4-Aminopyridine-sensitive outward currents in preBötzinger complex neurons influence respiratory rhythm generation in neonatal mice

Abstract

We measured a low-threshold, inactivating K⁺ current, i.e. A-current (I_A), in respiratory neurons of the preBötzinger complex (preBötC) in rhythmically active slice preparations from neonatal C57BL/6 mice. The majority of inspiratory neurons (21/34 = 61.8%), but not expiratory neurons (1/8 = 12.5%), expressed I_A. In whole-cell and somatic outside-out patches I_A activated at −60 mV (half-activation voltage measured −16.3 mV) and only fully inactivated above −40 mV (half-inactivation voltage measured −85.6 mV), indicating that I_A can influence membrane trajectory at baseline voltages during respiratory rhythm generation in vitro. 4-Aminopyridine (4-AP, 2 mm) attenuated I_A in both whole-cell and somatic outside-out patches. In the context of rhythmic network activity, 4-AP caused irregular respiratory-related motor output on XII nerves and disrupted rhythmogenesis as detected with whole-cell and field recordings in the preBötC. Whole-cell current-clamp recordings showed that 4-AP changed the envelope of depolarization underlying inspiratory bursts (i.e. inspiratory drive potentials) from an incrementing pattern to a decrementing pattern during rhythm generation and abolished current pulse-induced delayed excitation. These data suggest that I_A opposes excitatory synaptic depolarizations at baseline voltages of approximately −60 mV and influences the inspiratory burst pattern. We propose that I_A promotes orderly recruitment of constituent rhythmogenic neurons by minimizing the activity of these neurons until they receive massive coincident synaptic input, which reduces the periodic fluctuations of inspiratory activity.

Rhythmic motor behaviours originate from central pattern generator (CPG) networks in the brainstem and spinal cord (Marder, 2001). A key issue is to what degree proper network function (i.e. rhythmogenesis) depends on specific ion channels in constituent rhythm-generating neurons (Stein, 1997). The respiratory CPG is an excellent model for examining this question because its constituent rhythmogenic neurons are contained within the preBötzinger complex (preBötC) (Smith et al. 1991; Feldman & Del Negro, 2006) and the network output is measurable in vitro. Transverse medullary slices containing the preBötC spontaneously generate behaviourally relevant motor activity that can be monitored via the hypoglossal nerve (XII).

Most studies of respiratory rhythm generation have focused on the role of voltage-dependent inward currents (Mironov & Richter, 1998; Pierrefiche et al. 1999; Mironov et al. 2000; Thoby-Brisson et al. 2000; Del Negro et al. 2001, 2002a,b, 2005; Onimaru et al. 2003; Pena et al. 2004; Ptak et al. 2005; Pace et al. 2007b), neuromodulation (Johnson et al. 1996; Rekling et al. 1996b; Onimaru et al. 1998; Shao & Feldman, 2000; Pena & Ramirez, 2002, 2004; Ruangkittisakul et al. 2006), as well as excitatory and inhibitory synaptic currents (Greer et al. 1991; Funk et al. 1993, 1995; Brockhaus & Ballanyi, 1998; Pierrefiche et al. 1998; Shao et al. 2003; Paarmann et al. 2005; Pace et al. 2007a). Apart from an ATP-inhibited K⁺ current primarily activated during hypoxia (Pierrefiche et al. 1996; Mironov et al. 1998, 1999; Mironov & Richter, 2000, 2001; Haller et al. 2001a,b), K⁺ currents have not been well-characterized in the preBötC of neonatal rodents, nor have their contributions to rhythmogenesis been analysed.

Transient K⁺ currents are important in invertebrate CPGs and the primitive vertebrate lamprey CPG for swimming, where they control the sequence of cell activation during the behavioural patterns and influence spike timing and frequency (Getting, 1983; Tierney & Harris-Warrick, 1992; Hess & El Manira, 2001).

In the respiratory CPG, Rekling et al. (1996a) described a subset of inspiratory neurons that depolarized with a ramp-like trajectory and started spiking ∼400 ms prior to XII output, dubbed type 1 neurons, which are putatively rhythmogenic (Rekling & Feldman, 1998; Gray et al. 1999). Type 1 neurons held at hyperpolarized membrane potentials exhibited delayed excitation in response to 400 ms step pulses of depolarizing current (Rekling & Feldman, 1998). Delayed excitation is often attributed to transient K⁺ currents (i.e. A-currents, I_A) (Hagiwara et al. 1961; Getting, 1983; Dekin & Getting, 1987; Dekin et al. 1987; Nisenbaum et al. 1994), thus Rekling and colleagues proposed that rhythmogenic preBötC neurons expressed I_A, although they did not measure it in voltage clamp nor speculate on its role in rhythm generation (Rekling et al. 1996a; Rekling & Feldman, 1998). Inyushkin (2005) confirmed that preBötC neurons expressed I_A using voltage clamp, but stopped short of analysing its contributions to rhythmogenesis.

Here we studied preBötC neurons that became active preceding inspiratory bursts, which we refer to as ‘early inspiratory’ to distinguish them from more rostral ‘pre-inspiratory’ neurons (Onimaru et al. 2006), and sought to measure I_A and evaluate its role in rhythm generation.

Methods

The Institutional Animal Care and Use Committee at The College of William and Mary approved all protocols. Neonatal (P0–7) C57BL/6 mice were anaesthetized via hypothermia until mice lacked a tail-pinch response. Mice were rapidly decerebrated and then dissected. Transverse slices (550 μm thick) containing the preBötC and hypoglossal (XII) nerves were sectioned with a vibrating microtome from the medulla oblongata. The rostral cut captured the rostral-most XII nerves, the dorsomedial cell column and principal lateral loop of the inferior olivary nucleus while the caudal cut captured the obex.

Slices were perfused at 26–28°C with an artificial cerebrospinal fluid (ACSF) containing (mm): 124 NaCl, 9 KCl, 0.5 NaH₂PO₄, 25 NaHCO₃, 30 d-glucose, 1.5 CaCl₂.2H₂O, and 1 MgSO₄. We identified putatively rhythmogenic inspiratory preBötC neurons (see first section in Results, below) with 9 mm K⁺ in the ACSF to maintain rhythmic network function, and then switched to 3 mm K⁺ to isolate and measure the properties of I_A because the lower [K⁺] more closely matches the in vivo milieu. We acquired data from a total of 65 inspiratory neurons, 8 expiratory neurons, and 4 field-recordings in a total of 77 slices and 77 animals.

Most voltage- and current-clamp experiments were performed with a HEKA EPC-10 patch-clamp amplifier and Patchmaster software (Lambrecht, Germany). Dose–response experiments were performed with a Model 2400 patch-clamp amplifier (A-M Systems, Sequim, WA, USA) using Chart 5 software and a Powerlab 8/30 (AD Instruments, Colorado Springs, CO, USA) for stimulation. The remaining voltage-clamp experiments utilized a LabJack U3 (LabJack Corporation, Lakewood, CO, USA) as a waveform generator commanding the voltage-clamp amplifier controlled with custom C/C⁺⁺ software written for a G4 Powerbook (Apple Inc., Cupertino, CA, USA). Respiratory-related motor output was monitored from XII nerves with extracellular suction electrodes and a high-gain differential amplifier with band-pass filtering (0.3–1 kHz) (Dagan Instruments, Minneapolis, MN, USA). Raw XII activity was conditioned using a true RMS-to-DC converter (Analog Devices, One Technology Way, Norwood, MA, USA) to provide a full-wave rectified and smoothed XII waveform. Data were acquired digitally and analysed using Chart 5, Igor Pro 5 (WaveMetrics, Lake Oswego, OR, USA), Excel (Microsoft, Redmond, WA, USA) and custom software. An 8 mV liquid junction potential was corrected online in both current- and voltage clamp.

Whole-cell capacitance (C_M) was measured using 50 ms voltage steps from −60 mV to command potentials from −75 mV to −65 mV in a 10-step sequence. Charge (Q) was computed by integrating leak-subtracted capacitative current (ΔQ =∫I_C) and C_M was calculated from C_M=ΔQ/ΔV. Series (access) resistance (R_S) was monitored throughout voltage-clamp recordings according to the Thevenin equivalent circuit, which allows R_S to be calculated from the decay time constant (τ_m) in response to small voltage steps with R_S=τ_m/C_m as long as R_S was much less than the input resistance (R_N). We monitored R_N via P/N online leak protocols. To avoid voltage-clamp errors we discarded experiments in which R_S > 0.1 R_N. We compensated for R_S in whole-cell using analog feedback circuitry within the EPC-10 as much as possible without causing clamp oscillations that jeopardize stable recording. We rechecked R_S and R_N to assess voltage-clamp viability before running sequences of episodic protocols, ensuring the reliability of the acquired data.

We used the following standard patch solution containing (mm): 140 potassium gluconate, 5 NaCl, 0.1 EGTA, 10 Hepes, 2 Mg-ATP, and 0.3 Na(3)-GTP. KOH was used to equilibrate pH at 7.2. To isolate I_A in voltage clamp (Figs 1B, 2B, 3, 5 and 7A) and test for delayed excitation as in Figs 1C and 2C, we used a low Ca²⁺–high Mg²⁺ extracellular ACSF containing (mm): 124 NaCl, 3 KCl, 25 NaHCO₃, 30 d-glucose, 0.5 CaCl₂·2H₂O, 2 MgSO₄, 0.001 TTX, and 0.2 CdCl₂.

Figure 1. Phenotypic behaviours of inspiratory neurons located in the preBötC.

A, current-clamp recording showing early inspiratory activity that precedes the integrated XII nerve (∫XII) output in 9 mm[K⁺]_o, a hallmark property of rhythmogenic neurons. B, voltage-clamp recording from a holding potential of −100 mV illustrating the transient K⁺ currents evoked at depolarized membrane potentials. Inset, a different neuron showing transient K⁺ currents evoked from a holding potential of −60 mV. C, the same neuron as in A and B illustrating voltage-dependent delayed excitation where ΔI_app was 95 pA.

Figure 2. Phenotypic behaviours of expiratory neurons located in the preBötC.

A, current clamp showing tonic spiking during the expiratory phase that is inhibited during ∫XII output in 9 mm[K⁺]_o, a hallmark property of expiratory neurons. B, the corresponding voltage-clamp recording shows only minimal transient K⁺ currents. C, delayed excitation is not exhibited by the expiratory neuron where ΔI_app was 379 pA.

Figure 3. Isolation of I_A in inspiratory neurons.

A, voltage-clamp recording from a holding potential of −100 mV illustrating the transient K⁺ currents evoked at depolarized membrane potentials. B, voltage-clamp recording from a holding potential of −40 mV illustrating the lack of significant transient K⁺ currents evoked at depolarized membrane potentials. C, I_A was isolated by subtracting the current traces in B from A while the voltage-clamp protocols are shown superimposed.

Figure 5. Biophysics of I_A in inspiratory neurons.

A, the voltage dependence of activation measured from difference currents evoked by depolarizing step commands delivered from −100 mV and −40 mV in an outside-out patch. The voltage-clamp protocols are shown superimposed and alternating traces were eliminated to facilitate illustration. B, steady-state activation curve (m_∞) and inactivation curve (h_∞) from outside-out patches (top). The time constant of inactivation (τ_h) as a function of voltage (bottom).

Figure 7. 4-Aminopyridine attenuates I_A and blocks delayed excitation.

A, difference currents evoked from −100 and −40 mV in somatic outside-out patches in the presence of 1 μm TTX, 200 μm Cd²⁺ and 3 mm extracellular [K⁺] before and after application of 2 mm 4-AP. Two activation steps are shown: −80 mV and +30 mV. B, dose–response curve of I_A in outside-out patches. C, in a different cell from A, 2 mm 4-AP abolished delayed excitation in current clamp. This neuron is also analysed in Fig. 9. D, the effects of 4-AP on pulse-induced spikes averaged over 4 neurons.

We measured the voltage dependence and kinetics of I_A using Fitmaster software by HEKA (Lambrecht, Germany) and Igor Pro. Activation and inactivation functions took the form:

where x_∞ reflects voltage-dependent steady-state activation (m_∞) or inactivation (h_∞), θ_x is the membrane potential of half-activation (θ_m) or half-inactivation (θ_h), and σ_x is the slope factor.

We computed the time course of I_A using recorded voltage trajectories from current clamp (Fig. 6) and the chord conductance equation I_A=g_Am_A(∞)h_A (V–E_K). E_K was −71 mV to simulate in vitro conditions, the voltage-dependent parameters matched the values from Results, and dh_A/dt = (h_A(∞)–h_A)/τ_h(V), with τ_h(V) = (202–0.42·V) (fitted empirically, see Fig. 5B). The differential equation was integrated using the 4th order Runge–Kutta method in custom C/C⁺⁺ software run on Apple Macintosh G5 computers under OS 10.4. Integration step size was 0.25 ms to match the 4 kHz experimental sampling rate.

Figure 6. Simulating the role of I_A.

A, the membrane trajectory from Fig. 1A. B, the time course of I_A as computed from Hodgkin–Huxley-style equations (see Methods). Inset, a detail of the calculated I_A with expanded y-axis emphasizing the low magnitude (1–2 pA) of the current during the interburst interval. C, the time course of the computed h_A.

Inspiratory drive potentials were analysed using the Peak Parameters extension in Chart software. Leading and trailing slopes of the drive potential were calculated from digitally smoothed traces that minimize spikes but preserve the underlying drive potential characteristics (Pace et al. 2007b). Peak amplitude and baseline were automatically detected and the leading slope was computed from 20% of peak amplitude to 80% peak amplitude and trailing slope is calculated from 80% to 20%. Cycle-triggered averages were generated with custom software written in the Python programming language.

Sample means were generally compared using t test, or Fisher Exact test where indicated. Mean values are reported with standard error (mean ±s.e.m.) and significance was set at a P value of 0.05.

Results

Inspiratory preBötC neurons express I_A

Rhythmogenic neurons can be distinguished on the basis of small soma size (measurable via C_M) and an incremental pattern of depolarization followed by repetitive spike discharge several hundred milliseconds prior to inspiratory-related XII (or C4) motor activity (Fig. 1A) (Bianchi et al. 1995; Rekling et al. 1996a; Rekling & Feldman, 1998; Ballanyi et al. 1999; Richter & Spyer, 2001). In a prior study we showed that these intrinsic properties are a reliable means to identify rhythmogenic preBötC neurons (Hayes & Del Negro, 2007).

We isolated K⁺ currents in whole-cell voltage clamp using low Ca²⁺ ACSF containing 3 mm extracellular [K⁺], 1 μm TTX and 200 μm Cd²⁺. Depolarizing step commands from −100 mV (up to +10 mV) evoked sustained K⁺ currents in addition to transient K⁺ currents, i.e. I_A (Fig. 1B). I_A could also be evoked by depolarizing step commands from a −60 mV holding potential (Fig. 1B, inset), suggesting that I_A does not completely inactivate at baseline membrane potentials observed during normal inspiratory activity in vitro (e.g. Fig. 1A).

In current clamp, depolarizing current steps from a holding potential of −70 mV evoked a ramping depolarization (ΔV/Δt = 18 mV s^-1), whereas steps from −40 mV resulted in passive responses that quickly achieved steady state (Fig. 1C), which indicates that I_A is de-inactivated at hyperpolarized potentials, but steady-state inactivated at voltages above spike threshold.

Expiratory neurons in the preBötC are inhibited during XII motor activity but otherwise spike tonically (Fig. 2A). Their K⁺ currents were typically smaller overall, and only one expressed I_A (compare Figs 1B and 2B, note scale bars are the same, n = 8). In current clamp, depolarizing step commands did not generally evoke a ramping depolarization from any holding potential (Fig. 2C), which is consistent with the lack of I_A.

Biophysical properties of I_A

We separated I_A from non-inactivating K⁺ currents by subtraction. Using the same conditions as Figs 1B and 2B, we applied a sequence of 1 s step commands from −70 to +10 mV from a holding potential of −100 mV (Fig. 3A) and then repeated these steps from −40 mV (Fig. 3B). The difference current was defined as I_A, which activated at −60 mV and its maximum amplitude exceeded 1 nA at voltages greater than 0 mV (Fig. 3C).

Twenty-one of 34 (61.7%) inspiratory neurons expressed a peak transient outward current that exceeded the steady-state outward current and decayed exponentially with a time constant greater than 15 ms, which we defined as measurable I_A (Fig. 4A). This fraction of expression was significantly different from the 1/8 (12.5%) expiratory neurons found to express I_A (Fisher Exact test: P = 0.015).

Figure 4. Whole-cell characteristics of I_A in preBötC neurons.

A (top), the peak transient current normalized to the steady-state outward current plotted versus the measured whole-cell time constant of inactivation of outward currents (τ_h). The shaded area represents the boundary of measurements that were classified as not expressing I_A. Filled circles represent measured I_A-expressing inspiratory neurons, the downward pointed triangle represents the single I_A-expressing expiratory neuron, the crosses represent the inspiratory neurons that did not express I_A, and the upward pointing triangles represent non-I_A-expressing expiratory neurons. A (middle), the data points from top panel collapsed onto the τ_h axis. A (bottom), a histogram of the samples from middle panel in 2.5 ms bins. Points to the right of the vertical dashed line were called I_A-expressing neurons. B, the measured maximum conductance of I_A (g_A-max) versus C_M. The diagonal line represents a fit to the data with the y-intercept at the origin. Filled circles represent data points acquired using the subtraction protocol described in the text, while open circles represent data points acquired from a holding potential of −60 mV which may underestimate g_A-max. C, the g_A-maxversus R_N where the diagonal line is a linear fit to the data with no constraints. D, a pooled histogram showing the drive latencies of inspiratory neurons that expressed I_A(g_A⁺) or did not not (g_A^-) with up to 10 cycles per neuron. The maximum ordinate on each histogram is 16 bursts bin^-1. E, the drive latencies from each neuron with a given g_A. The line is fitted to the mean drive latencies of the pool of neurons. F, cycle-triggered averages of the membrane trajectory of inspiratory neurons with their associated XII recorded below. The vertical scale bar is 5 mV while the horizontal scale bar is 400 ms.

For inspiratory neurons expressing I_A, C_M measured 53.6 ± 5.9 pF (n = 21) and the difference between the onset of inspiratory-related EPSPs and the upstroke of XII activity (i.e. the drive latency) measured 305.2 ± 14.7 ms, which suggests these cells are putatively rhythmogenic (Rekling et al. 1996a; Rekling & Feldman, 1998; Feldman & Del Negro, 2006; Hayes & Del Negro, 2007). The C_M of neurons expressing I_A was directly related to g_A (Fig. 4B). Fitting the whole-cell conductance for I_A (g_A) linearly with C_M, with a y-intercept at zero, resulted in a slope of 0.219 ± 0.09 nS pF^-1. There was no obvious relationship between R_N of neurons expressing I_A and g_A (Fig. 4C).

The 13 of 34 (38.2%) preBötC neurons without measurable I_A exhibited drive latencies of 321.7 ± 14.6 ms and C_M of 44.4 ± 3.1 pF, which were indistinguishable from I_A-expressing neurons (latencies: P = 0.428; C_M: P = 0.181). Histograms of drive latencies for neurons with and without I_A are depicted in Fig. 4D showing the substantial overlap of the variability of burst activation. The g_A for neurons with I_A tends to increase as the drive latency decreases (Fig. 4E, latency =–0.008 g_A+ 13.6). Cycle-triggered averages (Fig. 4F) indicate that in neurons with I_A (n = 16) the voltage trajectory prior to the inspiratory phase shows a more abrupt rise that is statistically different from neurons without I_A (n = 8) (79.1 ± 11.6 ms versus 120.4 ± 14.4 ms, P = 0.037).

Detailed voltage-clamp analysis was precluded in whole-cell recordings because of inherent space-clamp limitations and series resistance errors attributable to large magnitude membrane currents (Armstrong et al. 1992). Therefore we studied I_A in somatic outside-out patches, repeating the subtraction protocol described above with step commands that reached +30 mV (Fig. 5A). The I_A activation function (see Methods) was fitted with the parameters θ_m=–16.3 mV and σ_m= 14.9 mV. Even in patches, I_A generally exceeded 200 pA with a maximum conductance of 1.14 ± 0.36 nS (n = 6).

We measured the steady-state inactivation of I_A at +10 mV for 500 ms following 1 s conditioning prepulses from −100 to +10 mV. The inactivation curve reaches its minimum above −40 mV and was fitted with the parameters θ_h = –85.6 mV and σ_h = –13.8 mV. These data explain why transient K⁺ currents can be evoked from a holding potential of −60 mV (e.g. Fig. 1B, inset); I_A is not fully inactivated at that potential, h_∞(–60) = 0.135 (Fig. 5B).

Over the range −30 to +30 mV, the inactivation time constant for I_A was 200–300 ms and could be empirically fitted with a line in the form, τ_h(V) = 202–0.42 V (Fig. 5B, bottom). Interestingly, τ_h(V) of ∼200 ms is commensurate with both the ramping depolarization responses observed in current clamp from baseline voltages of −70 mV (e.g. Fig.1C) and the transient ramp-like depolarization seen during endogenous network activity (e.g. Fig. 1A), suggesting the involvement of I_A in these membrane behaviours.

We computed the expected time course of I_A for the neuron in Fig. 1A to ascertain if and when the current would be active during the respiratory cycle (Fig. 6A). We used g_A of 11.7 nS, commensurate with the average maximum conductance, the chord conductance equation I_A=g_A*m_A(∞)*h_A*(V – E_K), the equation for inactivation gating (dh_A/dt), and empirically determined voltage dependence and kinetics (i.e. Fig. 5B). As the neuron begins to initiate the inspiratory burst, I_A rapidly achieves ∼60 pA, then diminishes throughout the burst and measures ∼40 pA at burst termination (Fig. 6B). This illustrates that I_A provides hyperpolarizing current during the burst with a particularly large influence at burst onset.

I_A exhibits window current between −60 and −40 mV that peaks at −52.2 mV with 0.6% of the maximum current active at steady state. This suggests that I_A does not substantially influence the baseline membrane potential during the majority of the quiescent phase of network activity but nonetheless resides in a sufficiently de-inactivated state that it can be rapidly evoked by depolarization; h_A(∞) is less than 0.2 through the expiratory phase (Fig. 6C). This is supported by the simulation where only 1–2 pA of I_A flows during the inter-inspiratory burst interval, yet I_A rapidly exceeds 50 pA at burst onset (Fig. 6B). Since voltage-clamp measurements may be subject to error, we examined whether disparities in the activation and inactivation curves could influence I_A. We shifted the half-activation and half-inactivation 10 mV in both the positive and negative direction and re-ran the simulations. When both curves were shifted, peak current was linearly related to the magnitude of the shift with a 60% increase for a +10 mV shift and a 42% decrease for a −10 mV shift, while the time course of I_A did not change. Regardless of shifts in voltage dependence, less than 2 pA of I_A flowed during the inter-inspiratory burst interval. These data suggest that I_A may play a role in inspiratory burst dynamics primarily as a result of the large conductance and its availability for activation, but I_A has little influence during the interburst membrane trajectory.

To test the role of I_A in the preBötC we first had to understand its pharmacology. We performed dose–response experiments with 4-aminopyridine (4-AP) in outside-out patches (Fig. 7B, n = 13). We measured the change in peak transient outward current due to 4-AP and the change in the total area of the transient outward component. The IC₅₀ for the peak response was 2.0 mm, which is close to the IC₅₀ of most A-currents (Rogawski, 1985). The IC₅₀ for the total area was 0.8 mm, while approximately 20% of the transient outward current could not be blocked by even saturating doses of 4-AP.

4-AP (2 mm) substantially attenuated I_A in outside-out patches (Fig. 7A); 4-AP likewise attenuated I_A in whole-cell recordings (n = 4), as previously shown (Inyushkin, 2005). Interestingly, non-inactivating outward currents evoked from the holding potential −40 mV were unaffected by 4-AP. We computed the 4-AP-sensitive current not attributable to I_A by subtracting the current evoked at +30 mV (from a holding potential of −40 mV) in control and 4-AP conditions, which measured < 25 pA. In contrast, the 4-AP-sensitive I_A regularly exceeded 200 pA (Fig. 7A), where I_A was defined as the subtracted current at +30 mV following prepulse holding potentials of −100 mV and −40 mV (Fig. 3). These data indicate that 4-AP attenuates I_A but does not affect sustained outward currents at baseline membrane potentials (e.g. channels related to the KCNQ subfamily). Consistent with blockade of I_A, delayed excitation in whole-cell current clamp, evoked by 500 ms current steps in the interval between XII discharge from a holding potential less than −70 mV, was also abolished by 4-AP (Fig. 7C). We also analysed the effects of 4-AP on pulse-evoked spikes (Fig. 7D). In 189 spikes of 4 neurons tested at rheobase, 2 mm 4-AP significantly increased the width at half-maximum of spikes 177 ± 26% from 2.7 ± 0.1 to 4.8 ± 0.6 ms (P = 0.031) and the area of the spike to 137 ± 13% from 304.7 ± 18.8 to 416.9 ± 23.2 mV·ms (P = 0.023). Some of this effect may have been attributable to attenuation of I_A (see simulated time-course and magnitude of I_A in Fig. 7D), but is more likely due to 4-AP effects on other channels in the Kv subfamily, which give rise to currents with delayed rectifier-like properties in addition to I_A.

Since I_A is expressed in more than half of the inspiratory preBötC neurons we recorded, in which it is available at typical baseline membrane potentials and generates large magnitude outward currents lasting several hundred milliseconds (Figs 1 and 3–7), we posited that I_A could influence rhythmogenesis.

4-AP affects rhythmic activity in the preBötC

We used 4-AP sensitivity to test the role of I_A in rhythm generation. In the context of respiratory network activity 4-AP caused erratic XII output that could not be straightforwardly interpreted. Therefore, we sought to determine whether the disorganized XII activity reflected a breakdown in rhythmogenesis by performing field recordings within the preBötC while recording the contralateral XII activity (Fig. 8B). The preBötC and XII activity patterns were coherent and recognizably rhythmic in control and washout, whereas 4-AP caused noisy preBötC rhythms that fluctuated in amplitude and period, and did not always match the output of the XII channel (Fig. 8A, n = 4). Under these conditions, the average rise time of the inspiratory activity within the preBötC decreased significantly in the presence of 4-AP from 141.3 ± 5.14 to 86.3 ± 8.8 ms (P = 0.016), while the falling slope did not change (P = 0.760, Fig. 8C).

Figure 8. Effects of 4-AP on the preBötC network.

A, ∫preBötC (top traces) and ∫XII (bottom traces) under control conditions, in the presence of 2 mm 4-AP and washout. Scale bar is 5 s. B, cartoon showing the configuration of the preBötC field-recording pipette (∫preBötC) and ∫XII suction electrode in A and C. The subcompact division of the nucleus ambiguus is indicated by scNA, and the hypoglossal motonucleus is indicated by XII. C, cycle-triggered averages of the field preBötC recordings under control conditions and in 2 mm 4-AP (n = 4).

As 4-AP perturbed rhythmic activity in preBötC field recordings we sought to determine the cellular basis for population-level fluctuations via whole-cell recordings and analyses (Figs 9 and 10).

Figure 9. I_A affects the discharge pattern of preBötC neurons.

A, on-cell activity of a preBötC neuron with early inspiratory spiking relative to the ∫XII. Current-clamp recording at 0 pA holding current illustrating similar activity as observed on-cell, i.e. early inspiratory spiking relative to XII output. Current-clamp recording at −40 pA holding current reveals temporal summation of EPSPs, beginning 400 ms prior to XII output. Voltage-clamp recording at a holding potential of −60 mV shows temporal summation of EPSCs prior to XII output. Baseline current was −25 pA. B, the burst discharge pattern changes from predominantly incremental (left) to decremental (middle) and back to incremental in washout (right). All recordings were at 0 pA bias current. Dashed lines at the bottom indicate the change in drive latency for each inspiratory burst illustrated. Traces in A, B and 7C were recorded in the same neuron. C, summary of the effects of 4-AP on the leading and trailing slope of inspiratory bursts (n = 8). Significance at P < 0.05 is shown with an asterisk.

Figure 10. Effects of 4-AP on the period and amplitude of inspiratory bursts.

A, current-clamp recordings of an inspiratory neuron in control and 2 mm 4-AP. All recordings were at 0 pA bias current. Smooth V_m denotes traces conditioned by a 5 Hz low-pass filter subsequently used to select inspiratory burst events (indicated by down arrows) and to measure their period and amplitude for the analyses in B and C. B, the mean change in burst period and coefficient of variation (n = 8). C, the mean change in burst amplitude and coefficient of variation (n = 8). Significance at P < 0.05 and P < 0.01 are shown with single and double asterisks, respectively.

The depolarizing activity occurring 400 ms (or more) prior to the inspiratory burst is largely synaptically driven, as we showed previously (Hayes & Del Negro, 2007). This early inspiratory activity can be observed in the on-cell recording configuration prior to whole-cell (Fig. 9A). In current clamp at baseline membrane potential (approximately −60 mV) the ascending ramp-like trajectory is accompanied by vigorous spiking. Hyperpolarizing the membrane potential reveals the temporal summation of EPSPs in the early inspiratory phase, which are also well resolved as summating EPSCs in voltage clamp (Fig. 9A, left to right). These data emphasize that a rapidly activating K⁺ current, which is de-inactivated at baseline voltages and subject to temporally summating synaptic inputs, may be uniquely poised to influence the dynamics of the inspiratory burst pattern.

We tested this idea by applying 2 mm 4-AP and measuring its effects on inspiratory bursts (Fig. 9B and C). The pattern changed from incrementing in control to decrementing in the presence of 4-AP (Fig. 9B). The leading slope of the inspiratory burst changed significantly in 4-AP from 53.1 ± 7.4 to 80.2 ± 5.3 mV · s^-1 (P = 0.032, n = 8, Fig. 9C). However, the trailing slope did not change significantly: −52.7 ± 2.5 mV s^-1 in control versus−65.5 ± 6.0 mV s^-1 in 4-AP (P = 0.095, n = 8, Fig. 9C). These general features are consistent with the average field-recording data analysed in Fig. 8C.

These data suggest that 4-AP removed a hyperpolarizing current that normally influenced the onset of the inspiratory burst, but its influence diminished during the inspiratory burst, which is consistent with the role of I_A we predicted (Fig. 6). Moreover, 4-AP did not change the baseline membrane (bias current is 0 pA in Figs 9B and 10A), which is consistent with the lack of significant window current measured in voltage clamp (see Fig. 5B) as well as the predictions of our simulation (Fig. 6B).

We also analysed the effect of 4-AP on drive potential amplitude and its coefficient of variation (CV) as well as respiratory cycle period and its CV. A representative experiment illustrates that burst sizes became more variable and that the timing of the neuronal activity became more variable (Fig. 10A). We applied a 5 Hz low-pass filter to the voltage trajectory, which filters out spikes and facilitates measurements of the underlying inspiratory drive potentials (Pace et al. 2007b). The amplitude and period of inspiratory activity fluctuated in 4-AP. The average period of drive potentials did not change significantly between control and 4-AP application (P = 0.052) while the CV for period approximately doubled from 0.32 ± 0.05 to 0.52 ± 0.03, which was statistically significant (P = 0.006, n = 8, Fig. 10B). The average amplitude of the drive potential did not significantly change (P = 0.408) but the CV for amplitude changed significantly from 0.26 ± 0.06 to 0.58 0.08 (P = 0.015, n = 8, Fig. 10C), which suggests that I_A influences the regularity of both the cycle period and the drive potential magnitude in preBötC neurons.

Discussion

We studied biophysical properties of I_A in preBötC neurons and analysed its role in respiratory rhythm generation. I_A recovers from inactivation at membrane potentials traversed during the interburst interval and remains relatively de-inactivated at baseline voltages, and thus available to be readily activated at the onset of the inspiratory phase of each respiratory cycle. I_A is prevalent among inspiratory neurons while being sparse in expiratory neurons. We found that inspiratory neurons expressing I_A received synaptic input at approximately the same time on average as inspiratory neurons that lack significant I_A (Fig. 4D), but that the rate at which these neurons responded to network activity was slower (Fig. 4E). Blockade of I_A increased the rate at which peak activity was achieved at both the network (Fig. 8C) and neuronal level (Fig. 9) which was correlated with an increase in the variability of both burst amplitude and period. Our data suggest that I_A may normally slow the onset of network inspiratory activity by counteracting excitatory synaptic input until there is massive excitatory drive to overcome I_A. We suggest this thereby promotes regular burst size and frequency throughout the network by suppressing spurious inputs.

The rhythm and pattern for breathing are generated in the brainstem. Neurons in the dorsal respiratory group (DRG), including the nucleus tractus solitarius (NTS) where I_A has been characterized (Champagnat et al. 1986; Dekin & Getting, 1987; Dekin et al. 1987), participate in afferent feedback and autonomic regulation. The ventral respiratory group (VRG) contains a bilaterally distributed column of neurons including rhythmogenic interneurons concentrated in the preBötC as well as premotoneurons that project to cranial and spinal motoneurons to carry out breathing movements (Bianchi et al. 1995; Onimaru et al. 1997; Ballanyi et al. 1999).

The role of I_A has been studied in mathematical models of respiratory networks that include the VRG and DRG, which predicted that I_A could influence the initial ramping trajectory of inspiratory neurons if it were sufficiently de-inactivated during the interburst phase of the respiratory cycle (Rybak et al. 1997). However, one caveat is that this network model did not explicitly consider the dynamics of preBötC neurons as a centre of rhythm generation and subsequent models of the preBötC have not analysed the role of I_A (Butera et al. 1999a,b; Del Negro et al. 2001; Rybak et al. 2003, 2004; Kosmidis et al. 2004). Our study is the first detailed characterization of I_A from putatively rhythmogenic preBötC neurons, as well as the first analysis of the role of I_A during endogenous respiratory network activity in vitro.

Role of I_A in the preBötC in vitro

We measured the voltage dependence and kinetics of I_A in somatic outside-out patches to minimize space-clamp limitations and series-resistance errors. I_A activates near −60 mV and is not fully inactivated until approximately −30 mV. Because these activation and inactivation functions encompass the range of membrane potentials visited during the interval between inspiratory bursts, I_A resides in a partially de-inactivated state at typical baseline membrane potentials in vitro. However, I_A has little window current, and consequently 4-AP application did not depolarize preBötC neurons. These properties indicate that I_A contributes little to the baseline membrane potential but can be rapidly recruited by synaptic depolarization at the onset of the inspiratory phase of the respiratory cycle (see Fig. 6).

I_A often causes delayed excitation, wherein the depolarization evoked by current pulses evolves with a ramp-like trajectory lasting several hundred milliseconds (or longer) because I_A activates rapidly and inactivates slowly (Hagiwara et al. 1961; Getting, 1983, 2989; Dekin & Getting, 1987; Gabel & Nisenbaum, 1998). Delayed excitation affects synaptic integration (Storm, 1988; Hoffman et al. 1997; Gulledge et al. 2005). This is particularly relevant in preBötC neurons with small C_M and early drive latency that are probably rhythmogenic, in which synaptic excitation builds up over several hundred milliseconds prior to respiratory-related motor output (Rekling et al. 1996a; Rekling & Feldman, 1998; also see Fig. 1A and 9; Hayes & Del Negro, 2007). Since I_A produces delayed excitation in preBötC neurons, is de-inactivated at baseline membrane potentials (Fig. 1B inset and 5B), and has a 200 ms inactivation time constant, we conclude that I_A plays a major role in shaping the ramp-like incremental discharge pattern characteristic of rhythmogenic neurons. Supporting evidence for this role is the dramatic increase in the leading slope of inspiratory activity following 4-AP application (Figs 8C and 9) as well as our comparison of membrane trajectories between neurons that express I_A and those that do not (Fig. 4F).

In invertebrate CPGs, I_A regulates the order in which rhythmogenic neurons discharge (Getting, 1983; Tierney & Harris-Warrick, 1992). In the stomatogastric ganglion of lobster Panulirus interruptus, I_A slows the frequency of anterior burster and pyloric dilator cells and delays the response to inhibition of follower lateral, early, and late pyloric cells (Tierney & Harris-Warrick, 1992). Likewise in the opistobranch mollusk Tritonia diomedea, I_A causes the appropriate tail escape–swim pattern to be generated by delaying the activation of ventral swimming interneurons that receive synaptic input from cerebral cell 2 (Getting, 1983). Additionally, in the inking behavioural circuits of Aplysia californica, I_A impedes ink release in the absence of substantial synaptic input to L14 neurons (Byrne et al. 1979; Byrne, 1980).

In the mammalian preBötC, network rhythms continued in the presence of 4-AP and could be measured in whole-cell and field recordings. 4-AP did not affect the mean period or amplitude of inspiratory burst-like discharges, but it did significantly increase period and amplitude variability (Fig. 10B and C). The increase in variability was correlated with the diminished ramp-like incremental discharge pattern in 4-AP. One possible explanation for the destabilization of periodic preBötC activity is that, like in Aplysia, I_A delays neuronal activation until the temporal summation of excitatory synaptic input builds up sufficiently to overcome and outlast the transient K⁺ current. We found that the more I_A is expressed in a neuron, the less the neuron is responsive to synaptic input preceding a burst (Fig. 4E and F) which is consistent with this hypothesis. However, the neurons lacking I_A are not expected to reach their peak activity until the neurons expressing some degree of I_A, which make up the majority of early inspiratory neurons, are also active. Thus I_A would influence the orderly recruitment of rhythm-generating neurons in the build-up to the inspiratory burst and ensure that the inspiratory phase does not begin until and unless a substantial fraction of the rhythmogenic population was involved. We speculate that this role for I_A would promote regularity in respiratory network behaviour, since the pharmacological attenuation of I_A caused substantial fluctuations in the amplitude and period of inspiratory burst activity.

An important caveat to this assertion is that 2 mm 4-AP extends spike duration significantly (Fig. 7D), which suggests that other outward currents may be affected by 4-AP such as Ca²⁺-activated K⁺ currents (Andreasen, 2002) or delayed-rectifier currents (Rusch & Eatock, 1996). However, the trailing slope of bursts did not change significantly at the network level (Fig. 8C) or neuronal level (Fig. 9C) suggesting the contribution of any outward currents active at the end of the burst were not significantly affected by 4-AP.

The above interpretation is further complicated by the fact that other neurons contained in slice preparations are also sensitive to 4-AP. Both raphé neurons (Aghajanian, 1985) and NTS neurons (Haddad & Getting, 1989) express I_A and project to preBötC neurons (Al-Zubaidy et al. 1996; Blessing, 1997; Pace et al. 2007b). These neurons may influence the preBötC by providing spurious excitatory or inhibitory input in 4-AP, but given that the slope of inspiratory activity changes in field recordings and in intracellular recordings, which is expected with a blockade of I_A, we are confident that at least some of this modulation is due to direct effects on rhythmogenic preBötC neurons. It is also unlikely that expiratory neurons in the preBötC that project to inspiratory cells can explain the effects of 4-AP on network activity because they did not express significant I_A.

Raphé and NTS neurons also project to hypoglossal motoneurons (Rekling et al. 2000). If these projection neurons are the same ones that have been shown to express I_A (Champagnat et al. 1986; Dekin & Getting, 1987; Dekin et al. 1987) then their response to 4-AP could further contribute to the very erratic XII discharge we observed. Evidence in support of this notion is that 4-AP is known to substantially increase spontaneous synaptic input to XII motoneurons and that I_A in XII motoneurons is relatively 4-AP-insensitive, fast-inactivating and mainly involved in spike repolarization (Haddad et al. 1990; Viana et al. 1993; Lape & Nistri, 1999).

Molecular identity of preBötC I_A

I_A in preBötC neurons shares similar voltage sensitivity and kinetic properties as members of the Shal-family (Kv4) K⁺ channels (Pak et al. 1991). In particular, mammalian Kv4.1 and Kv4.3 channels expressed in Xenopus oocytes (Serodio et al. 1994, 1996) closely match the I_A that we recorded in the preBötC. Kv4.1 subunits are sparsely expressed in mammalian brain tissue while Kv4.3 is widespread (Trimmer & Rhodes, 2004) and has voltage-independent kinetics (Serodio et al. 1996). This suggests that Kv4.3 expression may give rise to at least some of the I_A in preBötC neurons. Kv4.2 is also a good candidate because it is present in the nearby raphé neurons (Serodio & Rudy, 1998). Any of these, or other, Kv subunits could potentially make up preBötC I_A because the voltage dependence and kinetics could be influenced by Kv channel interacting proteins (An et al. 2000; Rhodes et al. 2004) or neuromodulation (Birnbaum et al. 2004). However, definitive molecular level identification will require single-cell reverse transcriptase polymerase chain reaction experiments or neuroanatomical approaches.

In addition to I_A, some of the K⁺ current affected by 4-AP probably includes voltage-gated K⁺ channels with delayed rectifier-like properties that derive from the Kv subfamily, such as Kv2.x and Kv3.x subtypes, because 4-AP extended action potential duration. However, the 4-AP-sensitive current is unlikely to include non-inactivating K⁺ currents from the KCNQ subfamily, since we found no 4-AP-sensitive sustained currents from the relatively depolarized holding potential of −40 mV.

The prevalence of I_A in rhythmogenic preBötC neurons

To identify inspiratory neurons in the preBötC that may be important for rhythm generation Rekling et al. (1996a) measured the drive latency and proposed that the earliest neurons to activate during the respiratory cycle are important for rhythmogenesis. Furthermore, we recently showed that membrane capacitance (C_M) of 30–65 pF is correlated with early drive latency in preBötC neurons that most likely play the foremost role in rhythmogenesis (Hayes & Del Negro, 2007). Since we found that the majority, but not all, putatively rhythmogenic neurons express somatic I_A based on these criteria, our data suggest that rhythmogenic neurons do not uniformly express I_A as originally suggested (Rekling et al. 1996a). However, we cannot exclude the possibility that many of the neurons we recorded may have expressed unrecognized dendritically localized I_A as our measurements strictly depend on the quality of our somatic voltage clamp.

Since an A-current as we have described is generally lacking in expiratory neurons, but prevalent in inspiratory neurons, we hypothesize that the molecular identity of channels that give rise to I_A may be a unique functional marker for putatively rhythmogenic neurons within the preBötC.

Role of I_Ain vivo

In the context of endogenous network activity in vitro E_K is approximately −71 mV, while in vivo it is probably closer to −98 mV – assuming 3.2 mm[K⁺] in the cerebrospinal fluid (Melton et al. 1991; cf. Richter et al. 1978). We therefore expect higher driving force for I_Ain vivo, and thus the window current would have a larger hyperpolarizing influence on baseline membrane potential. This could further de-inactivate I_A between inspiratory bursts, compared to what we observe in vitro. In this environment, where inspiratory neurons are expected to be under intensive bombardment of excitatory and inhibitory input, I_A may play an even more substantial role in properly synchronizing respiratory rhythmic activity through its ability to rapidly activate with a large outward current and quench spurious depolarizations.