Experiments were conducted on adult male Sprague Dawley rats (Charles River: Crl: CD-SD rats), weighing between 350 to 480 grams. Rats had access to food and water ad libitum and were kept in a 12-hour daily light-dark cycle. All procedures and experimental protocols were approved by the Institutional Animal Care and Use Committee at the University of Florida (Protocol # 201609515). Experiments performed in this manuscript also conform to the principles and regulations as described in the Editorial by Grundy (2015). All anesthetized animals were sacrificed via either transcranial perfusion, or anesthetic urethane overdose, permanently terminating heart rate, blood pressure and phrenic activity.

Morphine sulfate and (+)-naloxone were obtained from the National Institute of Drug Abuse Drug Supply Program. Morphine and (+)-naloxone were dissolved in a sterile saline solution and protected from light exposure. Stock drug solutions remained viable at room temperature for up to 3 weeks, after which the drugs were discarded, and a fresh solution was made as needed. Drugs or vehicle (sterile saline) were administered via subcutaneous (s.c.) injections (lower left abdominal quadrant) 3 hours prior to beginning terminal experiments.

Morphine was delivered at a dose of 4 mg/kg (s.c.): a dose shown to activate inflammatory signaling that persists for hours (Hutchinson et al., 2008a). This dose is predicted not to agonize mu opioid receptors effectively beyond 2 hours post-injection (Hutchinson et al., 2008a), and is not expected to compromise overall activity such as locomotion or respiratory motor output. This was an important consideration in our studies since our goal was to experiment with long-latency morphine effects that could activate an inflammatory cascade without causing a persistent respiratory depression. (+)-Naloxone was administered at 10 mg/kg (s.c). (+)-Naloxone is an inactive enantiomer of the corresponding mu opioid receptor antagonist (Iijima et al., 1978; Valentino et al., 1983; Lewis et al., 2012; Doyle & Murphy, 2018), and does not antagonize mu opioid receptors. Instead, (+)-naloxone antagonizes TLR4 (Hutchinson et al., 2008b; Lewis et al., 2012; Wang et al., 2016). The (+)-naloxone dose of 10 mg/kg was chosen because it results in plasma concentrations in the pharmacologically active range (Hutchinson et al., 2010; Lewis et al., 2012), but without the off-target effects known to occur at higher doses (≥ 32 mg/kg) (Tanda et al., 2016).

Anesthesia was induced with 3% inspired isoflurane in a Plexiglas chamber, and maintained with a nose cone (60% O2 balance N2). After confirming absence of any foot-pinch withdrawal reflex, a midline ventral cervical incision was made in the neck, the trachea was exposed and sectioned below the larynx; a tracheal tube (polyethylene catheter; PE 240; Intramedic, MD, USA) was inserted into the trachea to deliver isoflurane and mixed gases via artificial ventilation (2.5% isoflurane in 60% O2/balance N2). Artificial ventilation was achieved with a rodent ventilator (Tidal volume = 0.7 ml / 100g body weight; Rodent Respirator model 683, Harvard Apparatus, South Natick, MA, USA; ventilator rate 70 breaths/min). A rapidly responding flow-through CO2 analyzer (Capnogard, Novametrix, Wallingford, CT, USA) was placed on the expired side of a Y-tube connected to the tracheal cannula to monitoring end-tidal PCO₂ (PetCO₂). The tail vein was cannulated (24 gauge, Surflo, Elkton, MD, USA) so that rats could be slowly converted from isoflurane to urethane anesthesia (2.1 mg/kg; i.v.). Conversion to urethane was carried out at least one hour before beginning experimental protocols. Absence of foot-pinch withdrawal reflexes was used to test adequacy of anesthesia; supplemental anesthetic was given as required. Once urethane conversion was complete, fluids were given through the same tail vein cannula to maintain acid-base balance (1.5–2.5 ml/hour, started ~1 hour after the beginning of surgery; 1:4 solution of 8.4% sodium bicarbonate mixed in standard lactated Ringer’s solution). Body temperature was monitored with a rectal thermometer (Fischer Scientific, Pittsburgh, PA, USA) and maintained (37.5 ± 1°C) with a custom-made heated surgical table.

Rats were bilaterally vagotomized in the mid-cervical region to prevent entrainment of respiratory motor output with the ventilator, and paralyzed with pancuronium bromide (2 mg/kg; Sigma-Aldrich, St. Louis, MO, USA) to prevent the rats from “fighting” the ventilator. Flexible polyethylene tubing (PE 50; Intramedic MD, USA) was inserted into the femoral artery, and connected to a pressure transducer (Grass Instruments) to monitor arterial blood pressure. The same arterial line was also used to withdraw blood samples (70 ul) for blood gases measurements (i.e. PaO₂ and PaCO₂), acid-base balance and pH using a blood gas analyzer (ABL 90 Flex, Radiometer, Copenhagen, Denmark). Using a dorsal approach, the left phrenic nerve was isolated, cut distally and de-sheathed. Nerves were kept moist by a saline-soaked cotton ball until ready to be placed in suction recording electrodes to record nerve activity.

Custom-made glass suction electrodes were filled with 0.9% saline. Nerve activity was amplified (10K, A-M systems, Everett, WA), filtered (bandpass 100–5000 Hz), integrated (time constant 50 msec), digitized (Micro1401, Cambridge Electronic Design, UK) and analyzed using Spike 2 software (Cambridge Electronic Design, UK; version 8.08). Since peak amplitude of integrated inspiratory phrenic nerve bursts correlates with tidal volume in spontaneously breathing animals (Eldridge, 1976), integrated phrenic nerve burst amplitude served as an index of respiratory motor output. Inspiratory phrenic burst amplitude was determined immediately prior to blood samples drawn during baseline & hypoxic conditions, and at 30 and 60-minutes post-AIH. Measurements were made at equivalent times in time control experiments that did not receive AIH.

Triple immunofluorescence labeling was employed to stain for phospho-p38 MAPK, CD11b (microglial marker) and NeuN (neuronal marker). Rats used for immunohistochemistry did not undergo neurophysiological procedures. Approximately 4.5 hours following injections of the vehicle or drugs (see Study 3 below), rats were briefly anesthetized with isoflurane (~4-min under 3.5% isoflurane) and immediately perfused transcardially with ice-cold phosphate buffered saline (PBS, pH 7.4) followed by 4% buffered paraformaldehyde (PFA, pH 7.4). Cervical segment of the spinal cord (C3-C5) was excised, post-fixed in 4% paraformaldehyde overnight, and cryoprotected in 30% sucrose at 4°C. 40μm transverse sections were cut using a freezing microtome (Leica SM 2010R, Germany) and stored in anti-freeze solution (−20°C) until the day of staining. Transverse tissue sections were numbered sequentially and 2 sections per spinal segment (C3, C4 and C5) were used to represent C3-C5 for each animal (total of 6 spinal sections per animal). Free-floating sections were washed in 0.1M PBS containing 0.1% Triton-X100 (PBS-TX; 3 × 5-minute washes). Tissues were then blocked with 5% normal donkey serum (GeneTex, CTX30972) for 1 hour at room temperature to block non-specific binding sites. Staining was performed by incubating free-floating tissues with primary antibodies against CD11b (mouse host, 1:1000, EMD Millipore, USA), phospho-p38 MAPK (rabbit host, 1:500, Cell Signaling Technology, Inc.) and NeuN (chicken host; 1:1500, EMD Millipore, USA) over night at 4C (diluted in 2.5% donkey serum in PBS). Tissues were then washed, followed by incubation with secondary antibodies (1 hour room temperature in PBS-0.1% TX) to label CD11b (Donkey anti-mouse Alexa Fluor 488, 1:1000, Invitrogen), phospho-p38 MAPK (Donkey anti- rabbit, 1:500, Alexa Fluor 555, Invitrogen, USA), and NeuN (Donkey anti-chicken Alexa Fluor 647, 1:1500, Jackson Immuno Research, USA). Sections were then immediately washed in PBS and mounted on slides using VectaShield antifade hard-set mounting medium (Vector Laboratories, California, USA, Product # H-1400).

Serum was collected from a group of rats approximately 4.5 hours following a single systemic injection of morphine (s.c., 4mg/kg; n=6). Whole blood was collected into an anti-coagulant free serum separation plastic tube (600 ul per animal; BD Microtainer SST Gold Top, Item Number BD365967; Becton, Dickinson and Company, NJ, USA). Blood was allowed to clot at room temperature for 30-minutes. Clotted samples were then centrifuged at 2500 ×g for 10 minutes at 4 °C. Serum was removed and stored in cryovials at −80 °C until ready for analysis.

At least 1 h after conversion to urethane anesthesia, apneic and recruitment CO2 thresholds of respiratory nerve activity was determined by lowering inspired CO2 (or increasing ventilation rate in some cases) levels until rhythmic respiratory nerve activity ceased. After ~60 seconds, inspired CO2 was slowly increased until rhythmic respiratory nerve bursts resumed. The end-tidal PCO2 at which respiratory nerve activity stopped and then resumed were considered the apneic and recruitment thresholds, respectively. Baseline conditions were then established by holding end-tidal PCO2 ∼2 mmHg above the recruitment threshold and allowing sufficient time to establish stable nerve activity (> 20min). During baseline recordings, an arterial blood sample was taken to document baseline blood gas levels. Arterial PCO2 was maintained isocapnic (± 2 mmHg) with respect to this baseline value throughout experiments by actively manipulating inspired carbon dioxide concentration and/or ventilation rate. Baseline oxygen levels (∼60% inspired oxygen, balance N2 and CO2; PaO2 ≥ 150 mmHg) were maintained for the duration of experiments except during hypoxic challenges; targeted arterial PaO2 levels during hypoxic episodes were 35–50 mmHg. At the end of every experimental recording period, we assessed phrenic motor output response to a chemosensory stimulus composed of combined hypoxia and hypercapnia. Chemoreflex activation of breathing is an important component of the ventilatory control system. For example, chemosensory responses to hypoxia and hypercapnia are important feedback mechanisms that are critical for maintenance of blood gas homeostasis, as well as eliciting appropriate reflex ventilatory responses when facing challenges that compromise adequate lung ventilation. Chemosensory stimulation at the end of each neurophysiological experiment consisted of an inspired gas mixture consisting of 10% O2, 7% CO2 and balance N2 (total duration of two minutes).

Respiratory nerve activities were analyzed using Spike 2 software (Cambridge Electronic Design, UK; version 8.08). Integrated phrenic nerve inspiratory burst amplitudes were averaged over 1-minute bins at each experimental time point. Activities were analyzed during baseline, hypoxia, and at 30 and 60-minutes post-hypoxia. Changes (Δ) in nerve burst amplitudes were normalized and reported as percentage change from baseline (baseline = 0). Therefore, any value below zero is a decrease whereas values above zero are increases relative to baseline. Burst frequencies were also presented as number of breaths per minute. Respiratory activities were also analyzed at equivalent time points in time matched control animals that were not exposed to hypoxia. We also measured and analyzed mean arterial pressure (MAP), arterial CO2 pressure (PaCO2), arterial O2 pressure (PaO2), pH and standard base excess (SBEc) (Table 2). Values for these biological variables were not normalized and were presented as absolute values in respective figures and tables. Phrenic nerve inspiratory amplitude was also quantified during a chemosensory challenge at the end of each recording period. This was accomplished by analyzing steady-state maximal phrenic nerve amplitude for one minute during hypoxic hypercapnia (10% inspired O2 combined with 7% inspired CO2). Statistical comparisons between treatment groups were made using a One-Way ANOVA followed by the Tukey significance post hoc test for individual comparisons (SigmaPlot version 14; Systat Software Inc., San Jose, California, USA). Baseline phrenic nerve amplitudes, baseline respiratory frequency, baseline blood pressure, acute short-term hypoxic phrenic responses, and blood pressure change during hypoxia were also compared using a one-way ANOVA with Tukey post hoc test. Differences between groups were considered significant if p < 0.05. All average group data are presented with standard deviation.

Immunofluorescence images were analyzed using a custom-written MATLAB algorithm (MathWorks, Natick, MA, USA) designed for quantification of fluorescence intensities for molecules of interest in the ventral horn of the cervical spinal cord (a well-defined anatomical region containing phrenic motor neurons). The analysis was conducted by an investigator who was blinded to the experimental conditions. The algorithm identified CD11b and NeuN-labelled cells, and quantified signal intensity of phospho-p38MAPK immune-reactivity in each cell type. Methods for protein-specific signal intensity quantification within the ventral horn of the cervical spinal cord have been previously described elsewhere (Seven et al., 2018) in detail. CD11b (green color) and NeuN positive (purple color) cells were located within the ventral horn of the cervical spinal cord (cervical segments C3-C5) using a custom adaptive thresholding algorithm in MATLAB. The adaptive threshold was calculated by constructing a pixel intensity histogram from the image. First, a pixel intensity histogram is constructed in the immediate region of each detected cell in the ventral horn. The pixel intensity corresponding to fixed percentile value was used as the threshold value across all images to identify each cell type of interest. The pixels above the adaptive threshold were considered CD11b and NeuN-positive. Selection of a fixed percentile threshold returns a higher threshold value for an image with high signal and background intensities and a lower threshold value for an image with low signal and background intensities. CD11b and NeuN-positive images were binarized using the adaptive threshold, thus, CD11b and NeuN-positive areas were assigned the value of unity, whereas CD11b and NeuN-negative areas were set to zero. Intensity of the p38 MAPK labelling (red color) was calculated by averaging the pixel intensities within each identified cell type. The coordinates of CD11b and NeuN-positive pixels/areas were used to measure fluorescence intensities of phospho-p38 MAPK. Within each cell type, minimal detectable staining for “p38 MAPK” is expected due to minimal cellular auto-fluorescence that occurs with most staining procedures. Minimal detectable staining is a qualitative feature indicating the difference between the signal and the background fluorescence intensities. Although, we subtract general diffuse background fluorescence in our analyses, non-specific binding fluorescence and cellular auto-fluorescence remain throughout the cell bodies and projections. On the other hand, phospho-p38 MAPK positive regions are localized at the nucleus and clearly brighter than the background staining in other compartments of the cell. Therefore, final fluorescence intensities were determined after subtraction of local background labeling by determining the median value. For statistics, analyses of multiple comparisons were performed by ANOVA with Tukey’s significant difference test as post hoc test.

To determine if morphine had long-latency effects on respiratory motor plasticity, rats were pretreated with vehicle or morphine (4 mg/kg) 3 hours prior to initiation of LTF experiments. A second group of rats were pretreated with the TLR4 inhibitor, (+)-naloxone (10 mg/kg), 20-minutes prior to morphine. Baseline phrenic nerve burst amplitude was compared among all groups, including time controls and mAIH exposed rats. There was no significant difference in baseline phrenic nerve amplitudes in any group (p = 0.185 in the overall ANOVA, Fig. 1A). This demonstrated that 1) none of the drug injections affected basal phrenic nerve motor output during neurophysiological recordings, and 2) normalization to baseline measurements as a percentage change from baseline was appropriate to quantify and compare the magnitude of phrenic nerve amplitude in various treatment groups (see sections above). Like baseline nerve burst amplitude, there was no difference in baseline breathing frequency comparing the various treatment groups (p = 0.843 in the overall ANOVA; Fig. 1B). This suggested that in the time frame of our neurophysiological recordings, which was 4.5 hours following morphine or vehicle administration, drug treatments did not elicit a detectable depression of breathing as far as breathing frequency or motor drive (i.e. burst amplitude) were concerned.

We also compared the short-term hypoxic phrenic response in groups that were exposed to moderate AIH (Fig. 2A and B). As expected, isocapnic hypoxic episodes triggered a robust increase in phrenic nerve burst amplitude, however, this increase was similar among treatment groups (vehicle: 61 ± 26%; morphine: 53 ± 24%; TLR4-inhibitor + morphine: 71 ± 24% increase above baseline; p < 0.001 versus baseline in each group; p = 0.270 for the difference among groups as determined by the ANOVA). Similarly, we found no difference in the short-term hypoxic frequency response comparing the groups that were exposed to mAIH (p = 0.892 in the overall NOVA). Thus, drug treatments did not have long-lasting effects on the short phrenic responses during acute intermittent hypoxic exposures.

We next compared mean arterial blood pressure (MAP) during baseline conditions, as well as during hypoxic episodes (Fig. 2A and C). No difference in baseline MAP were detected (vehicle: 99 ± 11 mmHg; morphine: 109 ± 13 mmHg; TLR4-inhibitor + morphine: 103 ± 11 mmHg; p = 0.225). As expected, there was a significant decrease in mean arterial pressure during hypoxic episodes in each group (p < 0.001, within each group), but MAP during hypoxia did not differ between groups (vehicle: 54 ± 21 mmHg; morphine: 53 ± 11 mmHg; TLR4-inhibitor + morphine: 56 ± 13 mmHg; p = 0.927 between groups).

The lack of significant differences in baseline physiologic variables (breathing frequency, phrenic amplitude and MAP) between morphine and vehicle-pretreated rats suggests that the pretreatment time was sufficient for morphine to be metabolized and no longer circulating in physiologically relevant quantities at the start of our neurophysiological recordings. To confirm this, in a separate rat group (n=6), we measured serum levels of morphine and its two primary metabolites: morphine-3-glucuronide (M3G) and morphine-6-glucuronide (M6G). Serum was collected ~4.5 hours after a single acute morphine injection — an equivalent time delay to the execution of pLTF protocol in neurophysiological recordings. Morphine serum concentration was 15.5 ± 6 ng/ml, M3G was 323.7 ± 147 ng/ml and M6G was below the lower limit of quantification (<1 ng/ml) (Table 3). Lack of serum M6G was consistent with previous published findings, indicating that morphine is not metabolized to M6G in rats (Kuo et al., 1991). Representative detection traces by mass spectral analysis as achieved by multiple reaction monitoring is shown in Figure 3.

To determine if acute morphine delivery affects mAIH-induced pLTF, rats were pre-treated with morphine (4 mg/kg, s.c.) 3 hours prior to initiation of experiments. Results were compared to vehicle-treated controls. Representative traces of integrated phrenic activity before, during and after mAIH are shown in Figure 4. Vehicle-treated control rats displayed a persistent increase in phrenic amplitude following mAIH, consistent with many previous published reports from our laboratory (Baker-Herman et al., 2004; Devinney et al., 2015; Tadjalli & Mitchell, 2019). 60-minutes post-mAIH phrenic nerve burst amplitude was significantly increased by 55 ± 11% above baseline levels, confirming the presence of pLTF (Fig. 4i and Fig. 5; n=6; p < 0.0001). In contrast, pretreatment with systemic morphine abolished mAIH-induced pLTF (8 ± 24% increase above baseline 60-min post-AIH; p = 0.226; Fig. 4ii and Fig. 5; n=8); phrenic nerve burst amplitude remained close to baseline levels throughout the post-hypoxic period. Morphine alone had no lasting effect on the stability of phrenic motor output in the time frame of these studies since phrenic output remained unchanged relative to baseline in time control rats (4 ± 20% above baseline at 60-min; p = 0.916; Fig. 4iv and Fig. 5; n=6). At 60-minutes post-mAIH phrenic burst amplitude was significantly greater in vehicle-treated rats exposed to mAIH versus morphine + mAIH (p= 0.0028) or morphine treatment alone (p = 0.0003). Thus, morphine pre-treatment activates mechanisms that block mAIH-induced pLTF.

Even a mild systemic inflammation triggered by lipopolysaccharides (LPS)—endotoxins that signal through TLR4 — undermines pLTF expression (Vinit et al., 2011; Huxtable et al., 2013). Since morphine is known to activate inflammatory cascades via TLR4 signaling, and systemic morphine pre-treatment blocks pLTF expression (Figure 4 and 5), we reasoned that morphine may block pLTF in a manner parallel to LPS endotoxins. Thus, we hypothesized that blocking TLR4 signaling prior to morphine delivery will prevent the impairment of pLTF. To block TLR4 signaling we used (+)-naloxone, which is the opioid inactive stereoisomer of the prototypical opioid antagonist naloxone. Unlike (−)-naloxone, which has an affinity for mu opioid receptors of 1 nM, (+)-naloxone has an affinity of greater than 10,000 nM for mu opioid receptors and is considered opioid receptor inactive (Iijima et al., 1978; Valentino et al., 1983). Rather, (+)-naloxone blocks TLR4 receptor signaling in vitro and in vivo (Hutchinson et al., 2008b; Hutchinson et al., 2010; Wang et al., 2016). Two groups received the TLR4 inhibitor: one group received the TLR4-inhibitor followed by morphine without mAIH (time control) and another group received the same drug treatment followed by exposures to mAIH.

Pretreatment with the TLR4 inhibitor, (+)-naloxone, prevented the morphine-induced impairment of pLTF; at 60-minutes post-mAIH, phrenic burst amplitude was significantly increased by 50 ± 6% above baseline levels (p < 0.0001; Fig. 4iii and Fig. 5; n=6). The magnitude of pLTF in rats that received the combination of TLR4 inhibitor and morphine was not different compared to vehicle-treated controls (vehicle + mAIH), demonstrating that the TLR4 inhibitor fully prevented the morphine-induced impairment of pLTF (50 ± 6% versus 55 ± 11% above baseline; p = 0.655). In time-control rats injected with the TLR4 inhibitor and morphine (without mAIH; n=6), there was no discernable effect on phrenic motor output since phrenic burst amplitude did not change compared to baseline throughout the 60-minute recording period (5 ± 16% above baseline at minute 60; p = 0.747; Fig. 4v and Fig. 5).

Since (+)-naloxone prevented morphine-induced pLTF impairment, we hypothesized that morphine blocked phrenic motor plasticity via TLR4 signaling. The next aim was to ensure that (+)-naloxone did not antagonize mu opioid receptors. Thus, in separate rat groups, morphine (4 mg/kg, sc) was administered while monitoring phrenic nerve activity. In these anesthetized rats, morphine caused potent depression, silencing phrenic nerve activity (Fig. 6A). We then questioned if acute morphine-induced respiratory depression could be reversed using (+)-naloxone, which would indicate mu opioid receptor antagonism. However, (+)-naloxone (10 mg/kg, sc) administration had no effect on acute morphine-induced respiratory depression since phrenic nerve activity remained absent (Fig. 6A and B). We then administered racemic (+/−)-naloxone, which contains the opioid active stereoisomer (−)-naloxone. In contrast to (+)-naloxone, racemic naloxone (1 mg/kg, sc), quickly reversed respiratory depression, fully restoring phrenic motor output (Fig. 6B; 61 ± 16 % increase above baseline after racemic naloxone injection; p = 0.0005). To control for the time difference between morphine administration and injection of antagonist, in a separate group of rats, we also administered (+)-naloxone at the same time as racemic naloxone in the previous experiments. (+)-Naloxone again had no effect on morphine-induced respiratory depression (Fig. 6C) since phrenic nerve activity remained absent. These findings demonstrate that, at the concentrations employed in this study, (+)-naloxone does not antagonize mu opioid receptors, consistent with previous literature reports (Iijima et al., 1978; Valentino et al., 1983).

Chemoreflex responses to hypoxia and hypercapnia are critical for maintenance of adequate breathing and for appropriate ventilatory responses in conditions that hinder lung ventilation, such as opioid-induced respiratory depression (Pattinson, 2008). Therefore, we determined if morphine affects phrenic responses to maximal chemoreflex activation during hypoxic hypercapnia (Fig. 7 and 8). In vehicle-treated rats exposed to mAIH, hypoxic hypercapnia triggered a 141 ± 30% increase in inspiratory phrenic nerve burst amplitude (p < 0.0001). This increase was greater compared to rats that received morphine + mAIH (83 ± 33% baseline; p = 0.034 comparing the means), or morphine alone (68 ± 25% baseline; p = 0.006 comparing the means). There was a trend for the TLR4 inhibitor, (+)-naloxone, to attenuate the morphine-induced inhibition of maximal chemoreflex responses. In mAIH-exposed rats that received the TLR4 inhibitor prior to morphine, hypoxic hypercapnia triggered a 133 ± 29% increase in phrenic amplitude (p < 0.0001), a response significantly larger than in morphine time controls or in rats that received morphine + mAIH (p = 0.003 in the overall ANOVA). Further, maximal chemoreflex responses were not different versus rats receiving vehicle + mAIH, demonstrating full chemoreflex restoration (p = 0.679 comparing vehicle + mAIH versus TLR4-inhibitor + morphine + mAIH). Thus, morphine-induced blunting of maximal chemoreflex responses may involve mechanisms that require TLR4 signaling. However, there were no significant differences in chemoreflex responses when comparing TLR4-inhibitor + morphine time controls, morphine time controls, or morphine + mAIH (morphine time controls: 68 ± 25%; morphine + mAIH: 83 ± 33% baseline; TLR4-inhibitor + morphine time controls: 100 ± 53 %, p = 0.374).

Our next aim was to determine whether morphine-induced deficits in pLTF are paralleled by p38 MAPK activation— a well characterized MAPK pathway downstream from TLR4 known to orchestrate inflammatory cascades. To determine whether morphine modulates p38 MAPK phosphorylation levels, we evaluated morphine effects on dually phosphorylated (enzymatically activated) p38 MAPK expression in the ventral horn of the C3-C5 segment of the cervical spinal cord (an anatomical region containing the phrenic motor nucleus) using optical density immunofluorescence. We evaluated p38 MAPK phosphorylation levels in both CD11b and NeuN-positive cells—markers for microglia and neurons, respectively (Fig. 9). Results were compared to vehicle control rats and a separate rat group that received the TLR4 inhibitor before morphine injections. In vehicle control rats, ventral spinal CD11b and NeuN-positive cells displayed minimal detectable phospho-p38 MAPK staining (Fig. 9; n=3 rats; 93 images in total). After acute morphine delivery (~4.5 hours post-drug injection; n=4 rats; 105 images in total), phospho-p38 MAPK was significantly increased in CD11-b positive cells (23 ± 7% increase in phospho-p38 MAPK intensity in morphine-treated versus vehicle-treated controls, Fig. 9 Niii; p=0.025). While morphine increased p38 MAPK phosphorylation in CD11b-positive microglia, no significant change was observed in NeuN-positive cells (1 ± 6% change in phospho-p38 MAPK staining intensity in morphine treated rats versus vehicle controls; p=0.20). Thus, morphine enhanced p38 MAPK phosphorylation/activation in ventral cervical spinal microglia, but not neurons.

Next, we analyzed p38 MAPK optical density in rats pre-treated with the TLR4 inhibitor before morphine delivery. In rats that were pretreated with the TLR4 inhibitor (n=3 rats; 78 images in total), morphine no longer increased microglial p38 MAPK phosphorylation (7 ± 9 % decrease in microglial p38 MAPK staining intensity versus vehicle controls; p = 0.191; Fig. 9 Niii). This was in sharp contrast to rats treated with morphine alone since we had observed a significant increase in p 38 MAPK phosphorylation with morphine treatment alone (see above). Thus, in the time frame of our studies, morphine enhances microglial p38 MAPK activity by a mechanism that requires TLR4 signaling. Rats receiving the TLR4 inhibitor followed by morphine did not exhibit significant change in neuronal p38 MAPK signaling (2 ± 5 % increase in neuronal p38 MAPK intensity versus vehicle controls; p = 0.154; Fig. 9 Niii).

Long-term opioid use leads to unfavorable side effects such as tolerance, dependence, and hyperalgesia (Cui et al., 2006; Cui et al., 2008; Watkins et al., 2009; de Freitas et al., 2019). The underpinnings of these effects involve forms of synaptic plasticity that modulates the efficacy of synaptic transmission. Opioid-mediated synaptic plasticity has been studied extensively in the mesocorticolimbic system in the context of drug addiction (Luscher & Malenka, 2011), and in the dorsal spinal cord in the context of hyperalgesia (Sandkuhler & Gruber-Schoffnegger, 2012). Chronic opioid use elicits epigenetic changes and lasting neural plasticity in the dorsal spinal cord (Sandkuhler & Gruber-Schoffnegger, 2012; Liang et al., 2013; Liang et al., 2014; Chao et al., 2016). Less is known regarding mechanisms whereby opioids influence neural plasticity with acute opioid administration, particularly in motor networks such as the phrenic motor system. Some studies demonstrate that key forms of neuroplasticity such as hippocampal long-term potentiation or depression are sensitive to even a single exposure to opioids (Krug et al., 2001; Wagner et al., 2001; Nugent et al., 2007; Drdla et al., 2009; Dacher & Nugent, 2011). Acute opioid administration leads to synaptic long-term depression in dorsal striatum slices in vitro (Atwood et al., 2014), and long-term potentiation in the dorsal spinal cord (Drdla et al., 2009; Zhou et al., 2010). Further investigations concerning mechanisms whereby opioids modulate neuroplasticity is of importance given the high prevalence of opioid prescriptions for pain management.

Here, we aimed to shed further light on the influence of acute opioid use on neuro-adaptive behaviors in the spinal cord by examining the effect of acute morphine delivery on the expression of respiratory motor plasticity. Morphine was chosen for this investigation since 1) morphine reaches significant blood concentrations quickly (~15 minutes); and 2) morphine agonizes mu opioid receptors for ~2 hours at the dose studied here (Hutchinson et al., 2008a). Thus, in the time frame of the present study, the morphine dose used is not expected to elicit persistent mu opioid receptor-induced respiratory depression; nevertheless, downstream pro-inflammatory TLR4 signaling is expected to persist in this time frame (e.g. downstream p38 MAPK phosphorylation; Fig. 9). We did detect morphine and its metabolite M3G in serum samples taken at a time corresponding to the start of neurophysiological recordings. M3G has no affinity for opioid receptors, but does activate TLR4 (Lewis et al., 2010; Doyle & Murphy, 2018). The concentration of morphine in serum (15.5 ± 6 ng/ml) equates to approximately 54 nM. Since brain morphine concentrations are an order of magnitude less than blood concentrations (Xie et al., 1999; Quillinan et al., 2011), measured concentrations are not sufficient to impact respiratory motor output via mu opioid receptor activation (Levitt et al., 2015; Levitt & Williams, 2018). Collectively, serum morphine measurements and the observations that baseline phrenic nerve amplitude and respiratory frequency were similar among all groups are consistent with the interpretation that mu opioid receptor-induced respiratory depression was absent/minimal at the time of our neurophysiological recordings.

Since pLTF was absent >4 hours post-morphine delivery, we suggest off-target mechanisms other than morphine-induced mu opioid receptor activation are responsible for impaired pLTF. Presence of pLTF in morphine-treated rats that received (+)-naloxone demonstrates that TLR4 activation was necessary in the mechanism whereby morphine undermines pLTF. This finding is consistent with predicted pro-inflammatory profiles following TLR4 activation since even mild inflammation triggered by low-dose LPS (a potent TLR4 agonist) abolishes mAIH-induced pLTF in rats (Vinit et al., 2011; Huxtable et al., 2013; Tadjalli et al., 2021). Thus, we predict that drugs that inhibit pro-inflammatory cytokine signaling (Hocker & Huxtable, 2018) and/or p38 MAP kinase (Huxtable et al., 2015) may restore pLTF following morphine delivery.

To our knowledge, only one other study has examined the effect of acute opioid delivery on respiratory motor plasticity. Ivancev and colleagues (2013) examined the effect of continuous intravenous remifentanil (a short-acting, potent mu opioid agonist) on pLTF expression in rats (Ivancev et al., 2013). The fundamental difference between our study and that of Ivancev and colleagues is that pLTF was suppressed only with continuous infusion of remifentanil. In their study, when remifentanil infusion was stopped at 60-minutes post-AIH, phrenic nerve activity increased above baseline levels, which was interpreted as pLTF. In their study, however, continuous remifentanil infusion suppressed baseline phrenic nerve activity, and time control studies (without AIH) were not performed with remifentanil infusion. Therefore, the rebound increase in phrenic nerve activity following cessation of remifentanil infusion may have been due to removal of continuous mu opioid receptor-induced respiratory depression versus activation of endogenous mechanisms of plasticity (i.e. pLTF). Because of these differences in drug/protocol, it is difficult to compare our studies. Further studies concerning the impact of continuous remifentanil infusion on mAIH-induced pLTF is worthy of investigation.

Pioneering work demonstrate that morphine activates the innate immune system and activates glia (Watkins et al., 2005). Others confirmed this finding, implicating inflammation as a potential mechanism for at least some of the opioid side effects (e.g. loss of analgesia, tolerance) (DeLeo et al., 2004; Johnston et al., 2004; Shavit et al., 2005; Watkins et al., 2005; Hutchinson et al., 2011; Due et al., 2012). Morphine activates TLR4, increasing the expression of pro-inflammatory cytokines that can secondarily oppose opioid-induced analgesia (Hutchinson et al., 2008a; Watkins et al., 2009; Ellis et al., 2016). TLR4 antagonists prolong morphine analgesia, suggesting that morphine efficacy may be partially hindered by TLR4-mediated inflammatory signaling (Hutchinson et al., 2010). Since many opioids including morphine, oxycodone, buprenorphine and fentanyl activate TLR4 signaling (Wang et al., 2012), opioid receptor inactive TLR4 inhibitors such as (+)-naloxone may block certain side effects of opioids (Lewis et al., 2010; Lewis et al., 2012).

Although it is not fully understood how morphine activates TLR4 signaling, it may result indirectly from morphine metabolites. Morphine is metabolized primarily into morphine-3-glucuronide (M3G) and morphine-6-glucuronide (M6G) (Coughtrie et al., 1989; Zelcer et al., 2005). M6G is a mu opioid receptor agonist that contributes to morphine analgesic effects; in contrast, M3G has extremely low affinity for opioid receptors (van Dorp et al., 2006), but is a potent activator of TLR4 (Lewis et al., 2010; Due et al., 2012). Since we suggest that morphine suppresses pLTF through TLR4 signaling, anti-inflammatory drugs targeting TLR4 or p38 MAP kinase may prevent at least some unwanted side effects, and enable mechanisms of respiratory plasticity such as pLTF.

In the present study, we explored whether acute morphine delivery blocks mAIH-induced spinal respiratory (phrenic) motor plasticity via mechanisms that require TLR4 signaling. Moderate AIH-induced phrenic motor plasticity represents an important mechanism to ensure adequate breathing and/or breathing stability (Mahamed & Mitchell, 2007, 2008); 2) has major translational relevance since repetitive AIH is emerging as a novel treatment to restore breathing ability and non-respiratory motor behaviors (e.g. walking) in neuromuscular disorders that compromise movement such as spinal cord injury, ALS and MS (Gonzalez-Rothi et al., 2015; Fuller & Mitchell, 2017); and 3) highly sensitive to inflammation (Huxtable et al., 2013; Agosto-Marlin et al., 2018). Our demonstration that morphine blocks pLTF via a TLR4-dependent mechanism is of importance since it advances our understanding of mechanisms whereby opioids impair control of breathing and further suggests that co-administration of opioids to control pain may thwart the therapeutic efficacy of repetitive AIH since many clinical populations that may benefit from therapeutic AIH often suffer from chronic pain (Finnerup & Baastrup, 2012; Woller & Hook, 2013).

Although the location of the relevant TLR4 receptors was not investigated in this study, TLR4 receptors are expressed in the periphery and at multiple places in the central nervous system, including brainstem and spinal cord respiratory circuits. Since pLTF is induced by serotonin receptor activation on/near phrenic motor neurons within the spinal cord (Tadjalli & Mitchell, 2019), and the resulting plasticity appears to reside within phrenic motor neurons per se (Devinney et al., 2015; Dale et al., 2017), the relevant TLR4 receptors are likely expressed within the region of the phrenic motor nucleus located in the ventral cervical spinal cord. Two lines of evidence support this idea since activation of TLR4 receptors via systemic LPS: 1) undermines pLTF via cervical spinal cytokine signaling (Hocker & Huxtable, 2018); and 2) impairs pLTF via cervical spinal serine-threonine protein phosphatase activity (Tadjalli et al., 2021). Although the hypothesis that systemic morphine impairs pLTF via cervical spinal inflammatory cascades remains to be tested directly, our finding that morphine activates p38 MAPK within cervical spinal microglia is consistent with this hypothesis.

Downstream products of the innate immune response mediated by TLR4 signaling following morphine administration include transcription factors and cytokines. For example, morphine can induce activation of NF-κB and production of proinflammatory cytokines such as IL-1β and TNF-α (Wang et al., 2012; Wang et al., 2016). Furthermore, it has been demonstrated that morphine is able to induce these biological effects via LPS-like interactions with TLR4’s co-receptor MD-2 (Wang et al., 2012). Since cytokines such as IL-1β are known to negatively regulate mAIH-induced pLTF (Hocker & Huxtable, 2018), it is possible that deficits in pLTF following morphine could be secondary to IL-1β pro-inflammatory signaling. Although this concept was not tested in the present study, further investigations are warranted to answer this important question.

We show that a single analgesic dose of morphine results in a rapid biochemical change at the molecular level within the cervical spinal cord. Specifically, morphine exclusively enhanced p38 MAPK phosphorylation/activation within microglia of the ventral cervical spinal segments encompassing the phrenic motor nucleus. Morphine-induced p38 MAPK phosphorylation requires TLR4 signaling since this effect was prevented by pre-treatment with (+)-naloxone. Thus, even a single morphine dose can rapidly influence cellular networks capable of modulating the expression of phrenic motor plasticity.

The p38 MAPK family is well-known for its role in both promoting and responding to inflammation (Correa & Eales, 2012). p38 MAPK is a key regulator of inflammatory cascades, that include cytokines, chemokines, NF-κB, cyclooxygenase 2 and other proteins, often perpetuating a self-sustaining cycle (Kaminska, 2005; Cuadrado & Nebreda, 2010; Kyriakis & Avruch, 2012). p38 MAPK signaling is also an important regulator of synaptic plasticity (Thomas & Huganir, 2004; Moult et al., 2008; Falcicchia et al., 2020). We provide correlative evidence that morphine upregulates p38 MAPK phosphorylation/activation in spinal microglia, potentially orchestrating relevant cellular cascades that impair mAIH-induced pLTF. This possibility aligns with a prior report demonstrating that cervical spinal p38 MAPK activity undermines mAIH-induced pLTF in the context of neuro-inflammation elicited by one day of severe intermittent hypoxia—a physiologically relevant stimulus mimicking aspects of sleep apnea (Huxtable et al., 2015).

A unique aspect of our findings was that acute morphine enhanced p38 MAPK phosphorylation/activation in spinal CD11b-positive cells: the morphology of phospho-p38 immunoreactive CD11b-positive cells showed small cell bodies with ramified processes, indicative of microglia. To our knowledge, this is the first demonstration of morphine-induced enhancement of microglial p38 MAPK activity in a rapid time scale. Neurons had close contacts with phospho-p38 immunoreactive cells, but co-localization was not observed between phospho-p38 MAPK and NeuN. Morphine-induced p38 MAPK activation is coincident with impaired mAIH-induced pLTF, and reversal of p38 MAPK activation by (+)-naloxone pre-treatment is paralleled by pLTF restoration. Thus, selective microglial p38 MAPK activation via TLR4-dependent mechanisms may be a critical link in morphine-induced deficits in phrenic motor plasticity. This hypothesis awaits direct experimental verification. The proposed mechanisms by which morphine inhibits moderate acute intermittent hypoxia-induced phrenic long-term facilitation is illustrated in figure 10.

Chronic morphine delivery is known to enhance microglial (but not neuronal or astrocytic) p38 MAPK activation coincident with the formation of tolerance (Cui et al., 2006; Wang et al., 2009; Horvath et al., 2010). Spinal inhibition of p38 MAPK activity (Cui et al., 2006; Chen et al., 2008) or spinal intrathecal administration of minocycline, an inhibitor of microglia activation (Cui et al., 2008), attenuated morphine-induced tolerance. These findings indicate that activation of microglial MAPKs, particularly the p38 family, is critical to the development of some of the negative side effects of morphine analgesia (e.g. tolerance). While spinal MAPK involvement in tolerance and hyperalgesia following chronic morphine delivery is well documented, studies evaluating the short-term effects of acute morphine administration on p38 MAPK activity have been lacking. To the best of our knowledge, our study is the first to address this knowledge gap, illustrating that a single systemic administration of morphine (within a few hours) also enhances p38 MAPK activity exclusively in microglia. Understandings the early-onset mechanisms by which opioids influence neural network activity might facilitate development of therapeutic approaches to combat unfavorable opioid side-effects. This has been an important and enduring notion in pain management, not least because the soaring increase in opioid prescriptions for the treatment of acute pain is thought to be an important predictor/driver of persistent opioid use and addiction (Calcaterra et al., 2016; Sun et al., 2016; Brummett et al., 2017; Stark et al., 2017).

Diminishing reflex responses to changes in O₂ and CO₂ levels undermines the ability of the ventilatory control system to compensate for pathologies that compromise adequate lung ventilation. Opioids are widely known to suppress hypoxic and hypercapnic ventilatory responses in humans and animal models (Kirby & McQueen, 1986; Bailey et al., 2000; Dahan et al., 2001; Romberg et al., 2003; Modalen et al., 2006; Zhang et al., 2007, 2009; Zhang et al., 2011; May et al., 2013). Although these effects are assumed to arise from mu opioid receptor activation, our data suggest additional effects may arise from TLR4 activation. The morphine dose and timing of experiments in this study reflect the TLR4 versus mu opioid receptor actions due to the short half-life of morphine. Although differences in the short-term hypoxic phrenic response (Fig. 2B) were minimal among groups, responses to maximal chemoreflex activation with hypoxic hypercapnia were blunted in morphine-treated rats (Fig. 8). This finding suggests that morphine diminished the capacity (versus sensitivity) of phrenic (and ventilatory) responses to maximum chemoreflex activation (10% O₂/7% inspired CO₂). Since these data were collected at the end of experiments, when mu opioid receptors were no longer activated, and the effects were prevented by (+)-naloxone pretreatment, morphine appears to impair maximal phrenic responses to chemoreflex activation via TLR4 activation.

To date few studies have reported effects of inflammation on respiratory chemoreflexes (for review refer to (Huxtable et al., 2011)). Majority of the reports examined the impact of systemic inflammation on hypoxic ventilatory responses, with only one study examining the impact of systemic LPS (TLR4 agonist) on maximal chemoreflex stimulation (10.5% inspired O2 with 7% inspired CO2 / balance N2) of breathing. As demonstrated by Huxtable and colleagues (2011), there is evidence to suggest that LPS-induced inflammation reduces ventilation during maximal chemoreceptor stimulation. The precise mechanisms by which inflammation reduces max chemoreceptor stimulation of breathing was not investigated. Our findings in the present study conceptually agree with the report that max chemoreflex responses can be blunted in conditions where inflammatory processes are present. We extend this notion by demonstrating that max chemoreflex stimulation of respiratory activity (as measured by phrenic nerve motor output) is restored in rats that received the TLR4 inhibitor before morphine. We are unsure exactly which central sites are being affected by inflammatory processes; their identification was beyond the scope of this study. Overall, our data suggests that downregulation of chemoreflexes by immune driven inflammatory processes secondary to opioid use should be taken into consideration in patients with unstable ventilatory chemoreflex control.