KCNQ Current Contributes to Inspiratory Burst Termination in the Pre-Bötzinger Complex of Neonatal Rats in vitro

The pre-Bötzinger complex (preBötC) of the ventral medulla generates the mammalian inspiratory breathing rhythm. When isolated in explants and deprived of synaptic inhibition, the preBötC continues to generate inspiratory-related rhythm. Mechanisms underlying burst generation have been investigated for decades, but cellular and synaptic mechanisms responsible for burst termination have received less attention. KCNQ-mediated K+ currents contribute to burst termination in other systems, and their transcripts are expressed in preBötC neurons. Therefore, we tested the hypothesis that KCNQ channels also contribute to burst termination in the preBötC. We recorded KCNQ-like currents in preBötC inspiratory neurons in neonatal rat slices that retain respiratory rhythmicity. Blocking KCNQ channels with XE991 or linopirdine (applied via superfusion or locally) increased inspiratory burst duration by 2- to 3-fold. By contrast, activation of KCNQ with retigabine decreased inspiratory burst duration by ~35%. These data from reduced preparations suggest that the KCNQ current in preBötC neurons contributes to inspiratory burst termination.


INTRODUCTION
Inspiration is the dominant, active phase of the respiratory cycle in mammals. Its rhythm emanates from the pre-Bötzinger complex (preBötC) of the ventrolateral medulla (Smith et al., 1991;Del Negro et al., 2018). The most recent models of inspiratory rhythm generation favor synaptic mechanisms where burstlets or bursts derive from the progressive recruitment and excitation of a critical subpopulation of Dbx1-derived, glutamatergic neurons and their increased synchronization in the pre-inspiratory and ultimately inspiratory phases (Del Negro et al., 2018;Ashhad and Feldman, 2020;Kallurkar et al., 2020).
What about inspiratory burst termination? Under normal physiological conditions in many vertebrates, feedback from slowly adapting pulmonary stretch receptors plays an important role in terminating inspiration (Bonham et al., 1993). Most importantly, when stretch receptor feedback is removed by vagotomy for experimental preparations or due to medical necessity in humans with heart-lung transplants, inspiration is prolonged, but it still terminates (Lumsden, 1923;Smith et al., 1990;Bonham et al., 1993). Similarly, dorsolateral pontine networks normally contribute to inspiratory burst termination, but even the prolonged inspiratory (apneustic) events that emerge when these inputs are removed self-terminate (Lumsden, 1923;Stella, 1938;von Euler et al., 1976;Foutz et al., 1988;Feldman et al., 1992;Harris and Milsom, 2003).
Synaptic inhibition between central respiratory neurons is also a significant contributor to burst termination because a block of this inhibition prolongs the inspiratory burst . Nevertheless, even when synaptic inhibition is blocked in the preBötC in vivo  or throughout the brainstem in situ (Hayashi and Lipski, 1992) or in vitro (Feldman and Smith, 1989;Feldman and Del Negro, 2006;Sherman et al., 2015;Cregg et al., 2017;Baertsch et al., 2018), inspiratory bursts still terminate. Synaptic inhibition influences inspiratory bursts but is not required for burst termination.
Stripped of reflex pathways and contributions attributable to synaptic inhibition, inspiratory burst termination is probably accomplished by cellular and synaptic mechanisms intrinsic to core rhythmogenic preBötC interneurons. Putative mechanisms include desynchronization amongst rhythm-generating neurons as the inspiratory burst progresses (Ashhad and Feldman, 2020), excitatory synaptic depression (Guerrier et al., 2015;Kottick and Del Negro, 2015), and outward currents that activate due to inspiratory burst activity and thus hasten burst termination and promote transient post-burst refractoriness.
Ca 2+ -activated K + current (I K(Ca) ) contributes to burst termination in many vertebrate central pattern generator (CPG) networks (Grillner, 2006). However, I K(Ca) does not appear to contribute to burst termination in the preBötC because its block has only minor and variable effects on network rhythm and burst pattern (Onimaru et al., 2003;Crowder et al., 2007; reviewed in Rajani et al., 2016).
There is evidence for three cellular mechanisms that contribute to burst termination, including Na + -dependent K + current (I K-Na ), Na + /K + ATPase pump current, and ATP-dependent K + current (I K-ATP ; Del Krey et al., 2010). Whole-cell current-clamp recordings show that these currents contribute to the post-inspiratory burst hyperpolarization at the level of individual neurons, but effects at the network level are more difficult to interpret. The predicted effect of blocking mechanisms involved in burst termination is an increase in burst duration, but antagonists of these ion channels decrease burst duration (Krey et al., 2010), possibly due to additional effects on resting membrane potential.
Pharmacological potentiation of KCNQ currents slows inspiratory-related frequency in vitro and breathing frequency in vivo (Wei and Ramirez, 2019). In contrast, inhibition of KCNQ channels does not affect baseline inspiratory-related frequency in mice in vitro or breathing frequency in vivo. Inhibition of preBötC KCNQ channels does, however, partially counteract the well-known opioid-mediated depression in breathing through presynaptic mechanisms, even though these channels are not directly modulated by opioid receptor signaling (Wei and Ramirez, 2019). Human patients with mutations in KCNQ2/3 genes often develop neurological disorders associated with apnea (Ronen et al., 1993;Kato et al., 2013;Grinton et al., 2015). Finally, homozygous KCNQ2 knockout mice typically die in the first day of life due to respiratory insufficiency that may have a central etiology (Watanabe et al., 2001).
Therefore, we isolated the preBötC core oscillator in rhythmically active slices with the goal of evaluating whether KCNQ-mediated outward current may play a role in inspiratory burst termination.

Animals and Preparations
Male and female Sprague-Dawley neonatal rat pups ranging in age from postnatal day (P) 1 to 4 were obtained from timed-pregnant dams from Charles River or from Laboratories and Science Animal Support Services (University of Alberta). Animals were housed in University of Alberta's Health Sciences Laboratory Animal Services and maintained on a 12-h light/ dark schedule with ad libitum access to food and water.
Rhythmic medullary slice preparations were prepared as described previously (Smith et al., 1991;Ruangkittisakul et al., 2006;Lorier et al., 2007). Briefly, rat pups were anesthetized via inhalation of isoflurane. Once surgical plane anesthesia was verified by lack of withdrawal reflexes, animals were immediately decerebrated. The neuraxis was isolated in a dissection chamber containing cold (5-10°C) artificial cerebrospinal fluid (aCSF) composed of the following (in mM): 120 NaCl, 3 KCl, 1.25 NaH 2 PO 4 , 1 CaCl 2 , 2 MgSO 4 , 26 NaHCO 3 , and 20 d-glucose, equilibrated with 95% O 2 -5% CO 2 . The neuraxis was pinned to a wax chuck or glued to an agar block that was placed in the vice of a vibratome (Leica VT-1000S, or Leica VT-1200S, Concord, ON, Canada). Serial sections (100-300 μm) were then cut in the rostral to caudal direction and trans-illuminated to visualize anatomical landmarks identifying the rostral border of the preBötC. When the compact division of the nucleus ambiguus was no longer evident and the rostral margin of the inferior olive appeared in the thin section, we then cut a single 700-μm slice. The rostral border of the slice was ~0.35 mm caudal to the caudal aspect of the facial nucleus, which placed the caudal edge of the slice just caudal to the obex (Ruangkittisakul et al., 2006). The rhythmic slice contained the preBötC, rostral ventral respiratory group (rVRG), most of the XII motor nuclei, and the rostral XII nerve rootlets. For experiments that examined the effects of drugs on inspiratory network activity recorded from the XII Frontiers in Physiology | www.frontiersin.org nerves, slices were pinned rostral surface up in a Sylgardcoated recording chamber (volume, 10 ml) and perfused with aCSF that was recirculated at a flow rate of 9.5 ml/min. For whole-cell recording experiments, rhythmic slices were placed rostral surface up in the recording chamber (volume, 1 ml) of an upright microscope (Zeiss Axioskop 2 FS Plus, Toronto, ON, Canada) that was equipped with infrared-enhanced differential interference contrast optics (IRDIC). Slices were held in place by a U-shaped platinum frame with a mesh of parallel nylon fibers, i.e., a "harp, " and perfused with aCSF at a flow rate of 1-2 ml/min.

Whole-Cell Recording
Whole-cell recordings were made from inspiratory-modulated preBötC neurons (Rekling et al., 1996). A horizontal puller (P-97, Sutter Instruments, Novato, CA, United States) was used to pull whole-cell recording pipettes (3-4.5 MΩ) from filamented borosilicate glass (Clark/WPI, 1.2 mm O.D.). Patch pipettes were filled with intracellular solution containing (in mM) 140 K + -gluconate, 5 NaCl, 0.1 EGTA, 10 HEPES, 1 MgCl 2 , 1 glucose, and 1 Mg 2 ATP. Values of membrane potential (V m ) have not been corrected for the liquid junction potential of −14.1 mV (calculated using Clampex, pCLAMP Suite) that is associated with these solutions. Osmolarity was adjusted to 290-300 mOsm with sucrose, and pH was adjusted to 7.2-7.3 using KOH. Whole-cell signals (membrane current or voltage) were amplified and filtered with a patch-clamp amplifier (2-5 kHz low-pass filter, MultiClamp 700A, Axon Instruments) and sampled at 10 kHz via an analog-to-digital converter (Digidata 1322 A/D, Molecular Devices). Series resistance (R s ) was estimated and tracked throughout the experiment under voltageclamp conditions. R s measured 14.0 ± 1.4 MΩ (range: 12-16 MΩ). We excluded recordings in which R s varied by more than 20%.
The final concentration of DMSO in the aCSF or in the injection solutions never exceeded 0.003% for bath application of retigabine or XE991 and 0.02% for local application of XE991.
Drugs were either bath-applied or locally microinjected into the preBötC via pressure ejection from triple-barreled glass micropipettes (4-6 μm O.D. per barrel) pulled from borosilicate glass capillaries (cat no. 3B120F-4, WPI, Sarasota, FL, United States) using a vertical puller (P-35, Sutter Instruments). Drug concentration in the ejection pipette solution must be ∼10-fold greater than the bath-applied concentration to produce similar effects . For studies in which drugs were injected directly into the preBötC, the preBötC was first functionally identified by mapping the response to local injections of SP. In the site at which SP (1 μM, 10 s) evoked a rapid (next cycle), transient increase in inspiratory frequency that was at least double baseline was considered the preBötC core (see Zwicker et al., 2011).
For whole-cell recording experiments, drug ejection pipettes were placed 25-50 μm upstream of the target neuron at the superficial slice surface. With this arrangement, recorded neurons were typically 30-70 μm below the slice surface and 50-120 μm from the injection pipette tip. This placement, along with maintaining ejection pressure below 5 psi, was important to limit mechanical artifacts. In addition, membrane current was monitored carefully at pressure onset for abrupt, non-recoverable steps in holding current synchronous with pressure onset indicative of mechanical artifacts. Data were rejected if mechanical artifacts were noticed or R s changed Frontiers in Physiology | www.frontiersin.org by more than 20%. We have previously established that vehicle solutions containing DMSO at concentrations 0.2% (10-fold higher than used here) are without noticeable effect on membrane properties or inspiratory synaptic drive currents/ potentials/firing behavior (Funk et al., 1995;Ireland et al., 2012).
Note that after each experiment in which XE991 was bathapplied via the perfusion system, the inflow and outflow tubing (Tygon) was rinsed with at least 500 ml of ethanol followed by 2 L of double-distilled deionized water; and glass reservoirs and the recording chamber were thoroughly washed with ethanol and rinsed with double-distilled deionized water to remove residual XE991. Without this procedure, the baseline burst duration produced by the rhythmic slices on subsequent days would gradually increase during the first hour of slice recording (i.e., prior to the start of the experiment) to levels greater than twice the normal baseline values (i.e., >2,000 ms compared with a normal burst duration <1,000 ms). Thus, every experiment began with a 1-h recording of baseline activity to ensure that there were no residual effects of XE991. If a residual effect was observed, then we discarded those data, cleaned the recording chamber, and replaced the tubing.

Data Analysis and Statistical Comparisons
All measurements relating to frequency (inspiratory burst duration and inspiratory frequency) are reported as absolute values and relative to control, with mean ± standard deviation. Inspiratory XII burst amplitude is reported relative to control. The units of burst amplitude recorded from a suction electrode are not reported because they have no physiological meaning except as a relative comparison before/after drug application, such as in this context. Statistical tests were performed on raw data except for burst amplitude where relative values were compared. Student's paired t test was used to compare parameters between two groups. One-way analysis of variance (ANOVA) with repeated measures (RM) (used when no data were missing from any measurement category, i.e., when XE991 was bathapplied), or a mixed-effects model (used when data were missing from a measurement category, i.e., when linopirdine or retigabine were bath-applied) followed by Tukey's multiple comparison post hoc test (Prism 4.2 or 9.0.1, GraphPad Software, San Diego, CA, USA) was used to compare parameters between three or more groups. One-way ANOVA followed by Tukey's multiple comparison post hoc test was used to compare parameters between [K + ] e elevation, XE991, and retigabine. Differences between means were considered significant if p < 0.05.

RESULTS
We obtained voltage-clamp recordings from preBötC inspiratory neurons (input resistance measured 213 ± 144 MΩ, n = 5), as indicated by the presence of inward synaptic currents coincident with bursts of inspiratory motor output on the XII nerve ( Figure 1A). We bath-applied TTX (0.5 μM) to block sodium currents and ZD7288 (10 μM) to block the hyperpolarizationactivated mixed cation current (I h ). The activation of sodium currents or deactivation of I h (Thoby-Brisson et al., 2000) would otherwise obscure KCNQ current, which we aimed to evoke with our voltage-clamp protocol. We stepped the membrane from −60 to −20 mV for 90 s to evoke KCNQ current, which does not inactivate ( Figure 1B). Peak current measured ~600 pA, which then decreased at a constant rate of ~1 pA/s over 90 s. We assume the outward currents that declined at ~1 pA/s reflected non-KCNQ outward currents. XE991 blocks KCNQ in the open configuration (Greene et al., 2017), so we could only employ this drug after having opened the channel. The subsequent step evoked outward current matching the magnitude of the outward current at the termination of the first step. We monitored that current for 14-20 s, which again decreased at the rate of ~1 pA/s. We then applied XE991 (100 μM) locally (70-76 s, i.e., 90 s -the duration of monitoring). XE991 evoked an inward current (by blocking open KCNQ channels) that stabilized to the initial rate of decay. The current that developed between the onset of the XE991 application and the end of the −20-mV pulse was subtracted from the current that developed over the same time period in the control pulse to calculate the XE991-sensitive current.
XE991 reduced the outward current in five of five inspiratory neurons by −45 ± 26 pA ( Figure 1C). With the use of Ohm's law, this change in current was calculated to result in a −6.8 ± 1.5-mV hyperpolarization of membrane potential.
If KCNQ contributes to inspiratory burst termination, then its activation should hyperpolarize preBötC inspiratory neurons from potentials achieved at the end of an inspiratory cycle. We tested this idea in current-clamp recordings from seven preBötC inspiratory neurons synaptically isolated by 0.5 μM of TTX in the presence of 10 μM of ZD7288 (Figure 2A). Retigabine modulates KCNQ allosterically by hyperpolarizing its voltage dependence of activation, which accelerates, prolongs, and increases KCNQ channel openings in response to depolarization (Gunthorpe et al., 2012). The preBötC neurons were depolarized to −40 mV to approximate the membrane potential attained during the inspiratory drive potential. Retigabine (20 μM) was applied locally for 90 s. Figure 2B shows that retigabine elicited a hyperpolarization of −11.1 mV that recovered back to the initial holding potential several minutes after terminating the injection. Locally applied retigabine hyperpolarized preBötC inspiratory neurons by −4.7 ± 3.5 mV ( Figure 2C, n = 7, p = 0.01).
Note that the response to retigabine does not reveal the maximum KCNQ-mediated hyperpolarization but rather hyperpolarization achieved by the allosteric potentiation of the KCNQ current. Nevertheless, a response to retigabine does confirm the presence of KCNQ channels in preBötC inspiratory neurons and their influence on membrane potential in the range of −60 to −40 mV.
We next examined the impact of modulating KCNQ channels on inspiratory network activity. If KCNQ contributes to inspiratory burst termination, then KCNQ inhibition should increase burst duration, whereas its potentiation should reduce burst duration.
First, we tested whether inhibiting KCNQ increased burst duration. Rhythmic slices were exposed to XE991 (0.1 and 1.0 μM) via bath superfusion. Preliminary data indicated that XE991 required >30 min to achieve steady-state effects on network rhythm, so we assessed inspiratory activity after 60 min Frontiers in Physiology | www.frontiersin.org of incubation at each concentration. Figure 3A illustrates that XII motor output burst duration increased and that frequency decreased at both concentrations, although the effects on burst duration were stronger. The inset shows the burst profile of XII inspiratory-related output averaged from 10 consecutive cycles in control, 0.1, and 1 μM of XE991. The profile and duration of the XII motor output are reliable indicators of collective activity in the preBötC (Del Negro et al., 2001), although the absence of XII motor output does not indicate lack of collective activity in the preBötC (Kam et al., 2013;Ashhad and Feldman, 2020;Kallurkar et al., 2020).
After 60 min of washout, rhythmic inspiratory activity returned in two of four preparations in which inspiratory rhythm had stopped, while burst frequency increased in the A B remaining four preparations. However, none of the inspiratory burst parameters recovered fully: inspiratory burst duration was 404 ± 29 ms (72 ± 6% of control, Figure 5B, n = 6, p = 0.01), burst frequency was 5.2 ± 1.3 bursts/min (44 ± 13% of control, Figure 5B, n = 6, p = 0.02), and burst amplitude was 129 ± 9% of control ( Figure 5B, n = 6, p = 0.03). Bath-applied drugs may impact microcircuits throughout the slice. Therefore, changes in XII burst duration could reflect direct actions on preBötC rhythmogenic interneurons (Del Negro et al., 2018), output-related preBötC interneurons (Kam et al., 2013;Cui et al., 2016;Ashhad and Feldman, 2020;Kallurkar et al., 2020), neuronal populations upstream from the preBötC network that provide modulatory inputs (Ptak et al., 2009), or network elements downstream of the preBötC, specifically XII premotoneurons (Koizumi et al., 2008;Revill et al., 2015) or XII motoneurons. Therefore, to ascertain whether KCNQ modulation and its effects on XII motor output were mediated (at least in part) by preBötC neurons, we locally injected XE991 in the preBötC unilaterally and assessed its effects on inspiratory rhythm and pattern.
KCNQ activates below −60 mV and reaches full activation near zero mV (Marrion, 1997;Brown and Passmore, 2009;Gunthorpe et al., 2012). Thus, in these experiments where slices are perfused with aCSF containing extracellular K + ([K + ] e ) at 9 mM and baseline membrane potential of preBötC neurons is in the range of −50 mV (see experiments associated with Figures 1, 2), KCNQ will be partially activated under baseline conditions and become further activated with depolarization during the inspiratory burst. Thus, in addition to affecting the dynamics of the network by altering burst profile during the inspiratory phase, KCNQ will contribute to resting membrane potential. It is therefore challenging to determine whether the effects of KCNQ modulation on inspiratory output described above are due to its effect on resting membrane potential or the time-and voltage-dependent properties of KCNQ that will determine its pattern of activation during an inspiratory burst. To try and separate these effects, we compared the effects on burst duration and frequency that were produced by changing membrane potential through alterations in [K + ] e from 6 to 9 mM with those produced by inhibiting or activating KCNQ with XE991 (and linopirdine) or retigabine, respectively. This change in [K + ] e was selected because it produced a change in the membrane potential of preBötC neurons (~6-mV depolarization) similar in absolute value to that produced by retigabine (~5-mV hyperpolarization) or XE991 (~7-mV depolarization).
Elevating [K + ] e from 6 to 9 mM depolarized preBötC neurons and XII motoneurons on average from −57 ± 10.8 to

A B
FIGURE 6 | Effects of blocking KCNQ channels with XE991 locally within pre-Bötzinger complex (preBötC) on inspiratory burst parameters in the rhythmic slice preparation. (A) Rectified, integrated recordings from XII nerve rootlets demonstrating inspiratory burst parameter changes in the presence of local application of XE991 as a series of 5-s pulses for 30 min. (B) Group data from XII nerve rootlet recordings when XE991 was applied locally within preBötC, n = 6 for control and 10 μM of XE991 after 30 min, n = 4 washout. *p < 0.05, paired t-test.
Frontiers in Physiology | www.frontiersin.org −51 ± 3.8 mV (preBötC, n = 5), and −53 ± 5.1 to −47 ± 2.8 (XII, n = 9), an average of ~6 mV. If the effects of KCNQ modulators on inspiratory bursts are solely due to their effects on membrane potential, then the depolarizing actions of XE991 should match those evoked by the change in [K + ] e , while the effects of retigabine should be the opposite. Figure 7 shows the effects of elevated [K + ] e on resting membrane potential, burst duration, burst frequency, and burst amplitude next to the effects of XE991 and retigabine on these same measurements. [K + ] e elevation and XE991 have similar effects on baseline membrane potential yet disparate effects on burst duration and frequency. Elevating [K + ] e increased burst duration significantly by ~40% (from 577 ± 187 to 810 ± 80 ms, Figure 7A, n = 4, p = 0.04), but XE991 caused a >300% increase in burst duration, significantly greater than K + -mediated depolarization ( Figure 7A, three groups, 22 degrees of freedom, p < 0.0001).

DISCUSSION
Synaptic depression (Rubin et al., 2009;Guerrier et al., 2015;Kottick and Del Negro, 2015) and activity-dependent, Na + -linked mechanisms including Na + /K + ATPase pumps, I K-Na , and I KATP Krey et al., 2010) contribute to burst termination. Counterintuitively, pharmacological inhibition of those currents reduces, rather than lengthens, inspiratory burst duration. So, clearly, there are additional contributors to burst termination. We focused on KCNQ current because preBötC neurons express transcripts for KCNQ channels, it shapes burst profile in other networks, its voltage dependence and kinetics are suitable for burst termination, and Ca 2+activated K + currents have been excluded as contributors to inspiratory burst termination in the preBötC (reviewed in Rajani et al., 2016).
Here, our measurements of XII nerve burst duration in vitro show that blocking KCNQ (throughout the slice or locally within the preBötC) prolongs bursts, whereas activating KCNQ (throughout the slice) shortens burst duration and stops rhythmicity. We propose that KCNQ current represents a heretofore missing contributor to inspiratory burst termination.

Mechanisms Through Which KCNQ Modulation Influences Inspiratory Activity
Slow (tens of milliseconds) KCNQ activation is suitable for recruitment during inspiratory bursts, lasting ~300 ms, contributing to spike frequency adaption, and causing progressive membrane hyperpolarization (Storm, 1989;Gu et al., 2005;Honigsperger et al., 2015;Lombardo and Harrington, 2016;Dai et al., 2018). Presynaptically, KCNQ activation inhibits transmitter release (Wei and Ramirez, 2019). KCNQ activation during inspiratory bursts creates a shunt and progressively reduces membrane responsiveness to synaptic inputs (Brown, 1983;Brown and Passmore, 2009). Therefore, KCNQ at both pre-and postsynaptic sites in preBötC neurons can contribute to burst termination. (1 μM, n = 5) or retigabine (2 μM, n = 8) on inspiratory burst amplitude and resting membrane potential. The change in resting membrane potential evoked by increasing extracellular K + concentration from 6 to 9 mM was measured in pre-Bötzinger complex (preBötC) inspiratory neurons (n = 5); the change in resting membrane potential associated with XE991 was estimated using Ohm's law and the XE991-evoked current measured under voltage clamp and the input resistance of the recorded neurons (n = 5, Figure 1); the retigabine-evoked change in resting membrane potential was measured at −40 mV ( Figure 2, n = 7). *p < 0.05, one-way ANOVA with Tukey's multiple comparisons post hoc test.
Frontiers in Physiology | www.frontiersin.org KCNQ modulation affected burst duration more than frequency in rhythmic slices. XE991 and linopirdine attenuate KCNQ and increase burst duration 2-to 3.5-fold. KCNQ potentiation via retigabine decreases burst duration ~30%. Regarding frequency, inhibition of KCNQ has minimal effects on frequency (5-20% decrease). Wei and Ramirez similarly found no effect of XE991 (≥50 μM) or Chromanol 293B (another KCNQ blocker) on inspiratory frequency in mouse slices (Wei and Ramirez, 2019). Potentiation of KCNQ with retigabine (or ICA 69673) also decreased frequency and blocked rhythm (Wei and Ramirez, 2019). In summary, inhibition and potentiation of KCNQ increase and decrease burst duration, respectively. In contrast, frequency modulation was unidirectional -inhibition of KCNQ did not affect frequency, but KCNQ potentiation decreased frequency.
What might account for these differential effects of KCNQ modulation on frequency and burst duration? We propose that frequency effects are primarily related to the influence of KCNQ on resting membrane potential, while the voltage-and timedependent activation properties of KCNQ account for its effects on burst duration, consistent with Wei and Ramirez (2019). By contrast, modifications in [K + ] e potently modulate frequency yet exert a smaller effect on burst duration. Thus, this study focuses on KCNQ modulation of burst duration.
The postsynaptic actions of KCNQ can cause spike frequency adaptation, directly hyperpolarize the membrane potential, and shunt the membrane to synaptic drive, all of which can contribute to burst termination.
KCNQ activation during the inspiratory burst (hundreds of ms in vitro) progressively increases outward current (Brown, 1983;Brown and Passmore, 2009) that would counteract the ongoing excitation associated with inspiratory synaptic drive and activity of persistent currents such as persistent Na + current (I NaP ) and calcium-activated non-specific mixed cation current (I CAN ; Pace et al., 2007;Del Negro et al., 2018;Picardo et al., 2019). Inspiratory burst termination requires a reversal of the balance of inward and outward currents toward outward. KCNQ has the biophysical properties to transiently shift this balance and thus extinguish ongoing burst-like activity, as documented in other systems. In superior cervical ganglion sympathetic neurons, the duration of action potential bursts is determined by the balance between I NaP and KCNQ currents (Zaika et al., 2006). CA1 hippocampal neurons normally respond to short depolarizing excitatory current pulses with single action potentials, but those single spikes transform into bursts of action potentials in the presence of KCNQ inhibitor linopirdine (Yue and Yaari, 2004;Peters et al., 2005). Sympathetic neurons discharge tonically or exhibit bursts of action potentials. Tonic neurons lack KCNQ, whereas bursting neurons express a large KCNQ current. Moreover, neurons that discharge in bursts switch to tonic firing after KCNQ is blocked (Wang and McKinnon, 1995;Wang et al., 1998). Those data indicate that KCNQ biophysics provide the activitydependent outward current needed to transiently suppress spiking activity, which is obligatory for oscillatory bursting in single neurons, or as in the preBötC, a rhythmically active network.
It seems incongruent that the XE991-sensitive current measured 30-60 pA; i.e., this seems to be a relatively small current compared to the 2-to 3-fold increase in inspiratory burst duration evoked by KCNQ inhibition. Voltage clamp likely underestimates the true magnitude of the KCNQ current. First, there are voltage-clamp errors associated with uncompensated series resistance on the order of 12-16 MΩ. The maximum outward current evoked by voltage steps from −60 to −20 mV measured ~600 pA in one neuron and between 275 and 450 pA in the remaining four. By Ohm's law, a 600-pA outward current across 12-16 MΩ of series resistance will result in a 7-to 12-mV error. Thus, a step from −60 to −20 mV would have clamped the membrane potential to between −30 and −23 mV, thus underestimating the magnitude of the XE991-sensitive (KCNQ) current.
An incomplete block of KCNQ by XE991 may also have contributed to an underestimation of KCNQ current magnitude. Prolonged (>30 min) incubation periods are required for XE991 to block KCNQ channels (Martire et al., 2004;Yue and Yaari, 2004;Gu et al., 2005;Peters et al., 2005;Yeung and Greenwood, 2005;Wladyka and Kunze, 2006;Leao et al., 2009), which is consistent with our experiments where >30 min was required for steady-state effects of XE991. Nevertheless, 30-60 min may have been insufficient to achieve a maximum block. In addition, KCNQ blockers were used at concentrations near their respective IC 50 values (see below), which minimizes off-target actions but simultaneously ensures that the block would be incomplete.
It is also possible that the effects of XE991 on burst duration, recorded here from the XII nerve, are not mediated entirely through preBötC neurons but via other modulatory neurons or axons within the slice that project to preBötC, or through actions downstream from the preBötC at XII premotor neurons or motoneurons. Certainly, XII motoneurons express KCNQ channels, the blockade of which increases the number of spikes generated by a given depolarization (Ghezzi et al., 2017). Thus, for the experiments in which drugs were applied to the bath (Figures 3-5), we cannot exclude that KCNQ currents outside the preBötC contributed to the effects of KCNQ modulators on burst duration. However, the observation that the effects of XE991 on burst duration, recorded from the XII nerve, were similar whether applied to the bath or injected unilaterally into the preBötC (Figure 6) indicates that the majority of XE991 effects were mediated via neurons in the preBötC.
Finally, presynaptic KCNQ channel activation and the subsequent inhibition of transmitter release could contribute to the effects of KCNQ on burst duration but would not be measured by our voltage-clamp protocol in synaptically isolated preBötC neurons.
Linopirdine was applied at 3 μM, and its IC 50 is 3-7 μM (Costa and Brown, 1997;Lamas et al., 1997;Schnee and Brown, 1998;Wang et al., 1998). Linopirdine can inhibit large conductance Ca 2+ -activated K + current (BK) in hippocampal neurons with an IC 50 of ~16 μM (Lamas et al., 1997). However, off-target effects of linopirdine on BK current are unlikely in our context because more potent BK blockers show modest or inconsistent effects on inspiratory rhythm and pattern (reviewed in Rajani et al., 2016). Linopirdine also partially blocks leak, transient outward, and delayed rectifier K + currents and the slow component of the AHP but at 100 μM, which exceeds our dosage by orders of magnitude (Schnee and Brown, 1998). Thus, it is unlikely that the actions of linopirdine were mediated by anything other than KCNQ inhibition.

Physiological Significance
This study adds the KCNQ current to the list of candidate mechanisms that contribute to inspiratory burst termination in rodents. Of the four neuronal KCNQ isoforms (KCNQ2-5), isoforms 2-4 are the most likely to be involved in the preBötC because XE991 effectively blocks these isoforms with an IC 50 < 10 μM while its IC 50 at homomeric KCNQ5 channels is ~50 μM (Wang et al., 1998;Lerche et al., 2000;Schroeder et al., 2000;Schroder et al., 2001;Rajani et al., 2016). Data were generated using rhythmic medullary slices from newborn rat. Thus, the role of KCNQ in the intact, mature network remains to be explored. The role of KCNQ in burst termination relative to other mechanisms, including synaptic depression  and Na + -dependent outward currents Krey et al., 2010), also remains to be measured.
An additional point to consider is the potential role of KCNQ in the termination of burstlets, the putative rhythmogenic substrate, as opposed to bursts, which are obligatory for patterned motor output. Inspiratory rhythm generation is an emergent property of recurrently connected preBötC neurons that produce rhythmic "burstlets, " which are a subthreshold event from the standpoint of motor output. If network excitability is sufficiently high, then burstlets cross a threshold and trigger a full-amplitude burst that is transmitted out of the preBötC to premotoneurons, motoneurons, and inspiratory muscles (Kam et al., 2013;Del Negro et al., 2018;Ashhad and Feldman, 2020;Kallurkar et al., 2020). In the context of burst termination, synaptic depression, Na + /K + ATPase, I K-ATP , and I K-Na all depend on robust burst activity. But activity is much lower during a burstlet, and therefore, activity-dependent mechanisms may be less relevant in burstlet termination. KCNQ, in contrast, is voltage dependent and provides a compelling candidate that could contribute to burstlet termination as long as depolarization occurs.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, for reanalysis or meta analyses.

ETHICS STATEMENT
All experimental procedures were approved by the University of Alberta Faculty of Medicine Animal Welfare Committee and were performed in accordance with the Guidelines for the Care, Handling, and Treatment of Experimental Animals from the Canadian Council on Animal Care.

AUTHOR CONTRIBUTIONS
AR, AK, WM, and GF participated in the study design. AR and AK collected data. AR, AK, CDN, and GF wrote the manuscript. All authors were involved with manuscript revision, and read and approved the submitted version.