Abstract
Inhibitory 5-HT1a receptors are located on serotonin (5-HT) neurons (autoreceptors) as well as neurons of the respiratory network (heteroreceptors). Thus, effects on breathing of 5-HT1a agonists, such as (R)-(+)-8-hydroxy-2-(di-N-propylamino) tetralin (8-OH-DPAT), could either be due to decreased firing of 5-HT neurons or direct effects on the respiratory network. Mice in which the transcription factor LMX1B is genetically deleted selectively in Pet1-1-expressing cells (Lmx1bf/f/p) essentially have complete absence of central 5-HT neurons, providing a unique opportunity to separate the effect of activation of downstream 5-HT1a heteroreceptors from that of autoreceptors. We used rhythmically active medullary slices from wild-type (WT) and Lmx1bf/f/p neonatal mice to differentiate autoreceptor versus heteroreceptor effects of 8-OH-DPAT on hypoglossal nerve respiratory output. 8-OH-DPAT transiently increased respiratory burst frequency in Lmx1bf/f/p preparations, but not in WT slices. This excitation was abolished when synaptic inhibition was blocked by GABAergic/glycinergic receptor antagonists. Conversely, after 10 min of application, frequency in Lmx1bf/f/p slices was not different from baseline, whereas it was significantly depressed in WT slices. In WT mice in vivo, subcutaneous injection of 8-OH-DPAT produced similar biphasic respiratory effects as in Lmx1bf/f/p mice. We conclude that 5-HT1a receptor agonists have two competing effects: rapid stimulation of breathing due to excitation of the respiratory network, and delayed inhibition of breathing due to autoreceptor inhibition of 5-HT neurons. The former effect is presumably due to inhibition of inhibitory interneurons embedded in the respiratory network.
Introduction
5-HT1a receptors are located throughout the respiratory network, on both 5-HT source neurons of the raphe and on downstream neurons of target nuclei. These receptors are coupled to inhibitory G-proteins. Thus ligand binding to somatodendritic 5-HT1a autoreceptors inhibits firing of 5-HT neurons, reducing 5-HT release and therefore decreasing activation of other receptors by endogenous 5-HT (McCall and Clement, 1989; Sharp et al., 1989; Veasey et al., 1995). 5-HT1a heteroreceptors are found in several respiratory-related nuclei including the hypoglossal motor nucleus, the retrotrapezoid nucleus, and the pre-Bötzinger complex (preBötC; Okabe et al., 1997; Liu and Wong-Riley, 2010), and there has been much interest in pharmacological targeting of these receptors to reverse life-threatening instabilities of breathing in clinical conditions (Wilken et al., 1997; El-Khatib et al., 2003; Richter et al., 2003; Manzke et al., 2009). Determining the effect of these downstream 5-HT1a receptors on the output of the respiratory network, however, is challenging when studied in an intact network because it is not always clear whether the target neurons cause excitation or inhibition of respiratory output. Additionally, the agonists and antagonists used to query the role of 5-HT1a receptor function can have nonspecific, off-target effects.
Much of our knowledge describing the role of 5-HT1a receptors in control of breathing has been based on studies using 8-OH-DPAT, which binds to both 5-HT1a autoreceptors and downstream heteroreceptors, making it difficult to differentiate between the role of the two receptor populations. Additionally, 8-OH-DPAT has moderate affinity for 5-HT7 receptors (Bard et al., 1993; Eriksson et al., 2008). Low levels of mRNA expression for 5-HT7 receptors are found in the preBötC (Richter et al., 2003) and immunoreactivity has been reported in raphe serotonergic cells (Muneoka and Takigawa, 2003). Our knowledge of the role of 5-HT7 receptors in breathing is limited, although they appear to be involved in restoring phrenic nerve activity following fentanyl-induced depression of respiratory discharge (Richter et al., 2003). The potential facilitation of respiration by activation of 5-HT7 receptors could confound interpretation of the effects of 8-OH-DPAT on both 5-HT1a autoreceptors and heteroreceptors.
Here we report a unique approach to distinguish between 5-HT1a receptor populations in which we use mice lacking central 5-HT neurons (Lmx1bf/f/p mice). Recent investigation into the importance of the 5-HT system in ventilatory chemosensation has exploited the use of this transgenic mouse (Ding et al., 2003; Zhao et al., 2006; Hodges and Richerson, 2008b; Hodges et al., 2008; Buchanan and Richerson, 2010). An experimental benefit of the loss of central 5-HT neurons in this mouse model is the coincident loss of 5-HT1a autoreceptors, but not heteroreceptors. As such, application of 8-OH-DPAT in Lmx1bf/f/p mice will only affect heteroreceptor populations, providing a tool to identify the role of these receptors by comparing the results with those obtained from WT mice in which 5-HT1a autoreceptors and downstream heteroreceptors are both present. Additionally, we examined the effect of coadministered 8-OH-DPAT with a 5-HT7 receptor antagonist, removing potential confounding effects of 5-HT7 receptors in interpretation of the role of 5-HT1a receptors in breathing.
Materials and Methods
5-HT neuron-deficient mouse model.
LMX1b is a transcription factor that is required for 5-HT neuron differentiation during embryogenesis (Ding et al., 2003). The procedure using Cre-Lox site-specific recombination targeting this transcription factor to produce mice lacking 5-HT neurons (Lmx1bf/f/p mice) has been previously described by Zhao et al. (2006). Briefly, females homozygous for the Lmx1b sequence flanked by LoxP (floxed Lmx1b;Lmx1bf/f) are mated with males homozygous for floxed Lmx1b and hemizygous for ePet1-Cre (Lmx1bflox/flox,ePet−Cre/+, also referred to as Lmx1bf/f/p). This strategy results in a 1:1 ratio of Lmx1bf/f (referred to here as WT; containing a normal number of central 5-HT neurons that are phenotypically normal) and Lmx1bf/f/p pups (expressing loxP sites flanking both Lmx1b alleles, targeted by Cre recombinase produced specifically in Pet1-expressing cells, and missing >99% of central 5-HT neurons). Genotyping was performed on tail tissue samples obtained from each mouse and using procedures described by Zhao et al. (2006). All experiments were performed on neonatal mice, or on slice preparations derived from neonatal mice [postnatal day (P)0–P4], to allow for direct comparison between the various methods used. Note that use of the in vitro medullary slice preparation is typically used at this age due to instability of the respiratory-related activity and lower viability of slice preparations from older animals.
Tryptophan hydroxylase immunohistochemistry.
Brainstems were fixed overnight in 4% paraformaldehyde in 0.1 m phosphate buffer (containing the following in g/L: NaH2PO4 3.18, Na2HPO4 10.94, pH 7.4), allowed to sink in cryoprotectant [30% sucrose in PBS containing the following in g/L: NaCl 87, NaH2PO4 (monobasic) 3.86, Na2HPO4 (dibasic) 10.2] for 24 h, frozen, and cut into 30-μm-thick sections (WT n = 6, 3.2 ± 0.3-d-old; Lmx1bf/f/p n = 7, 3.0 ± 0.3-d-old). Consecutive coronal sections were mounted on microscope slides and stained for tryptophan hydroxylase (TpOH) as follows: sections were rinsed in PBS, incubated for 30 min in 0.15% H2O2 in methanol, for 5 min in PBS with 0.1% Triton X-100 (PBST), for 60 min in PBST with 5% horse serum, and overnight at room temperature in PBST with 2.5% horse serum and a primary antibody against TpOH (Sigma-Aldrich, T0678; 1:1000 dilution). Sections were rinsed three times in PBST, incubated for 30 min in biotinylated anti-mouse IgG antibody (1:250 dilution) in PBST with 2.5% horse serum, and rinsed three times in PBST. Sections were then processed using the avidin-biotin-peroxidase method with the Vectastain ABC kit (Vector Laboratories), and washed three times in PBS. Serotonin neurons were visualized using NovaRED (Vector Laboratories) as the chromogen, and washed three times in water. Sections were dehydrated with an ethanol series, cleared with xylenes, and coverslipped using Permount (Fisher Scientific). Stained sections were observed with brightfield illumination.
The number of TpOH-positive neurons was counted in every available and intact section between −1.8 to 2.8 mm rostral to obex. Within each section, TpOH-positive neuron counts were made from two regions: a midline region containing nucleus raphe obscurus and pallidus (identified by a 0.5-mm-wide rectangle centered on the midline; Hodges et al., 2011) and a bilateral ventrolateral medulla region that included any neurons not located within the midline rectangle. For each animal, the total number of cells counted from all sections was divided by the number of sections examined, resulting in an average number of cells/section for each individual animal. These were then averaged for WT and Lmx1bf/f/p. Differences between the average number of positive cells per section in WT and Lmx1bf/f/p were determined using a Mann–Whitney rank sum test with p < 0.05 as significant.
5-HT1A receptor autoradiography.
P3 mice were decapitated and the brains were rapidly removed and frozen. Brains were sectioned 20 μm thick using a cryostat and slices were thaw mounted on Superfrost Plus slides (Fisher Scientific). Slides were maintained at −80°C until processing. For 5-HT1a receptor autoradiography, slides were preincubated for 30 min at room temperature in 0.17 m Tris-HCl buffer, pH 7.6, containing 4 mm CaCl2 and 0.01% ascorbic acid. For total binding, slides were incubated for 1 h at room temperature in the same buffer with the addition of 4 nm 3H-8-OH-DPAT (New England Nuclear) and 10 μm pargyline. Nonspecific binding was determined by processing slides similarly except 10 μm 5-HT was added as a displacer for 3H-8-OH-DPAT. Sections were subsequently rinsed, dried and exposed to BAS-TR2025 phosphoimaging plates (Fujifilm Medical Systems) for 3 weeks with a set of 3H-standards (GE Healthcare). Quantitative densitometry was completed using NIH ImageJ software.
Eight WT (5 females and 3 males) and 8 Lmx1bf/f/p (2 females and 6 males) were analyzed. Every fourth section was processed for specific binding and serial sections were processed for nonspecific binding. For each area analyzed, specific binding was measured on three sections per mouse; nonspecific binding was measured on the three serial sections. Three areas in the medulla were analyzed: caudal raphe (CR) consisting of an area 0.5 mm wide centered on the midline, and lateral to that the caudal ventrolateral medulla (CVL). The spinal nucleus of V (Sp5) was also sampled. Mean specific binding was determined per mouse and these were averaged together to generate a group mean. Nonspecific binding values were subtracted from those of total binding to obtain specific binding. For each area, differences between WT and Lmx1bf/f/p were determined using a Student's t test with p < 0.05 as significant.
In vivo plethysmography.
Whole-body plethysmography was used to measure ventilation in P2 mice (described previously by Hodges et al., 2009). In neonates, this method allows for unrestrained breathing measurements, resulting in less stress and accurate determination of breathing frequency and relative tidal volume; however, it does not allow for accurate calculation of absolute tidal volume (which is thus expressed as relative to baseline; for review, see Cummings et al., 2004). Pups were separated from their mother and placed in a 9 ml Plexiglas chamber equipped with inflow and outflow gas ports, a pressure transducer, and a temperature sensor. Ambient temperature was maintained at 30°C using a temperature controller (TCAT-2AC, Physitemp Instruments) and an incandescent, infrared lamp. Pressure transducer data were digitized at 100 Hz using an analog-to-digital converter (PCI-6224, National Instruments) and recorded using MATLAB data acquisition software (MathWorks) with a custom-written programs used to monitor data in real-time and analyze off-line.
Studies consisted of a stop flow protocol, where room air flowed through the chamber for 20 s and was stopped for 60 s (repeated five times). This method allows for low noise recording of small amplitude breaths generated by neonatal mice. Baseline ventilatory measurements were made before (baseline), and 30, 60, and 120 min following subcutaneous (intrascapular) injection of 8-OH-DPAT (Sigma-Aldrich; 100 μg/kg body weight, dissolved in 150 mm NaCl). Pups were in the chamber for ∼7 min at a time, and returned to their mother between recordings. Data are presented as mean ± SEM. Differences were determined using a two-way repeated-measures (RM) ANOVA with Bonferroni post hoc analysis for multiple comparisons with statistical significance attributed to p < 0.05.
In vitro medullary slice preparations.
Experiments were performed on brainstem slices prepared from neonatal mice (P0–P4). Mice were decapitated, and the brainstem and spinal cord removed under a flow of chilled dissection solution composed of the following (in mm): 124 NaCl, 25 NaHCO3, 3 KCl, 1.5 CaCl2, 1.0 MgSO4, 0.5 NaH2PO4, 30 d-glucose, and equilibrated with 95% O2 and 5% CO2, pH ∼7.4. The brainstem was pinned to a wax block, and transverse slices of the medulla (500–600 μm thick) were prepared using a vibratome (Smith et al., 1991). Slices from the rostral medulla containing the preBötC, the hypoglossal motor nucleus, the most rostral hypoglossal nerve rootlets and 5-HT neurons of the medullary raphe (Ptak et al., 2009) were placed in a recording chamber with a volume of 0.7 ml and superfused with dissection solution at a rate of 0.7 ml/min via a syringe pump (Harvard Apparatus). The temperature of aCSF within the chamber was increased to 29°C using a temperature controller (Warner Instruments; TC-324B). After stabilization of the slice, the potassium concentration in recording solutions was elevated to 8 mm to enhance rhythmic activity in the slice, as per standard procedures (Smith et al., 1991; Ptak et al., 2009).
Inspiratory-related motor discharge of the hypoglossal rootlets was recorded continuously using glass suction electrodes. Signals were amplified × 10,000 and bandpass filtered (0.3–1 kHz) using a Grass LP511 AC amplifier (Grass Technologies). A Hum Bug noise eliminator (Quest Scientific) was used as a digital filter of 60 Hz noise. Data were digitized at 10 kHz, stored on a computer hard drive and analyzed off-line using MATLAB software. Analysis of hypoglossal root discharge included full wave rectification, integration, and calculation of the moving average with a bin size of 50 ms.
After 30 min of baseline recording, the superfusate was switched to one containing one of the drugs listed below and applied continuously to the slice for 10 min, which was sufficient to achieve steady-state responses in the recorded hypoglossal nerve activity. The last 2 min under each condition (control and drug application) was used to measure values of each parameter to determine the effect of drug relative to baseline (“steady-state” response). We also analyzed a “peak” response in all slices exposed to 8-OH-DPAT, defined as the maximum hypoglossal burst frequency observed for any 1 min time period during the 10 min exposure. Only one slice was taken and recorded from each animal and a 2 min average value for each condition and/or parameter was used from each slice for quantitative analyses. Data are presented as mean ± SEM. Differences were determined using a two-way RM ANOVA with Bonferroni post hoc analysis for multiple comparisons or a Student's paired t tests and statistical significance was attributed to p < 0.05. Each slice was exposed to a single drug, with the exception of the coapplication of 8-OH-DPAT and SB 269970, and coapplication of 8-OH-DPAT and a gabazine/CGP 55845/strychnine mixture to block endogenous GABAA, GABAB, and glycine receptors, respectively.
Results
5-HT neuron and 5-HT1a receptor distributions in Lmx1bf/f/p mice
As reported by Zhao et al. (2006) and Ding et al. (2003), we found near-complete absence of 5-HT neurons in neonatal Lmx1bf/f/p brainstems. Immunohistochemical staining identified a very small number of TpOH-positive neurons. An average of 0.18 ± 0.04 TpOH-positive neurons per section was counted in the midline raphe region of P2 Lmx1bf/f/p pup medullary tissue, compared with 28.1 ± 3 in WT (Fig. 1A,B), representing a >99% depletion in Lmx1bf/f/p mice (p < 0.001). A similar reduction was observed in the ventrolateral medullary region (p < 0.001). We also evaluated 5-HT1a receptor binding sites in neonatal Lmx1bf/f/p tissue using quantitative tissue autoradiography. In nonraphe areas, such as the Sp5 and the CVL, WT and Lmx1bf/f/p mice had equivalent levels of 5-HT1a receptor binding. However, there was selective loss of 5-HT1a receptor binding along the midline in the area occupied by the caudal raphe nuclei in Lmx1bf/f/p compared with WT mice (Fig. 1C,D). Note that the nonraphe areas (CVL and Sp5) are where downstream heteroreceptors are located and the raphe area (CR) is where we expect somatodendritic autoreceptors and possibly downstream 5-HT1a receptors.
5-HT neurons and 5-HT1a receptor binding are specifically lost within the caudal raphe of Lmx1bf/f/p mice. A, Transverse sections of the caudal medulla from P2 WT and Lmx1bf/f/p mice immunostained for TpOH-positive neurons. B, Neuron counts (average no. TpOH-positive neurons/section; n = 6 WT, n = 7 Lmx1bf/f/p) revealed a significant depletion of 5-HT neurons in Lmx1bf/f/p mice in both the midline raphe and ventrolateral medulla regions. Note that sections were 30 μm thick. C, Autoradiographs of transverse sections through the medulla from P3 WT and Lmx1bf/f/p mice illustrating distribution of 5-HT1a receptor binding. In WT mice, binding was seen in the midline (arrow) in the area of the CR that was not apparent in Lmx1bf/f/p mice. D, Quantification revealed Lmx1bf/f/p mice had significantly less 5-HT1a receptor binding than WT mice in the CR, but there was no difference in receptor binding in the CVL or Sp5 (n = 8 WT, n = 8 Lmx1bf/f/p). *Significant difference between genotypes; p < 0.05, t test. Data are mean ± SEM.
Respiratory discharge in the rhythmogenic medullary slice
Hypoglossal motor output was characterized in Lmx1bf/f/p slices. Because 5-HT is an excitatory modulator for breathing, the loss of excitatory drive might cause Lmx1bf/f/p slice preparations to be incapable of bursting (particularly because these slice preparations are devoid of many normal inputs and modulators). Hypoglossal nerve bursting was present, although it was slower in Lmx1bf/f/p (7.9 ± 0.42 bursts/min; n = 46) compared with WT preparations (10.499 ± 0.52 bursts/min; n = 51; Fig. 2; p < 0.001). Additionally, the burst pattern in Lmx1bf/f/p slices was more irregular than WT slices as assessed by an elevated (∼ 42%) coefficient of variation of the interburst interval (CVIBI; Mann–Whitney rank sum test; p < 0.001). Burst duration was equivalent between genotypes (0.65 ± 0.03 and 0.67 ± 0.03 s).
Hypoglossal nerve bursting was slow and irregular in Lmx1bf/f/p mouse pup rhythmogenic slice preparations. A, Traces of raw hypoglossal (XII) and integrated XII (∫XII) nerve activity in WT and Lmx1bf/f/p slices under control conditions. B, XII burst frequency was lower in Lmx1bf/f/p preparations. Bursting was also more irregular in Lmx1bf/f/p preparations than WT, as quantified by the coefficient of variation of the interburst interval (CVIBI). Burst duration was the same in both genotypes (n = 51 WT, n = 46 Lmx1bf/f/p).*Significant difference between genotypes; p < 0.05, t test. Data are mean ± SEM.
In brain slices from WT mice, hypoglossal nerve output is dependent on continuous activation of 5-HT2A receptors by endogenous 5-HT (Peña and Ramirez, 2002; Ptak et al., 2009). To ensure that endogenous 5-HT was indeed absent in Lmx1bf/f/p slices, we bath-applied MDL 11939 (MDL; α-Phenyl-1-(2-phenylethyl)-4-piperidine methanol; Tocris Bioscience), a highly selective 5-HT2a receptor antagonist, to the preparation (50 μm, dissolved in 0.1% DMSO in aCSF; n = 7). There were no significant effects on burst frequency (Fig. 3; p = 0.234), although a mild reduction in burst amplitude was observed. This is in contrast to effects of MDL in WT slices (n = 6), which resulted in a marked depression of burst activity (∼50%; p = 0.004), as previously reported by Ptak et al. (2009). We thus conclude that these slices lack endogenous 5-HT and activation of 5-HT2a receptors.
There was no evidence for endogenous 5-HT release in Lmx1bf/f/p slices, or compensation by an increase in adrenergic tone. A, MDL 11939 (selective 5-HT2a receptor antagonist; 50 μm) significantly decreased hypoglossal burst frequency in WT rhythmogenic slices (p = 0.004, n = 6 WT) but not in Lmx1bf/f/p slices (p = 0.234, n = 7 Lmx1bf/f/p; two-way RM ANOVA followed by Bonferroni test for multiple comparisons). No differences were observed between genotypes in the responses to prazosin hydrochloride (α1 receptor antagonist; 50 μm; n = 6 WT, n = 6 Lmx1bf/f/p) or yohimbine hydrochloride (α2 receptor antagonist; 50 μm; n = 6 WT, n = 6 Lmx1bf/f/p; two-way RM ANOVA followed by Tukey test for multiple comparisons; p < 0.001). B, C, Time course of burst frequency (B) and burst amplitude (C; relative to baseline) during a 10 min bath application of MDL11939 (50 μm) in WT and Lmx1bf/f/p preparations (Friedman RM ANOVA on ranks followed by Dunn's methods for multiple comparisons; p < 0.001). *Significant difference from baseline. Data are mean ± SEM.
Because the Lmx1bf/f/p slice, devoid of 5-HT, retains some (albeit altered) rhythmogenic activity, whereas respiratory output is blocked by 5-HT2 receptor antagonists in WT mice (Peña and Ramirez, 2002; Ptak et al., 2009; Viemari et al., 2011), we hypothesized that there may be a compensatory mechanism or altered network that provides an alternative source of excitation. We thus examined whether there were differences in noradrenergic modulation (Viemari et al., 2011) by testing responses to adrenergic antagonists in WT and Lmx1bf/f/p mice (Fig. 3A). None was detected. Neither genotype showed any significant changes in hypoglossal burst frequency in response to the α1-adrenergic receptor antagonist prazosin hydrochloride (Tocris Bioscience; 50 μm, dissolved in 0.5% DMSO in aCSF; n = 6 WT, n = 6 Lmx1bf/f/p). Yohimbine is an antagonist at α2-adrenergic receptors with effects at a variety of 5-HT and other receptors including partial agonist effects at 5-HT1a receptors. Yohimbine hydrochloride (Tocris Bioscience; 50 μm, dissolved in aCSF) depressed frequency in both WT and Lmx1bf/f/p mice by a similar degree (to 0.65 ± 0.08 and 0.61 ± 0.07 relative to baseline, respectively; n = 6 WT, n = 6 Lmx1bf/f/p). The compensatory mechanism to maintain excitation remains unknown.
Effects of 8-OH-DPAT in vitro and in vivo
We examined the effect of 8-OH-DPAT in vitro and in vivo in both WT (containing both 5-HT1a autoreceptors and downstream heteroreceptors) and Lmx1bf/f/p neonates (containing only downstream 5-HT1a heteroreceptors). Because Lmx1bf/f/p mice lack raphe 5-HT neurons and 5-HT1a autoreceptors, the effect of 8-OH-DPAT in this mouse is due to activation of downstream 5-HT1a receptors alone.
We examined the effect of a range of doses of 8-OH-DPAT (Sigma-Aldrich; 10 nm–100 μm, dissolved in aCSF) on hypoglossal nerve bursting activity in slices (Fig. 4). We determined the peak response (the maximum burst frequency observed during the 10 min exposure) and the steady-state response (average of the final 2 min of the 10 min exposure). In slices from WT mice (n = 3, 4, 4, 6, 6, 6, 5 slices for each successive dose of 8-OH-DPAT) burst frequency began to decrease within the first few minutes (Fig. 5A), stabilizing below baseline at the end of a 10 min exposure to higher doses (3–100 μm; Fig. 5B). Lower doses did not depress burst frequency. Unlike WT slices, Lmx1bf/f/p slices (n = 4, 5, 4, 5, 4, 9, 5 for each successive dose in Lmx1bf/f/p), did not exhibit an attenuation of frequency at any dose. In fact, they responded to 8-OH-DPAT in an opposite manner: with a transient increase in burst frequency (peak response; Fig. 5C). This peak response was significant at higher doses (1–100 μm) and was not observed in WT preparations, with the exception of one dose (0.3 μm). 8-OH-DPAT had very little effect on burst amplitude at any dose in either genotype (Fig. 5D). Only at 100 μm was burst amplitude significantly lower than baseline in WT preparations (0.57 ± 0.09 relative to baseline; p = 0.022).
Traces of raw hypoglossal (XII) nerve activity in WT and Lmx1bf/f/p slices, before (baseline) and during (peak and steady-state) bath application of 8-OH-DPAT (100 μm). Represented are examples of a transient peak response (defined as the maximum burst frequency during the 10 min drug exposure) and a steady-state response (attained at the end of the 10 min drug application).
8-OH-DPAT differentially affected hypoglossal nerve discharge in WT and Lmx1bf/f/p rhythmogenic slices. A, Time course of burst frequency (relative to baseline) during a 10 min bath application of 8-OH-DPAT (100 μm) in WT and Lmx1bf/f/p slices. B,C, 8-OH-DPAT dose–response curves (10 nm–100 μm) showing changes in burst frequency relative to baseline at steady-state (B) and during the peak response (C; n = 3, 4, 4, 6, 6, 6, 5 slices for each successive dose of 8-OH-DPAT in WT; n = 4, 5, 4, 5, 4, 9, 5 for each successive dose in Lmx1bf/f/p). D, Steady-state responses of burst amplitude (relative to baseline) at different doses of 8-OH-DPAT. *Significant difference from baseline; paired t test, p < 0.05). Data are mean ± SEM.
In vivo results were similar to those obtained in vitro (Fig. 6). Baseline ventilation, as assessed by whole-body plethysmography, was lower in Lmx1bf/f/p P2 pups compared with WT pups (n = 7 Lmx1bf/f/p; n = 7 WT). Breathing was slower in Lmx1bf/f/p pups (Fig. 7A; average breathing frequency of 2.12 ± 0.17 Hz in Lmx1bf/f/p compared with 3.49 ± 0.15 Hz in WT pups). Breaths were approximately half the size of those in WT mice (Fig. 7B; 56.7 ± 5.7% of WT; normalized to weight). Additionally, the loss of central 5-HT resulted in unstable breathing (Fig. 7D,E), with long respiratory pauses (>1 s) and a large percentage of time spent apneic (∼17% of the time). These observations are similar to those reported by Hodges et al. (2009) in neonatal Lmx1bf/f/p mice.
Ventilatory traces from whole-body plethysmographic recordings in WT and Lmx1bf/f/p P2 mouse pups at baseline, 30 min, and 120 min after subcutaneous injection of 100 μg/kg 8-OH-DPAT.
Breathing was augmented in neonatal mouse pups lacking central 5-HT neurons after injection of 8-OH-DPAT. A, B, Breathing frequency (A) and breath amplitude (B; index of tidal volume) before and after injection of 8-OH-DPAT in WT and Lmx1bf/f/p pups (n = 7 WT, n = 7 Lmx1bf/f/p). C, Minute ventilation (VE) expressed relative to baseline. D, E, Percentage of time spent apneic (defined as a respiratory pause >1 s) (D) and frequency of apneas (E) before and after injection of 8-OH-DPAT. One-way RM ANOVA followed by Bonferroni post hoc for multiple comparisons, compared with baseline; p < 0.05. *Significant difference from baseline. Data are mean ± SEM.
Subcutaneous injection of 8-OH-DPAT altered breathing differently in the two genotypes (Figs. 6, 7). Although 8-OH-DPAT significantly augmented frequency in Lmx1bf/f/p pups 30 min postinjection (p < 0.001), there was no change in breathing frequency in WT pups until 2 h postinjection (Fig. 7A; p = 0.011). There was a twofold increase in minute ventilation (frequency × amplitude; Fig. 7B) in Lmx1bf/f/p mice (p < 0.001) and a nonsignificant ∼20% trend toward a reduction in WT mice (Fig. 7C). Stimulation of 5-HT1a receptors also normalized breathing in Lmx1bf/f/p pups by reducing the number of apneas (Fig. 7D; p < 0.001) and percentage time spent apneic (Fig. 7E; p < 0.001).
Coapplication of 8-OH-DPAT and a 5-HT7 antagonist
One caveat of using 8-OH-DPAT to examine effects of 5-HT1a receptors is that it is a partial agonist at 5-HT7 receptors, thus confounding interpretation. To isolate the effect of 5-HT1a receptors, we coapplied 8-OH-DPAT (3 μm) with the 5-HT7 antagonist SB 269970 hydrochloride (Tocris Bioscience; 20 μm, dissolved in aCSF) to rhythmogenic medullary slices (n = 6 WT; n = 9 Lmx1bf/f/p). SB 269970 alone did not alter hypoglossal bursting in either type of preparation (Fig. 8). After addition of 8-OH-DPAT to the aCSF containing SB 269970, we examined both peak and steady-state responses. As with 8-OH-DPAT alone, coapplication of 8-OH-DPAT and SB 269970 did not evoke an increase in hypoglossal burst frequency in WT slices. In contrast, coapplication resulted in a transient increase in Lmx1bf/f/p preparations (from 3.78 ± 0.59 bursts/min to 5.00 ± 0.53 bursts/min; p = 0.003). For both genotypes, the steady-state responses to the coapplication were similar to effects of 8-OH-DPAT alone; a decrease in burst frequency in WT (from 8.00 ± 1.13 bursts/min to 4.67 ± 0.53 bursts/min; p = 0.017) and no change in Lmx1bf/f/p preparations. We thus conclude that observations made in our experiments using 8-OH-DPAT were mediated by 5-HT1a receptor mechanisms and not confounded by action of this drug on 5-HT7 receptors.
8-OH-DPAT elicited a transient increase in frequency (peak response) in Lmx1bf/f/p preparations, despite concurrent blockade of 5-HT7 receptors. A, XII nerve and ∫XII nerve activity in an Lmx1bf/f/p slice during control conditions, bath application of SB 269970 (20 μm), and coapplication of SB 269970 and 8-OH-DPAT (3 μm). B, Hypoglossal burst frequency at baseline, during application of SB 269970, and during coapplication of SB 269970 and 8-OH-DPAT (n = 6 WT, n = 9 Lmx1bf/f/p). Shown are both the peak and steady-state responses to 8-OH-DPAT (attained at the end of 10 min of drug application; one-way RM ANOVA followed by Bonferroni post hoc for multiple comparisons; p < 0.05). *Significant difference. Data are mean ± SEM.
Coapplication of 8-OH-DPAT and glycinergic/GABAergic receptor antagonists
We hypothesized that a mechanism for the transient excitation caused by 8-OH-DPAT in Lmx1bf/f/p mice involves disinhibition of the respiratory network because downstream 5-HT1a receptors are predominantly located on inhibitory glycinergic respiratory interneurons including the preBötC inspiratory rhythm generator (Dutschmann et al., 2009; Manzke et al., 2009). Inhibition of these inhibitory neurons by 5-HT1a receptor-activation could cause network excitation. To test this hypothesis, experiments were performed with inhibitory receptor (glycinergic and GABAergic) antagonists in the Lmx1bf/f/p slice preparations to determine endogenous inhibitory control of the inspiratory bursting frequency, and to then test for occlusion of 8-OH-DPAT-induced excitation by the antagonists (Fig. 9). We first applied a mixture of gabazine (0.5 μm), CGP 55845 (1 μm) and strychnine (0.5 μm), respectively. In 8 of 11 preparations, these blockers significantly augmented inspiratory burst frequency (Fig. 9; average increase in frequency of 167%; one-way RM ANOVA, p < 0.001), indicating ongoing inhibitory control of the inspiratory rhythm. Subsequent coapplication of the inhibitory blocker mixture and 8-OH-DPAT (100 μm; n = 8 Lmx1bf/f/p) did not cause a significant transient further increase in peak or steady-state burst frequency, indicating that the inhibitory blockers effectively occluded the excitatory actions of 8-OH-DPAT.
Inhibitory receptor blockers occluded the 8-OH-DPAT-induced excitation in Lmx1bf/f/p slice preparations. Hypoglossal burst frequency at baseline, during application of an inhibitory blocker mixture consisting of gabazine (0.5 μm), CGP 55845 (1 μm), and strychnine (0.5 μm), and during coapplication of inhibitory blockers and 8-OH-DPAT (100 μm; n = 8 Lmx1bf/f/p). Shown are both the peak and steady-state responses to 8-OH-DPAT (attained at the end of 10 min of drug application; one-way RM ANOVA followed by Bonferroni post hoc for multiple comparisons; p < 0.05). *Significant difference. Data are mean ± SEM.
Discussion
This study exploits a transgenic mouse model to differentiate and identify the roles in respiratory control of 5-HT1a autoreceptors located on serotonergic neurons, versus heteroreceptors located on downstream nuclei. The bulk of the evidence shows that 5-HT neurons have a net stimulatory effect on the respiratory network (for review, see Richerson, 2004; Hodges and Richerson, 2008a; Ptak et al., 2009; Depuy et al., 2011). However, it has alternately been proposed that the net effect of 5-HT is inhibitory (Lalley, 1986, 1994), or that 5-HT acts as a “modulator” of respiratory output (Hodges and Richerson, 2008a). We propose that the discrepancies regarding the effect of 5-HT on breathing resulted from expression of multiple 5-HT receptor subtypes at different locations with varying effects on respiratory output. In particular, activation of 5-HT1a receptors results in two competing effects: initial stimulation of breathing due to excitation of the respiratory network and a secondary decrease in stimulation of breathing due to silencing of 5-HT neurons.
Validation of the Lmx1bf/f/p mouse as a tool for studying 5-HT1a receptor distribution and functions
These experiments are the first time the in vitro rhythmic medullary slice preparation has been used from Lmx1bf/f/p mice (Zhao et al., 2006). Here we show that despite lacking >99% of central serotonergic neurons, slices prepared from neonatal Lmx1bf/f/p mice produced a relatively consistent, albeit altered, respiratory-related rhythmic discharge. There was a slower and more irregular hypoglossal inspiratory burst pattern than observed in WT preparations. This pattern of discharge is reminiscent of phrenic nerve recordings from the in vitro brainstem–spinal cord preparation of Lmx1bf/f/p mice (Hodges et al., 2009).
These findings were paralleled in vivo, where Lmx1bf/f/p neonates ventilate more slowly than WT siblings and with shallower breaths, and also exhibit an increased number of apneas (Hodges et al., 2009).
Interestingly, previous observations point to an absolute requirement for 5-HT in generation of the respiratory rhythm in neonatal mouse and rat slices (Peña and Ramirez, 2002; Ptak et al., 2009). In contrast, we recorded spontaneous respiratory bursting in slice preparations devoid of central 5-HT neurons, suggesting that there might be some compensatory mechanism in Lmx1bf/f/p mice to maintain respiratory output. However, our results indicate that this compensatory excitatory drive in Lmx1bf/f/p mice is not due to an increase in noradrenergic tone. The maintenance of respiratory output could be due to an increase in other inputs, such as glutamatergic tonic drive (Crone et al., 2012), or there could be changes in expression of ion channels or synaptic connections within the respiratory network consistent with homeostatic plasticity reported in other systems (Turrigiano and Nelson, 2004; Marder and Goaillard, 2006).
In order for the Lmx1bf/f/p mouse to be a valid tool for differentiating between 5-HT1a autoreceptor and heteroreceptor function, it was necessary to confirm that the numbers of 5-HT neurons and their 5-HT1a autoreceptors were diminished. Evidence presented includes immunohistochemical staining for TpOH-positive cells, which revealed a >99% absence of 5-HT neurons in the medulla of Lmx1bf/f/p mice, in agreement with previous observations (Ding et al., 2003; Zhao et al., 2006). Additionally, there was a decrease in 5-HT1a receptor binding sites in raphe nuclei as assessed by autoradiography. Interestingly, the number of 5-HT1a receptors outside the raphe (which likely included the preBötC in the CVL sampled) remained unchanged, suggesting that the loss of 5-HT did not lead to upregulation or downregulation of 5-HT1a heteroreceptors in these areas. We also assessed the functional consequence of a loss of 5-HT neurons by demonstrating loss of the 5-HT2a receptor-mediated respiratory drive that normally stimulates the respiratory network (Peña and Ramirez, 2002; Ptak et al., 2009). The 5-HT2a receptor antagonist MDL 11939 (50 μm) reduced respiratory output in slices from WT mice as expected, but had no effect in slices from Lmx1bf/f/p mice, consistent with the absence of endogenous 5-HT.
These data support the use of this mouse as a unique tool to investigate the effect of activating downstream 5-HT1a heteroreceptors on respiratory output in the absence of a confounding effect of 5-HT neuron inhibition by activation of 5-HT1a autoreceptors.
Differentiation of inhibitory autoreceptor and downstream heteroreceptor activation in respiratory control
5-HT1a receptors are widely distributed throughout the brain including regions involved in respiration such as the preBötC, retrotrapezoid nucleus/ parafacial respiratory group, raphe nuclei, nucleus of the solitary tract, and hypoglossal nuclei (Liu and Wong-Riley, 2010). Those receptors located on raphe serotonergic cells serve a unique function, providing autoinhibition via negative feedback (McCall and Clement, 1989; Sharp et al., 1989; Veasey et al., 1995; Aghajanian and Sanders-Bush, 2002).
Our results demonstrate downstream excitation of respiratory output via 5-HT1a heteroreceptors. Using Lmx1bf/f/p mice, in which only downstream 5-HT1a heteroreceptors exist, we found that 8-OH-DPAT increased ventilation in vivo, and increased hypoglossal inspiratory burst frequency in vitro. These observations are consistent with previous work demonstrating that 5-HT1a receptors facilitate breathing via involvement in the hypoxic ventilatory response (Hilaire and Duron, 1999; Bou-Flores et al., 2000; Nucci et al., 2008) and in the counteraction of respiratory depression caused by spontaneous central apneas, apneusis, and opioid-induced ventilatory depression (Lalley et al., 1994; Sahibzada et al., 2000; Meyer et al., 2006; Stettner et al., 2008; Dutschmann et al., 2009; Guenther et al., 2009; Manzke et al., 2009). The latter compensation was confirmed as 5-HT1a receptor activation since the effect of 8-OH-DPAT was blocked by WAY-100635 (a 5-HT1a receptor antagonist). Other 5-HT1a agonists have also been used to treat apneusis and apneic periods (Wilken et al., 1997; El-Khatib et al., 2003).
The mechanisms for downstream excitation of the respiratory system are unclear. 5-HT1a receptors are coupled to inhibitory G-proteins (Raymond et al., 1999), so a potential mechanism for the excitatory action of 8-OH-DPAT is through disinhibition of the respiratory network. Downstream 5-HT1a receptors are predominantly located on inhibitory glycinergic respiratory interneurons, therefore their activation by a ligand would inhibit those inhibitory neurons, resulting in excitation (Dutschmann et al., 2009; Manzke et al., 2009). This could possibly also involve a complex reorganization of respiratory network operation in vivo (Shevtsova et al., 2011). This hypothesis of network disinhibition was confirmed by the occlusion of 8-OH-DPAT-induced excitation by inhibitory receptor (glycinergic and GABAergic) antagonists (Fig. 9).
In contrast, 8-OH-DPAT depressed breathing activity in WT mice in vivo, and in WT slices in vitro. WT mice have 5-HT1a autoreceptors and heteroreceptors, thus interpretation of 8-OH-DPAT data must consider the net effect of both types of 5-HT1a receptors. Pharmacological activation of 5-HT1a receptors in the medullary raphe decreases 5-HT neuron firing and reduces heart rate and ventilation (Ootsuka and Blessing, 2006; Nakamura and Morrison, 2007). Additionally, activation of inhibitory 5-HT1a autoreceptors by focal injection of 8-OH-DPAT into the raphe results in blunting of the hypercapnic ventilatory response (Messier et al., 2004; Taylor et al., 2005). The action of 5-HT1a autoreceptors is further demonstrated by their involvement in other autonomic functions. Mice with excessive 5-HT inhibition due to overexpression of 5-HT1a in serotonin reuptake transporter-expressing cells exhibit sporadic autonomic crises and increased spontaneous death occurring between P25 and P80 (Audero et al., 2008). Our results demonstrate that in neonatal mice, the inhibitory 5-HT1a autoreceptor effect of 8-OH-DPAT dominates over the downstream excitatory effect. The inhibition of 5-HT neurons via autoreceptor activation and resulting decrease in endogenous serotonin release onto downstream excitatory 5-HT2a receptors can explain the observed depression in respiratory motor output.
Distinguishing between downstream mechanisms
We have ruled out the possibility that the moderate affinity of 5-HT7 receptors for 8-OH-DPAT (Bard et al., 1993; Eriksson et al., 2008) could explain the excitatory action on respiration. Not much is known regarding the role of 5-HT7 in breathing, however, this G-protein coupled receptor is positively linked to adenylate cyclase, is expressed in the preBötC, and has been detected in the raphe magnus. It colocalizes with 5-HT neurons and is involved in restoring phrenic nerve activity following fentanyl-induced depression of respiratory discharge (Barnes and Sharp, 1999; Muneoka and Takigawa, 2003; Richter et al., 2003). We administered a 5-HT7 antagonist (SB 269970) alone, and then with 8-OH-DPAT to isolate the effect of the latter drug on 5-HT1a receptors. A similar approach has been used previously in studies on thermoregulation and learning (Hedlund et al., 2004; Faure et al., 2006; Eriksson et al., 2008). SB 269970 alone did not alter hypoglossal output in either Lmx1bf/f/p or WT preparations. It also failed to inhibit the 8-OH-DPAT-induced increase in frequency observed in Lmx1bf/f/p preparations, supporting the conclusion that downstream 5-HT1a receptors facilitate breathing.
Our study provides a unique view identifying which receptors mediate the effect of 8-OH-DPAT on respiration, and rule out 5-HT7 receptors as targets of 8-OH-DPAT in the context of respiratory control. Our results argue the need for cautious interpretation when using 8-OH-DPAT as a 5-HT1a agonist in respiratory control studies. We confirm differential roles for 5-HT1a receptors in breathing, depending on their synaptic location. There is a dominant inhibitory effect due to inhibition of 5-HT neurons, and a stimulatory effect on the respiratory network due to activation of 5-HT1a heteroreceptors. The latter stimulatory effect would be the dominant effect when 5-HT is released under normal conditions in vivo.
Footnotes
This work was supported in part by the Intramural Research Program of the NIH, NINDS, and from the NIH Grants 2U54NS041069, NIH/NICHD, P01HD36379, and R01HD052772.
The authors declare no competing financial interests.
- Correspondence should be addressed to Dr. Andrea E. Corcoran, Department of Physiology and Neurobiology, Geisel School of Medicine at Dartmouth, 1 Medical Center Drive, Lebanon, NH 03756. andrea.e.corcoran{at}dartmouth.edu
