Presynaptic Mechanisms and KCNQ Potassium Channels Modulate Opioid Depression of Respiratory Drive

Abstract

Opioid-induced respiratory depression (OIRD) is the major cause of death associated with opioid analgesics and drugs of abuse, but the underlying cellular and molecular mechanisms remain poorly understood. We investigated opioid action in vivo in unanesthetized mice and in in vitro medullary slices containing the preBötzinger Complex (preBötC), a locus critical for breathing and inspiratory rhythm generation. Although hypothesized as a primary mechanism, we found that mu-opioid receptor (MOR1)-mediated GIRK activation contributed only modestly to OIRD. Instead, mEPSC recordings from genetically identified Dbx1-derived interneurons, essential for rhythmogenesis, revealed a prevalent presynaptic mode of action for OIRD. Consistent with MOR1-mediated suppression of presynaptic release as a major component of OIRD, Cacna1a KO slices lacking P/Q-type Ca²⁺ channels enhanced OIRD. Furthermore, OIRD was mimicked and reversed by KCNQ potassium channel activators and blockers, respectively. In vivo whole-body plethysmography combined with systemic delivery of GIRK- and KCNQ-specific potassium channel drugs largely recapitulated these in vitro results, and revealed state-dependent modulation of OIRD. We propose that respiratory failure from OIRD results from a general reduction of synaptic efficacy, leading to a state-dependent collapse of rhythmic network activity.

Keywords: KCNQ; opioid; preBötC; presynaptic; respiratory depression.

FIGURE 1

DAMGO, a mu-opioid receptor specific agonist, suppresses fictive inspiratory bursts recorded from in vitro preBötC medullary slices. (A) Integrated extracellular bursts recorded from an isolated preBötC, in response to bath application of increasing titers of DAMGO (30, 100 nM), followed by wash. Two minute segments shown at expanded time scale for baseline (black), 30 nM DAMGO (blue), 100 nM DAMGO (red), and wash (green). (B) Summary of burst frequency suppression by DAMGO, yielding an IC₅₀ of ∼30 nM (N = 19 slices; ^∗∗∗p < 0.0001; paired t-test). Wash restores burst frequency. Mean and SE plotted, with individual replicant values.

FIGURE 2

Exemplary records of fictive inspiratory bursts recorded from in vitro preBötC slices in response to increasing titers of activators or blockers of KCNQ and GIRK potassium channels, mimicking or reversing opioid-induced respiratory depression (OIRD). Integrated recording (top), instantaneous burst frequency (below). (A) Inspiratory bursts as a function of increasing titers of ICA 69673, a KCNQ activator (0.1, 0.5, 2.0, in mM), mimicking OIRD. (B) Inspiratory bursts as a function of increasing titers of retigabine (RTG), an FDA-approved KCNQ activator (0.1, 0.5, 1.0, 3.0, in mM), mimicking OIRD. (C) Rescue of DAMGO-induced OIRD (100 nM) with increasing titers of Chromanol 293B (293B), a KCNQ blocker (10, 50, 100, in mM). (D) Rescue of DAMGO-induced OIRD (100 nM) with increasing titers of XE991, a KCNQ blocker (1, 3, 30, 60, in mM). (E) Failure to rescue DAMGO-induced OIRD (100 nM) with increasing titers of TertiapinQ (TPQ), a GIRK blocker (5, 15, 60, 100, in nM). (F) Failure to mimic OIRD by increasing titers of ML297, a GIRK activator (3, 10, 30, in mM). Transient apneas observed at the beginning of washes (A–D) are an artifact of temperature drop or oxygen desaturation from perfusion exchanges.

FIGURE 3

Pharmacological interrogations of in vitro preBötC slices suggests a modulatory role for KCNQ potassium channels in opioid-induced respiratory depression (OIRD), and a minor role for GIRK potassium channels. (A, left) Dose-response curves of inspiratory burst frequencies in response to increasing titers of DAMGO, and two activators of KCNQ potassium channels (ICA 69673, Retigabine) which act on different structural domains of the channel subunit. Both activators inhibit burst frequencies with an IC₅₀ = ∼0.7–1.0 μM, comparable to their EC₅₀ for activation of KCNQ channels. (A, right) Dose-response curves of inspiratory burst frequencies in response to increasing titers of ML297, a GIRK1 subunit specific activator. Modest depression of burst frequencies, at concentrations 10–100-fold higher than the EC₅₀s (0.16–0.9 μM) for GIRK1 containing heteromeric channels. (B, left) Dose-response curves of inspiratory burst frequency in response to increasing titers of the KCNQ blockers XE991 and Chromanol 293B (293B), applied in the presence of DAMGO (100 nM) to suppress respiratory rhythms. (B, right) Dose-response curves of inspiratory burst frequency in response to increasing titers of TertiapinQ (TPQ), a GIRK-specific blocker, applied in the presence of DAMGO (100 nM). No recovery of respiratory burst frequency was observed at the highest concentration of TPQ (100 nM). By contrast, both KCNQ blockers partially rescued respiratory rhythms suppressed by DAMGO, at relatively high concentrations (20–100 μM). Shown above plots are the EC₅₀s and IC₅₀s of each compound for specific molecular species of homo- and hetero-tetrameric GIRK and KCNQ channels reported in the literature (see text and Table 2 for references). Mean and SE plotted for DAMGO and TertiapinQ; median and IQR plotted for all other datasets, along with individual replicant values: DAMGO (N = 35), ICA 69673 (N = 14), Retigabine (N = 6), ML297 (N = 6), XE991 (N = 7), Chromanol 293B (N = 5), TertiapinQ (N = 9).

FIGURE 4

KCNQ blockers (XE991, Chromanol 293) have no effect on baseline inspiratory frequencies from in vitro preBötC transverse slice preparations. (A) Examples of integrated extracellular preBotC recordings with corresponding voltage traces (left) and summary plots (right) for XE991. (B) Integrated extracellular preBotC recordings with corresponding voltage traces (left) and summary plots (right) for Chromanol 293B (293B). No statistical significance observed between baseline and all measured concentrations, for XE991 (N = 7) or Chromanol 293B (N = 8). Median and IQR plotted for XE991; mean and SE plotted for Chromanol 293B, with individual replicant values.

FIGURE 5

RT-PCRs of micro-dissected preBötC “islands” reveal transcripts for Kcnq1-5, Girk1-4, and Mor1. However, heterologous co-expression of MOR1 with KCNQ4/3 or KCNQ5/3 in Xenopus oocytes provide no evidence for direct coupling downstream of activated MOR1. (A) Transcripts for all Kcnq1-5 (Q1–Q5) subunits were detected, along with mu-opioid receptor (Mor1) and Girk1-4 subunits. Known preBötC molecular markers also robustly detected, including Substance P receptor (Tacr1), gastrin-release peptide receptor (Grpr1), somatostatin (Sst). Weaker reactions detected for somatostatin receptor (Sstr2), and Substance P (Tac1). For each primer set, paired RT-PCRs were performed with first-strand single-stranded cDNAs generated with (+) and without (-) reverse transcriptase. Expected sizes for PCR products (in bp): Kcnq1 (70), Kcnq2 (264), Kcnq3 (117), Kcnq4 (123), Kcnq5 (89), Mor1 (114), Girk1 (102), Girk2 (174), Girk3 (85), Girk4 (152), Sstr2 (246), Tacr1 (221), Grpr1 (169), Sst (250), Tac1 (110). (B) Two-electrode voltage clamp recordings from Xenopus oocytes co-injected with cRNAs encoding MOR1 and either GIRK1(F127S), KCNQ4+KCNQ3 (KCNQ4/3) or KCNQ5+KCNQ3 (KCNQ5/3). Current traces are shown before (left) and after (right) bath application of DAMGO (200 nM). Activation of MOR1 by DAMGO augments GIRK1(F127S) currents twofold, but had no effect on KCNQ4/3 or KCNQ5/3 current amplitudes. Reduction of KCNQ5/3 currents reflects time-dependent “rundown,” independent of DAMGO application. (C) Summary of current amplitudes after DAMGO (200 nM) for GIRK1(F127S) + MOR1 (N = 7), KCNQ4/3 + MOR1 (N = 5) and KCNQ5/3 + MOR1 (N = 11), normalized to controls without DAMGO. Current amplitudes plotted as median and IQR, with individual replicant values. Statistical significance: ^∗∗∗p = 0.0006, ^∗∗p = 0.0022–0.0025, by Mann–Whitney.

FIGURE 6

Genetic reduction of Ca_V2.1 (Cacna1a) function sensitizes in vitro preBötC inspiratory rhythms to depression by DAMGO. (A) Continuous plot of instantaneous burst frequency from a heterozygous Ca_V2.1 KO/+ preBötC slice, in response to increasing titers of DAMGO (0.1, 0.6, 1, 3, 6, 10, 30, in nM), followed by Naloxone (NX) (1, 2, in mM); dashed lines mark solution exchanges. Steep reduction of burst frequency at 6 nM DAMGO (B) WT (C3H strain) preBötC slice instantaneous burst frequency plot, in response to increasing titers of DAMGO (1, 3, 6, 10, 30, in nM); dashed lines mark solution exchanges. Slight depression of burst frequency with 6–10 nM DAMGO. (C) Summary of burst frequencies in response to DAMGO titers for heterozygous Ca_V2.1 KO/+ (N = 6) preBötC slices. (D) Summary of burst frequencies in response to DAMGO titers for WT (N = 16) preBötC slices. Heterozygous Ca_V2.1 KO/+ respiratory burst frequencies are increased ∼10-fold in sensitivity to DAMGO depression, relative to WT (Ca_V2.1 KO/+ IC₅₀ = ∼1 nM; WT IC₅₀ = ∼10 nM). Median and IQR plotted, with individual replicant values. Statistical significance: ^∗∗∗p = 0.0003–0.0009, ^∗∗p = 0.0012, ^∗p = 0.0159–0.0188 by Mann–Whitney.

FIGURE 7

Sparse labeling of Dbx1-derived (Dbx1⁺) preBötC neurons using Dbx1^Cre–ET2 mouse line, without tamoxifen treatment. (A) Embryonic tamoxifen injection (e10.5) labels a dense column of Dbx1⁺ cells with tdTomato extending radially from the dorsal ventricular surface through the ventrally-located preBötC, viewed by Dodt-IR optics (Ai) and RFP fluorescence (Aii). This slice from a P7 animal heterozygous for Dbx1^Cre–ERT2 and Ai27 (Dbx1^Cre–ERT2/+; Rosa26^Ai27/+). (B) Similar slice without tamoxifen sparsely labels individual Dbx1⁺ cells in a similar pattern as in 7A. This slice from an P7 animal homozygous for Dbx1^Cre–ERT2 and Ai27 (Dbx1^{Cre–ERT2/Cre–ERT2}; Rosa26^Ai27/Ai27). Viewed by Dodt-IR (Bi) and RFP fluorescence (Bii). (C) Sparsely-labeled fluorescent neurons and astrocytes (D) from the same slice in 7B at higher magnification.

FIGURE 8

mEPSCs recorded from Dbx1-derived (Dbx1⁺) preBötC neurons support a presynaptic site of action for DAMGO and KCNQ potassium channels. (A) Dbx1⁺ preBötC neuron labeled with tdTomato, without tamoxifen, imaged with Dodt-IR (left) and RFP fluorescence (right). Slice homozygous for Dbx1^Cre–ERT2 and Ai14 (Dbx1^{Cre–ERT2/Cre–ERT2}; Rosa26^Ai14/Ai14). (B) Representative mEPSCs recorded from a identified Dbx1⁺ inspiratory neuron in response to sequential bath application of TTX (1 μM), DAMGO (100 nM), and XE991 (20 μM). Currents measured at a holding potential of -60 mV. (C) Representative cumulative fractional distribution plot of mEPSC inter-event intervals (IEIs) recorded from an inspiratory Dbx1⁺ neuron, in response to TTX (black), DAMGO,TTX (red), and XE991,DAMGO,TTX (blue). DAMGO,TTX (red) distribution is significantly shifted toward longer IEIs relative to either TTX (black) or XE991,DAMGO,TTX (blue) distributions; p = 0.05 (Paired Wilcox–Signed Rank test; modified Kolmogorov-Smirnov), whereas TTX (black) and XE991,DAMGO,TTX (blue) distributions are not significantly different (Paired Wilcox–Signed Rank test; modified Kolmogorov–Smirnov). (D) Summary of pairwise comparisons of mEPSC cumulative fractional inter-event interval distributions (IEIs) from individual neurons in response to TTX (N = 7), DAMGO,TTX (N = 7) and XE991,DAMGO,TTX (N = 7), by paired Wilcox–Signed Rank tests (modified Kolmogorov-Smirnov). Significant ranked differences in mEPSC IEI distributions denoted by colored boxes [p = 0.05; orange (>), blue (<)]. Orange denotes statistically significant shifts toward longer IEIs, blue denotes significant shifts toward shorter IEIs. Non-significance denoted by green (=). (E) No consistent changes in mEPSC amplitudes from Dbx1⁺ neurons in TTX (1 μM) after application of DAMGO (100 nM) and XE991 (20 μM). Mean (bar), SE (box) and minimum/maximum (whiskers) plotted for each recorded neuron following sequential application of TTX (black), DAMGO (red), and XE991 (blue). (F) Pooled mEPSC means from each cell (N = 7) for each drug treatment (TTX; DAMGO,TTX; XE991,DAMGO,TTX). Pairwise comparisons between all combinations of drug conditions revealed no statistically significant differences (Mann–Whitney).

FIGURE 9

Whole-body plethysmography from awake, unanesthetized neonatal and adult mice, in response to systemic delivery of GIRK1 activator (ML297) or KCNQ activator (retigabine; RTG). Profound suppression of respiratory frequency by KCNQ activator, but only modest suppression by GIRK1 activator. (A) Modest reduction of respiratory frequency with ML297 (50 mg/kg, IP) from a neonatal pup. Representative plethysmography recordings before (Ai, top), and after drug injection (Ai, bottom). (Aii) Distributions of instantaneous breath frequencies from the same animal, fitted to single Gaussian distributions. (B) Large reduction of respiratory frequency with retigabine (10 mg/kg, IP) from a neonatal pup. Plethysmography records before (Bi, top) and after drug injection (Bi, bottom). (Bii) Distributions of instantaneous breath frequencies from the same animal, fitted to single Gaussian distributions. (C) Summary of plethysmography data for neonates (P7–13) and adults (P25–60), normalized to baseline respiratory frequencies. Both neonates and adults exhibited modest and similar reductions (∼10%) in respiratory frequency with ML297. However, retigabine caused large reductions in respiratory frequency in both neonates (46% reduction) and adults (20% reduction). Adult respiratory suppression by retigabine was significantly stronger than ML297 (p = 0.012; t-test). Results from vehicle controls (DMSO; IP) with neonates and adults were not significantly different and pooled. Population mean and SE plotted, along with individual responses derived from means of fitted Gaussian curves, for each condition, for neonatal retigabine (N = 11), neonatal ML297 (N = 12), adult retigabine (N = 9), adult ML297 (N = 6), DMSO control (N = 8). Statistical significance: ^∗∗∗p < 0.0001, ^∗∗p = 0.0003, by un-paired t–test.

FIGURE 10

State-dependent reversal of respiratory suppression by morphine with XE991. Whole-body plethysmography from unanesthetized neonatal and adult mice, in response to systemic IP injections with morphine and XE991, a KCNQ-specific blocker. (A) Representative plethysmography recordings from an adult “responder” mouse during baseline breathing (Ai, blue), after morphine injection (150 mg/kg) (Ai, red), and subsequent injection with XE991 (3 mg/kg) (Ai, green). (Aii) Distributions of instantaneous breath frequencies from the same animal under each condition, fitted to single Gaussian distributions. XE991 rescue of respiratory suppression by morphine was variable, and dependent upon the respiratory frequency observed after morphine injection. (B) Neonatal “responders” exhibited strong suppression of respiratory frequency to morphine, below a 2.9 Hz “threshold” frequency. All animals partially recovered respiratory frequency with XE991 (N = 14, p < 0.0001; paired t-test). (C) Adult “responders” similarly exhibited strong respiratory suppression to morphine, below a 2.0 Hz “threshold” frequency. All animals partially recovered respiratory frequency with XE991 (N = 8, p = 0.0003; paired t-test). (D) Neonatal “non-responders,” by contrast failed to exhibit morphine suppression of respiratory frequency below a 2.9 Hz “threshold.” None of these animals increased respiratory frequency more XE991 (N = 5, ns; paired t-test). (E) Adult “non-responders” similarly failed to exhibit morphine suppression of respiratory frequency below a 2.0 Hz “threshold.” None of these animals responded to XE991 with an increased respiratory rate (N = 5, ns; paired t-test). ^∗∗∗p < 0.0001, ^∗∗p = 0.0003, ns, non-significant.

FIGURE 11

Model for presynaptic site of action by opioids underlying opioid-induced respiratory depression (OIRD). We hypothesize that glutamatergic presynaptic terminals onto Dbx1-derived (Dbx1⁺) inspiratory interneurons within the preBötC compartmentalize mu-opioid receptors (MOR1) and KCNQ potassium channels, in addition to the normal complement of presynaptic calcium channels associated with vesicular release, including P/Q-type (Ca_V2.1) and N-type (Ca_V2.2) calcium channels. Activation of MOR1 leads to suppression of presynaptic calcium channel function, via direct binding with Gβ/γ generated by receptor activation. This leads to reduced synaptic transmission across the excitatory inspiratory circuit, defined by an essential kernel of Dbx1⁺ glutamatergic neurons. Because rhythmogenesis within this circuit is largely network-driven, reduced synaptic efficiency leads to collapse of rhythmogenic capacity, and opioid-induced apnea. We further propose that active KCNQ potassium channels contribute to maintaining presynaptic resting membrane potentials. Blocking these channels may depolarize presynaptic terminals, mitigating the effects of opioid-mediated suppression of calcium channel function. We emphasize that our model proposes that presynaptic KCNQ channels act independent of direct signaling downstream from MOR1 activation, but function in parallel with MOR1-mediated inhibition of presynaptic calcium channels by virtue of co-compartmentalization to the presynaptic terminal. Sufficient restoration of synaptic transmission may thus revive rhythmogenesis, in spite of ongoing signaling by activated MOR1. Our model of OIRD suggests a relatively minor role for MOR1-coupled activation of GIRK potassium channels, however somatic GIRK currents have been reported by other studies.