This was followed by a rapid increase in excitability and concurrent recovery of spiking activity within minutes. In contrast, when presynaptic inputs to pyloric neurons were blocked, there was no temporary loss of spiking activity in elevated [K+]. This is a case of rapid adaptation that restores neuronal activity disrupted by global depolarization.
Introduction
Neuronal circuits must be robust to various environmental challenges. This is especially true for central pattern generators (CPGs) that produce essential motor patterns such as breathing, walking, and chewing (Marder and Calabrese, 1996). Maintaining stability over a range of perturbations involves multiple intrinsic and synaptic mechanisms that operate across minutes to days (Von Euler, 1983; Marder and Bucher, 2001; Harris-Warrick, 2010). Complicating matters, both theoretical and experimental evidence suggest that robust CPGs with similar activity patterns can have widely variable underlying cell intrinsic and synaptic conductances (Prinz et al., 2004; Marder and Goaillard, 2006; Schulz et al., 2006; Schulz et al., 2007; Goaillard et al., 2009; Norris et al., 2011; Roffman et al., 2011). These individually variable circuits, nevertheless, must be reliable, and it remains an open question how such circuits respond and adapt to environmental challenges.
In the context of neuronal circuits, global perturbations involve changes to properties of the environment in which the neurons reside and thus may exert a wide influence over most circuit neurons. For instance, changes in temperature and pH will influence cellular function by altering how basic biochemical processes occur. Global perturbations also include changes to the ionic composition of extracellular fluid that then alter the electrochemical driving forces important for neuronal activity. In particular, elevated extracellular potassium concentration ([K+]), is a physiologically relevant depolarizing stimulus associated with a wide array of conditions including thermal stress, epileptic seizures, kidney failure, traumatic brain injury, and stroke (Baylor and Nicholls, 1969; Katayama et al., 1990; Pérez-Pinzón et al., 1995; Jensen and Yaari, 1997; Rodgers et al., 2007; Krishnan and Kiernan, 2009; Morrison et al., 2011; Arnold et al., 2014; Chauvette et al., 2016). Experimentally, increased extracellular [K+] is often used to induce changes in neuronal activity. For instance, in rodent hippocampal cultures, increased [K+] were used to depolarize neurons and activate calcium-dependent transcription pathways (Lin et al., 2008; Sharma et al., 2015). Increased [K+] is often applied to pre-Bötzinger complex neurons in acute slice preparations to reestablish rhythmic firing lost in the absence of excitatory inputs (Ballerini et al., 1999; Panaitescu et al., 2009; Ruangkittisakul et al., 2011; Rybak et al., 2014). Although our understanding of the basic depolarizing effect of increased [K+] on neuronal membrane potential is well defined (Somjen, 1979), the circuit level and long-term consequences of global changes in [K+] are less well understood. Here, we characterized the effects of increased [K+] on a well described motor circuit of the crab, Cancer borealis.
The crustacean stomatogastric nervous system (STNS) has been extensively studied and is a highly advantageous system for the study of fundamental mechanisms of circuit dynamics, pattern generation, and neuromodulation (Selverston, 1976; Selverston and Moulins, 1987; Marder and Bucher, 2007; Marder et al., 2014). The combination of intrinsic variability, well-established connectivity, and production of behaviorally relevant fictive activity, make the STNS an attractive model to study underlying network dynamics and robustness in response to a global perturbation (Selverston and Miller, 1980; Eisen and Marder, 1982; Miller and Selverston, 1982; Selverston et al., 1982). Previous studies have shown that the pyloric rhythm of the STNS is extraordinarily robust to changes in both temperature and pH (Tang et al., 2010; Tang et al., 2012; Soofi et al., 2014; Marder et al., 2015; Haddad and Marder, 2018; Haley et al., 2018; Kushinsky et al., 2019). Extracellular potassium concentrations can act on all neurons in a circuit simultaneously and affect a multitude of cellular processes, making it an attractive model with which to study the effects of a global perturbation on neural circuits (Somjen, 2002; Misonou et al., 2004).
Here, we characterize the effects of elevated [K+] on the pyloric rhythm and quantify the time course of these effects. We describe both some expected and unanticipated effects of the treatment.
Materials and Methods
Animals and dissections
Adult male Jonah Crabs, Cancer borealis, (N = 73) were obtained from Commercial Lobster (Boston, MA) between December 2016 and March 2019 and maintained in artificial seawater at 10-12°C in a 12-hour light/dark cycle. On average, animals were acclimated at this temperature for one week before use. Prior to dissection, animals were placed on ice for at least 30 min. Dissections were performed as previously described (Gutierrez and Grashow, 2009). In short, the stomach was dissected from the animal and the intact stomatogastric nervous system (STNS) was removed from the stomach including the commissural ganglia, esophageal ganglion and stomatogastric ganglion (STG) with connecting motor nerves. The STNS was pinned in a Sylgard-coated (Dow Corning) dish and continuously superfused with 11°C saline.
Solutions
Physiological Cancer borealis saline was composed of 440 mM NaCl, 11 mM KCl, 26 mM MgCl2, 13 mM CaCl2, 11 mM Trizma base, 5 mM maleic acid, pH 7.4-7.5 at 23°C (approximately 7.7-7.8 pH at 11°C). High - 1.5x, 2x, 2.5x, and 3x[K+] - salines (16.5, 22, 27.5 and 33mM KCl respectively) were prepared by adding more KCl salt to the normal saline formula. Picrotoxin (PTX) used to block glutamatergic synapses was added to normal or 2.5x[K+] saline at a 10-5M concentration for isolated pacemaker experiments (Marder and Eisen, 1984).
Electrophysiology
Intracellular recordings from somata were performed in the desheathed STG with 10–30 MΩ sharp glass microelectrodes filled with internal solution (10 mM MgCl2, 400 mM potassium gluconate, 10 mM HEPES buffer, 15 mM NaSO4, 20 mM NaCl (Hooper et al., 2015). Intracellular signals were amplified with an Axoclamp 900A amplifier (Molecular Devices, San Jose). Extracellular nerve recordings were made by building wells around nerves using a mixture of Vaseline and mineral oil and placing stainless-steel pin electrodes within the wells to monitor spiking activity. Extracellular nerve recordings were amplified using model 3500 extracellular amplifiers (A-M Systems). Data were acquired using a Digidata 1440 digitizer (Molecular Devices, San Jose) and pClamp data acquisition software (Molecular Devices, San Jose, version 10.5). For identification of Pyloric Dilator (PD) and Lateral Pyloric (LP) neurons, somatic intracellular recordings were matched to action potentials on the pyloric dilator nerve (pdn), lateral pyloric nerve (lpn) and/or the lateral ventricular nerve (lvn).
Elevated [K+] saline application
Prior to all applications of elevated [K+] saline, baseline activity was recorded for 30 minutes in physiological saline. Following the baseline recording, the entire preparation was superfused with elevated [K+] saline in concentrations ranging from 1.5x to 3x [K+] for 90 minutes. The preparation was then washed with physiological saline for 30 minutes. Recordings from PD neurons in the isolated pacemaker kernel were made by superfusing with 10-5M Picrotoxin (PTX) saline until all inhibitory synaptic potentials in the PD neurons disappeared (for at least 20 minutes). These preparations were then exposed to 2.5x[K+] PTX saline for 90 minutes, followed by a 30-minute wash in PTX saline.
Threshold and excitability measurements
To measure the action potential threshold and excitability of PD neurons, two-electrode current clamp was used to apply slow ramps of current from −4nA to +2nA over 60s. Resting membrane potential and input resistance were measured during both baseline conditions and after the application of elevated [K+] to ensure the integrity of the preparation (neurons with input resistances <4MΩ were discarded). Three ramps were performed during baseline at 10-minute intervals, and ramps were performed in 2.5x[K+] at 5, 10, 20, 30, 40, 50, 60, 70, 80 and 90 minutes after the start of elevated [K+] superfusion. After the preparation was returned to physiological saline, three ramps were performed again at 10-minute intervals. In recordings from the PD neurons with glutamatergic synapses blocked by PTX, baseline ramps were performed as described above, followed by three ramps in PTX saline. Preparations were then superfused with 2.5x[K+] PTX saline and washed in PTX saline following the same ramp procedure as above.
Data acquisition and analysis
Recordings were acquired using Clampex software (pClamp Suite by Molecular Devices, San Jose, version 10.5) and were visualized and analyzed using custom MATLAB waveform analysis scripts. These scripts were used to detect and measure voltage response amplitudes and membrane potentials, plot raw recordings and processed data, generate spectrograms, and perform some statistical analyses.
Spectral analysis
Spectrograms were calculated using the Burg (1967) method for estimation of the power spectrum density in each time-window. The Burg method (1967) fits the autoregressive (AR) model of a specified order p in the time series by minimizing the sum of squares of the residuals. The fast-Fourier transform (FFT) spectrum is estimated using the previously calculated AR coefficients. This method is characterized by higher resolution in the frequency domain than traditional FFT spectral analysis, especially for a relative short time window (Buttkus, 2000). We used the following parameters for the spectral estimation: data window of 3.2 s, 50% overlap to calculate spectrogram, number of estimated AR-coefficients p=window/4+1. Before the analysis, voltage traces were low pass filtered to 5 Hz using a six-order Butterworth filter and down-sampled. PD neuron burst frequency was calculated as the mean frequency at the peak spectral power in each sliding window.
Analysis of interspike interval distributions
Intracellular voltage traces were thresholded to obtain spike times. Distributions of inter-spike intervals (ISIs) were calculated within 2-minute bins. Hartigan’s dip test of unimodality (Hartigan and Hartigan, 1985) was used to obtain the dip statistic for each of these distributions. This dip statistic was compared to Table 1 in Hartigan and Hartigan (1985) to find the probability of multi-modality. The test creates a unimodal distribution function that has the smallest value deviations from the experimental distribution function. The largest of these deviations is the dip statistic. The dip statistic shows the probability of the experimental distribution function being bimodal. Larger value dips indicate that the empirical data are more likely to have multiple modes (Hartigan and Hartigan, 1985).
Activity pattern plots
For all recordings, we determined the time of spikes over the course of the experiment. For the recovery time plots, silence was defined as no more than 2 spikes in a 30-second sliding window. Otherwise, all spike behaviors (even if irregular) were counted as active spiking.
To more broadly determine the activity pattern of each PD neuron across the experiment, we analyzed the distribution of inter-spike intervals (ISI) in 2-minute bins using Hartigan’s dip statistic, as described above. If the dip statistic was 0.05 or higher the neuron was considered to be bursting. If the dip statistic was lower than 0.05 the neuron was considered to be tonically firing. Neurons with some spikes, but not enough ISIs to calculate the dip, were classified as sparsely firing. Neurons with no ISIs in the observed window were classified as silent. We then plotted the activity pattern of the neuron in these four categories – bursting, tonic, sparse firing and silent – for each PD neuron across the entire experiment.
Identification of the spike threshold
The spike threshold was identified as the voltage point of the maximum curvature before the first spike. Specifically, we calculated the first derivative of the voltage (dV/dt) and defined the spike onset as the point when dV/dt crosses the threshold value of 10 mV/ms.
All electrophysiology analysis scripts are available at the Marder lab GitHub (https://github.com/marderlab).
Results
Network activity in the pyloric circuit
The entire stomatogastric nervous system (STNS) of the crab Cancer borealis was isolated intact from the stomach and pinned in a dish, allowing us to change the composition of the continuously flowing superfused saline (Fig.1A). The stomatogastric ganglion (STG) contains identified neurons that drive the pyloric rhythm that filters food through the animal’s foregut. Figure 1B illustrates the stereotypical triphasic pyloric pattern which is comprised of the activity from lateral pyloric (LP), pyloric (PY) and pyloric dilator (PD). The rhythm is recorded extracellularly from motor axons contained in the lateral ventricular nerve (lvn) and other nerves. Figure 1C illustrates the connectivity diagram of the pyloric network. The anterior burster (AB) neuron, the intrinsic oscillator that drives the circuit, is strongly electrically coupled to two PD neurons, which burst synchronously with the AB neuron. Together the AB and two PD neurons form the pacemaker kernel of the network, and their coordinated burst of spikes initiates each triphasic cycle (Maynard, 1972). Synaptic connections between neurons in the STG are all inhibitory and are both graded and spike mediated (Graubard et al., 1980; Manor et al., 1997). Rhythmic inhibition from the pacemaker drives bursting activity resulting from post-inhibitory rebound in the LP neuron and PY neurons which reciprocally inhibit each other (Hartline and Gassie, 1979; Selverston and Miller, 1980). As a result, LP bursting and PY bursting compose the second and third phases of the pyloric rhythm respectively.
The pyloric rhythm is disrupted by high extracellular potassium
To test the response of the pyloric rhythm to changes in extracellular [K+], we switched from superfusion of normal physiological saline to superfusion of saline with elevated [K+] over the STNS while continuously recording the activity of pyloric neurons extracellularly from the lvn. We tested concentrations of [K+] that were 1.5, 2, 2.5 and 3-times physiological concentrations to study the dose-dependent responses of pyloric neurons to changes in extracellular [K+]. When extracellular [K+] was changed slightly, to 1.5x the physiological concentration (Fig. 1D, N = 4), the pyloric rhythm remained triphasic and was negligibly affected. When exposed to 2x [K+] (Fig. 1E, N = 20), the response of pyloric neurons was more variable; in some cases (N = 9/20) there was a short disruption of the triphasic pyloric rhythm, which was followed by the recovery of spiking activity.
Higher concentrations of extracellular [K+] produced more pronounced effects on pyloric activity. Superfusion of 2.5x [K+] (Fig. 1F, N = 12 extracellular recordings) reliably and profoundly altered the pyloric rhythm. During the application of 2.5x[K+] saline, all preparations exhibited a surprising disruption of action potentials from pyloric neurons, followed by recovery of spiking activity during continued exposure to 2.5x[K+] saline. Similarly, application of 3x[K+] saline to the STNS resulted in consistent cessation of the pyloric rhythm (Fig. 1G, N = 5). However, at 3x[K+], very few of the preparations recovered consistent activity. Based on these responses, we settled on a concentration of 2.5 times the control saline K+ concentration (2.5x[K+]) for further study, which reliably disrupted the pyloric rhythm and was accompanied by a consistent recovery of spiking or bursting activity during the continued application of 2.5x[K+].
The pattern of loss and recovery of pyloric activity in 2.5x[K+] saline was consistent across all experiments; however, the precise nature of each neuron’s response cannot be determined from extracellular data alone. Therefore, to obtain more detailed information on the effects of increased [K+], we recorded intracellularly from pyloric neurons while superfusing 2.5x[K+] saline over the STNS.
Pyloric neurons PD and LP depolarize and temporarily lose spiking activity in high extracellular potassium
With intracellular recordings, we saw a marked loss of spiking activity (crash) of pyloric neurons in response to the application of 2.5x[K+] saline that was consistent with the previously described extracellular recordings. In the representative example shown in Figure 2, the PD and LP neurons burst robustly in normal physiological saline (Fig. 2Ai). Within a few minutes of the start of 2.5x[K+] saline application, the minimum membrane potential of the PD and LP neurons depolarized by 15 and 22mV respectively (Fig. 2Aii), which was coincident with a reduction in firing frequency. The activity of both the LP and PD neurons became more burst-like over the course of the 90-minute application of 2.5x[K+] saline (Fig. 2Aiii-iv) and recovered to normal baseline behavior when returned to physiological saline (Fig. 2Av). The pattern of depolarization and recovery of spiking can be more clearly depicted by plotting time-condensed voltage traces for the PD neuron (the response of the LP neuron closely resembles that of the PD neuron) for the entire 150 minute experiment. In this trace, the membrane potential depolarizes in 2.5x[K+] saline followed by a loss of spiking activity (Fig. 2B). To visualize spiking behavior over the course of the experiment, we plotted the instantaneous interspike intervals (ISI) of the PD neuron for the whole experiment on a log scale (log10(ISIs), Fig. 2C). All healthy PD neurons in physiological saline have regular bursting activity that yields a bimodal distribution of ISIs, reflecting the relatively longer ISI period between bursts and the shorter ISIs of spikes within a burst. Although the initial depolarization in 2.5x[K+] saline caused a very brief increase in burst and spike frequency, the increase was immediately followed by a loss of all spiking activity (Fig. 2C). Over the course of the 2.5x[K+] saline application, both the PD and LP neuron recovered rhythmic bursts of action potentials, which is clear from the re-emergence of two ISI bands (Fig. 2C). Bursting activity is suggestive of the re-appearance of slow membrane potential oscillations. These slow oscillations are best visualized by spectrograms of the neuron’s membrane potential; recovery of bursting activity in elevated [K+] saline can be seen by the re-appearance of a strong frequency band in the voltage spectrogram (Fig. 2D). We then calculated the most hyperpolarized point of the membrane potential in each burst averaged over five-minute bins for all PD neurons to determine the overall depolarizing effect of 2.5x[K+] saline. Individual PD neurons depolarized upon application of 2.5x[K+] saline and remained depolarized throughout the application with no repolarization of the membrane potential (Fig. 2E). Over all preparations, PD neurons depolarized upon application of 2.5x[K+] saline (Fig. 2F, repeated measures ANOVA, Tukey post-hoc p<0.05, average depolarization after 10 minutes 14.5 ± 3.3mV) After the initial change in the first 10 minutes in 2.5x[K+], the minimum membrane potentials did not change for the remainder of the elevated [K+] application (p > 0.05). The minimum membrane potential returned to baseline levels when the preparations were returned to physiological saline (n.s., p > 0.05). The behavior of LP neurons in 2.5x[K+] was very similar to that of PD neurons; LP neurons depolarized by 16.5 ± 2.9mV after 10 minutes in 2.5x[K+] (n = 5) and remained depolarized for the duration of the elevated [K+] application.
Variability in the response of PD neurons to 2.5x[K+] saline
Although pyloric activity and circuit connectivity is highly conserved across animals, responses of individual PD neurons to 2.5x[K+] saline varied substantially across preparations. Superfusion of 2.5x[K+] saline led to a period of silence in 9 of the 13 PD neurons, with striking variability in the duration of silence and extent of recovery across animals. Figure 3 shows the responses of four PD neurons from four different animals to a 90-minute application of 2.5x[K+] saline. Across all preparations (n = 13), the time of silence elicited by 2.5x[K+] saline application varied from 1 to 62 minutes (time of silence was 10.9 ± 5.8 minutes SD). We characterized the recovery of PD neuron activity by comparing the ISI distributions over time. In some cases, the PD neurons exhibited only tonic firing activity in 2.5x[K+] saline which is reflected by the presence of a single ISI band (Fig. 3A, B). In other cases, PD neurons regained burst-like activity, which was reflected in the re-emergence of two ISI bands in 2.5x[K+] saline (Fig. 3C, D).
In all PD neuron recordings, the variable period of silence upon application of 2.5x[K+] saline was followed by the recovery of spiking activity. For all PD neurons, we calculated the “time to recovery,” defined as the length of time between the silencing of the neuron and when the neuron recovered at least two action potentials in a 30-second sliding window during the application of 2.5x[K+] saline. We were interested in determining whether aspects of baseline activity of each PD neuron influenced the neuron’s time to recovery. Therefore, for each PD neuron, we calculated the mean minimum membrane potential during baseline recordings (Fig. 3E), the change in membrane potential upon application of 2.5x[K+] saline (Fig. 3F) and the baseline bursting frequency (Fig. 3G) and compared these values to the time to recovery for each corresponding neuron. We found no correlation between any of these values and the time elapsed until recovery of spiking activity (R2=0.203, R2=0.104, R2=0.012 respectively).
PD neurons in the isolated pacemaker kernel continue spiking in 2.5x[K+] saline
From the previous experiments, it was unclear to what extent the crash and recovery of activity in 2.5x[K+] saline was due to presynaptic inputs to PD neurons. Because the PD and AB neurons receive only glutamatergic input from other pyloric network neurons, the pacemaker kernel can be studied in isolation from the pyloric network neurons by superfusing saline with 10-5M picrotoxin (PTX), which blocks ionotropic glutamatergic synapses in the STG (Fig. 4A) (Bidaut, 1980).
The response of PD neurons in the presence of PTX (PTX(+)) to 2.5x[K+] saline was markedly different from the behavior of PD neurons in control conditions (in the absence of PTX) (PTX(-)). Figure 4 illustrates a representative example of the responses of PD neurons to 2.5x[K+] PTX saline; the preparation initially switched from rhythmic bursting to tonic spiking activity in 2.5x[K+] PTX saline (Fig. 4Biii), followed by recovery of bursting activity that became more pronounced with time (Fig. 4B iv – vi). There was no interruption of (firing) activity upon the superfusion of 2.5x[K+] PTX saline (Fig. 4C) in PD neurons as compared to 25.x[K+] alone. The recovery of bursting in 2.5x[K+] PTX saline can also be seen in the emergence of two distinct ISI bands (Fig. 4D) and the emergence of a robust frequency band in the spectrogram of the PD intracellular voltage trace (Fig. 4E).
We further quantified the response of PD neurons to 2.5x[K+] PTX saline by calculating the mean minimum membrane potential in five-minute bins for each neuron across the experiment (n = 8, Fig. 4F). Similar to PD neurons in the intact circuitry, all PTX(+) PD neurons depolarized in 2.5x[K+] saline (Fig. 4G, repeated measures ANOVA, Tukey post-hoc p < 0.05, average depolarization after 10 minutes 12.6 ± 3.0mV). The minimum membrane potential of PD neurons in PTX then remained stable during the application of 2.5x[K+] and returned to baseline levels when returned to physiological extracellular [K+] (Fig. 4G, p > 0.05). Importantly, the fact that PTX(+) PD neurons maintain tonic firing upon the application of 2.5x[K+] saline despite marked depolarization indicates that the crash observed in PTX(-) PD neurons is unlikely to be due to depolarization block.
Synaptic inputs alter response of PD neurons to 2.5x[K+] saline
The initial responses of PD neurons to 2.5x[K+] saline application differed in the presence or absence of PTX. This difference indicates the existence of a circuit-driven response to elevated [K+]. To quantify this effect, we used values from the Hartigan’s dip test on 2-minute bins of log(ISI) to determine the time that each PD neuron was either bursting, tonically firing, or silent throughout the experiment and plotted the activity of each PD neuron over time (see methods).
In the intact circuit, the majority (N = 9 of 13) of PD neurons exhibited a period of silence following the application of 2.5x[K+] saline, then recovered spiking activity over a variable amount of time (Fig. 5A). In 2.5x[K+] PTX saline, PD neurons either remained active or only briefly went silent, then demonstrated robust recovery of spiking or bursting activity (Fig. 5B). The average time of PD silence in 2.5x[K+] saline was significantly different with the addition of PTX (Fig. 5C, Wilcoxon rank-sum test p = 0.013), as was the average time of PD sparse firing between PTX(+) and PTX(-) PD neurons (Fig. 5D, Wilcoxon rank-sum test p = 0.034). In both the time of silence and sparse firing comparisons, the difference between PTX(+) and PTX(-) PD neurons was still significant when the largest point in the PTX(-) group was removed. Neither the average time of tonic firing (Fig. 5E, n.s., p = 0.77), nor the average time of burst firing significantly were significantly different in the presence or absence of PTX (Fig. 5F, n.s., p = 0.51).
Excitability of PD neurons changes rapidly during exposure to 2.5x[K+] saline
We wished to see whether periods of silence in high K+ were due to depolarization block. Therefore, we applied 60 second steady ramps of current from - 4nA to +2nA to PD neurons at several time points during the period of silence elicited by application of 2.5x[K+] saline. In silent PD neurons, action potentials could always be induced by injecting positive current, indicating that this period of silence is not due to depolarization block (Fig. 6A, ramps at 5-minutes and 10-minutes).
To determine the excitability of PD neurons during application of 2.5x[K+], we repeated the slow current ramps from −4nA to +2nA at 5, 10, 20, 30, 40, 50, 60, 70, 80, and 90 minutes after the beginning of the 2.5x[K+] saline application in the presence or absence of PTX. In the representative example shown, there was a clear change in the number and frequency of spikes elicited in the PD neuron by the current ramp as a function of time in 2.5x[K+] saline (Fig. 6A). In addition, as time in 2.5x[K+] increased, more spikes were elicited at the same membrane potentials during the current ramp (Fig. 6B).
For each individual PTX(-) preparation (N = 5), spike threshold became more hyperpolarized in the PD neuron as a function of time in 2.5x[K+] saline (Fig. 6C). In contrast, when the same current ramps were applied to PTX(+) PD neurons (N = 6), there was no clear trend in spike threshold as a function of time in 2.5x[K+] saline (Fig. 6D). Across all PD neurons, we calculated the percent change in spike threshold from baseline for each current ramp. There was a significant drop in spike threshold over time for PD neurons in 2.5x[K+] saline between the 5-minute the 60 through 90-minute time points and also between the 10-minute and the 60 through 90-minute time points (Fig. 6F blue, one-way repeated measures ANOVA, Tukey Post-Hoc test, all p < 0.05). For most of these neurons the biggest changes in firing rates and spike threshold occurred during the first 10 minutes of 2.5x[K+] saline application, which is similar to the time in which most PD neurons recover spiking activity in elevated [K+]. Across all PTX(+) PD neurons there was no significant percent change in spike threshold over time in 2.5x[K+] (Fig. 6G red, one-way repeated measures ANOVA, Tukey Post-Hoc test, n.s., all p > 0.05).
To visualize overall changes in PD neuron excitability, we then calculated average F-I curves for all neurons at the beginning (5-minute ramp) and end (90-minute ramp) of the 2.5x[K+] saline application. Control PD neurons become more excitable as time in 2.5x[K+] increases; there is a shift in the average F-I curve between the 5- and 90-minute time points (Fig. 6F solid lines, two-way repeated measures ANOVA, p = 0.009). In addition, at the 5-minute time point in 2.5x[K+], PTX(+) PD neurons are more excitable than PTX(-) PD neurons (Fig 6F purple lines, two-way repeated measures ANOVA, p = 0.036). This initial difference in excitability is consistent with the fact that PTX(+) neurons remain active upon application of 2.5x[K+], while PTX(-) PD neurons typically lose spiking activity. The excitability of PTX(+) PD neurons does not change over time in 2.5x[K+]; there was no significant difference between the 5- and 90-minute PTX(+) PD neuron F-I curves (Fig 6F dashed lines, two-way repeated measures ANOVA, p = 0.097). In addition, by the end of the 2.5x[K+] application (90-minutes), there was no significant difference between the F-I curves of PTX(+) and PTX(-) PD neurons (Fig 6F red lines, two-way repeated measures ANOVA, p = 0.375).
Discussion
Surprising inhibitory effect of a depolarizing stimulus
In the STG, the pyloric rhythm is robust to multiple perturbations including changes in temperature, pH and neuromodulatory state (Tang et al., 2010; Tang et al., 2012; Temporal et al., 2014; Hamood et al., 2015; Haddad and Marder, 2018; Haley et al., 2018). Here, we tested whether pyloric neurons are similarly robust to global depolarization through changes to extracellular [K+]. It is generally assumed that positive current input or depolarization of a neuron’s membrane potential will lead to an increase in neuronal activity, and that extreme depolarizations can lead to loss of activity through depolarization block. Instead, we found a type of transient neuronal silence in response to elevated [K+] that is not due to a depolarization block.
In the pyloric circuit, all local synaptic connections are inhibitory (Eisen and Marder, 1982; Miller and Selverston, 1982). Synaptic transmission in the STG is both graded and spike-mediated, meaning that action potentials are not required for inhibitory synapses to function (Graubard et al., 1980; Manor et al., 1997; Nadim et al., 1997). Therefore, the observed decrease in both bursting and spiking activity in elevated extracellular [K+] may be caused by global depolarization leading to an increase in graded inhibition that suppresses spiking and bursting activity. In support of this theory, elevated [K+] does not have the same inhibitory effect on PD neurons with glutamatergic synapses blocked compared to PD neurons with intact synaptic connections. Similarly, in the proprioceptive neurons of the blue crab and the muscle receptor organ of the crayfish, it was recently reported that increased [K+] also has an inhibitory effect at concentrations not thought to cause a depolarization block (Malloy et al., 2017).
Adaptation to global perturbation
The PD/AB pacemaker unit of the pyloric circuit exhibited rapid adaptation to the disruptive stimulus of increased extracellular [K+]. Although the triphasic pyloric rhythm was not fully restored in 2.5x[K+] saline, PD neurons exhibited rapid changes in excitability over several minutes, which corresponded to the recovery of spiking and, in many cases, bursting activity.
Classical homeostatic plasticity takes place over hours to days and requires changes in intracellular calcium concentrations which then subsequently drive changes in gene expression, leading to cell-intrinsic changes in excitability (Desai et al., 1999; Cudmore and Turrigiano, 2004; Turrigiano, 2012). There is evidence that this mechanism can be induced by changes in extracellular [K+]. Rat myenteric neurons cultured in elevated [K+] serum for several days exhibit long-lasting changes in Ca2+ channel function (Franklin, 1992). Similarly, culturing rat hippocampal pyramidal cells in high [K+] medium for several days leads to activation of calcium-dependent changes in input resistance and the resting membrane potential that regulates the intrinsic excitability of the neurons (O’Leary et al., 2010). Changes in K+ channel densities are also associated with homeostatic regulation of neuronal activity. Depolarization of crustacean motor neurons with current pulses for several hours alters K+ channel densities in a cell-specific manner through a calcium-dependent mechanism (Golowasch et al., 1999). In addition, cerebellar granule cells in which inhibitory GABA receptors are blocked for several days maintain their response to excitatory input by strengthening voltage-independent K+ conductances (Brickley et al., 2001). Adaptation to global perturbation over several hours to days is well described by models via calcium signals that influence expression levels of ion channels (O’Leary et al., 2014; O’Leary, 2018). Pyloric neurons in the STG also exhibit these long-term adaptations; preparations in which neuromodulatory inputs are removed initially lose rhythmicity and gradually recover over the course of several days (Thoby-Brisson and Simmers, 1998, 2002; Luther et al., 2003; Gray and Golowasch, 2016; Gray et al., 2017).
Intriguingly, many cases of rapid adaptation in neuronal circuits are related to changes in signaling between neurons. At the Drosophila neuromuscular junction, blocking glutamatergic signaling results in long-lasting changes in synaptic strength within minutes (Frank et al., 2006; Müller and Davis, 2012; Müller et al., 2012). In the crustacean cardiac ganglion, blocking delayed rectifier K+ currents led to adaptation of the circuit and an increase in electrical coupling strength within one hour (Lane et al., 2016). Although these rapid changes in synaptic strength are often associated with Hebbian plasticity and learning, rapid alterations in synapses can also lead to a homeostatic-like maintenance of circuit activity.
The rapid adaptation of PD neurons in elevated extracellular [K+] is most likely due to a combination of cell-intrinsic conductance changes and synaptic modulation. In crustacean motor neurons, cell-intrinsic changes in current densities have been shown to be modulated by second-messenger kinase pathways activated by global depolarization (Ransdell et al., 2012). Rapid changes in circuit state could also be influenced by neuromodulation, as these experiments were conducted with the entire STNS, including connections to the upstream modulatory ganglia which were also exposed to the elevated [K+]. In the STNS, the upstream commissural ganglia and the esophageal ganglion release a wide range of neuromodulators onto the STG that affect excitability of cells and the pyloric rhythm (Marder and Bucher, 2007; Marder, 2012). In particular, proctolin released by the MPN and MCN1 neurons endogenously drives the pyloric pacemaker neuron AB (Hooper and Marder, 1987; Nusbaum and Marder, 1989; Wood et al., 2000; Stein et al., 2007). Dopamine is also released by modulatory neurons in the STNS, and exogenous application of dopamine can cause rapid changes in electrical coupling strength and has been shown to rapidly regulate potassium currents in both mammalian and invertebrate neurons (Tornqvist et al., 1988; Harris-Warrick et al., 1998; Gruhn et al., 2005; Rodgers et al., 2013).
Variable responses to similar perturbations
The effects of perturbations are often represented by averaging the experimental results from a number of individuals, but there are several reasons why this approach is not always appropriate. Circuits with apparently identical outputs under control conditions can present distinctly different responses to perturbation due to underlying differences in network parameters (Tang et al., 2012; Hamood and Marder, 2014; Haddad and Marder, 2018; Haley et al., 2018; Alonso and Marder, 2019). Indeed, in this study we observed a wide range of responses in identified PD neurons to the same experimental conditions, 2.5x[K+], which may uncover individual differences that are not evident during control conditions. Within the STG, conductance densities and strengths of synaptic connections can vary up to five-fold in magnitude between individuals (Schulz et al., 2006; Schulz et al., 2007; Goaillard et al., 2009; Shruti et al., 2014; Temporal et al., 2014). Variability in neuronal conductances underlying similar activity patterns has also been demonstrated across phyla (Swensen and Bean, 2005; Nelson and Turrigiano, 2008; Tran et al., 2019). In addition, computational modeling of the pyloric network revealed that multiple combinations of independent circuit parameters can give rise to functionally identical activity patterns (Goldman et al., 2001; Prinz et al., 2004; Taylor et al., 2009; Marder et al., 2015). The application of elevated [K+] saline to identified STG neurons provides additional evidence that differences in individual conductance parameters influence circuit responses to global perturbation (Alonso and Marder, 2019).
Implications for disease states
Hyperkalemia is well documented in many nervous system disorders; however, it remains unclear how the changes in [K+] acutely affect neuronal activity. Chronic Kidney Disease (CKD) leads to increases in serum [K+] up to three times normal levels, directly affecting neuronal excitability, and changes in neuronal properties (Krishnan and Kiernan, 2009; Arnold et al., 2014). Similarly, increased activity of a group of neurons can increase the extracellular [K+] in the surrounding tissue (Baylor and Nicholls, 1969; Kříž et al., 1974) and epileptic seizures and brain trauma can lead to increases in [K+] in surrounding brain regions (Moody et al., 1974; Katayama et al., 1990; Silver and Erecinska, 1994; Fröhlich et al., 2008). These transient changes in extracellular [K+] have been tied to long lasting changes in the organization and phosphorylation pattern of K+ channels (Misonou et al., 2004), which could lead to long-lasting changes in circuit function (Somjen, 2001, 2002; Rodgers et al., 2007).
In this study, pyloric neurons respond to elevated [K+] with a rapid adaptation of excitability; another example of how the such adaptations occur. The fast adaptation of pyloric neurons to global changes in [K+] also suggests that rapid swings in [K+] may be more damaging to neurons than gradual shifts or sustained changes in concentration.
Reassessing global perturbation
The results of this study highlight the unexpected complexity of seemingly simple perturbations. In this classical manipulation of increased extracellular [K+] we observed a paradoxical decrease in the activity of PD neurons upon bath application of high [K+], followed by a recovery of activity in a short period time. Despite knowing the network connectivity, circuit properties and behavior of identified neurons within the STG, we were still unable to predict or fully explain the effects of increased [K+] on circuit performance. The complex interaction between circuit level effects and cell intrinsic responses to simple changes in ion concentrations underscores the importance of recording and reporting neuronal activity during such manipulations in any experiment.
Acknowledgments
Supported by NSF Grant R35 NS097343. The authors would like to thank Daniel Shin for assistance with dissections, and Stephen Van Hooser for statistical advice, and Sonal Kedia and Jason Pipkin for careful reading of this manuscript.