Abstract
In rodent neonates, the neurotransmitter serotonin (5-HT) modulates the activity of both the medullary respiratory rhythm generator and the cervical phrenic motoneurons. To determine whether 5-HT also contributes to the maturation of the respiratory network, experiments were conducted in vitro on the brainstem–spinal cord preparation of neonatal mice originating from the control strain (C3H) and the monoamine oxidase A-deficient strain, which has a brain perinatal 5-HT excess (Tg8). At birth, the Tg8 respiratory network is unable to generate a respiratory pattern as stable as that produced by the C3H network, and the modulation by 5-HT of the network activity present in C3H neonates is lacking in Tg8 neonates. In addition, the morphology of the phrenic motoneurons is altered in Tg8 neonates; the motoneuron dendritic tree loses the C3H bipolar aspect but exhibits an increased number of spines and varicosities. These abnormalities were prevented in Tg8 neonates by treating pregnant Tg8 dams with the 5-HT synthesis inhibitor p-chlorophenylalanine or a 5-HT2A receptor antagonist but were induced in wild-type neonates by treating C3H dams with a 5-HT2A receptor agonist. We conclude that 5-HT contributes, probably via 5-HT2A receptors, to the normal maturation of the respiratory network but alters it when present in excess. Disorders affecting 5-HT metabolism during gestation may therefore have deleterious effects on newborns.
- serotonin
- serotonergic receptors
- medullary respiratory network
- electrophysiology of phrenic motoneurons
- morphology of phrenic motoneurons
- fetal mice
- neonatal mice
- maturation
- modulation of neonatal respiratory activity
The respiratory network comprises a medullary network responsible for setting up the rhythmic drive (Bianchi et al., 1995; Koshiya and Smith, 1999) and several pools of motoneurons located in the medulla and the spinal cord (Monteau and Hilaire, 1991; Hilaire and Monteau, 1997). This network is able to produce rhythmic activity in phrenic motoneurons (PhMns) in vitro by embryonic day 15 (E15) in rats, i.e., 1 week before birth (Hilaire and Duron, 1999), although the rhythm and amplitude of this activity are very variable at this age (Di Pasquale et al., 1992). At E18, endogenous serotonin (5-HT) is indispensable to rhythmogenesis because blocking medullary 5-HT1A receptors (5-HT1AR) induces respiratory arrests, whereas activating them increases the respiratory frequency (Di Pasquale et al., 1994a). E18–E20 is a critical period in network maturation; respiratory bursting pacemaker neurons begin to fire in the medullary network, the respiratory rhythm and amplitude stabilize, and 5-HT is no longer required (Di Pasquale et al., 1994b, 1996). At birth, 5-HT exerts complex effects, including facilitatory effects on the rhythm generator via medullary 5-HT1AR, the recruitment of PhMns via 5-HT2AR located on their somatic membrane, and inhibitory effects on the transmission of the central drive to PhMns via presynaptic 5-HT1BR (Morin et al., 1991a,b; Lindsay and Feldman, 1993; Di Pasquale et al., 1994a,1997; Monteau et al., 1994; Hilaire et al., 1997b).
5-HT is a widely distributed neurotransmitter that is expressed from an early embryonic age and is involved in several nervous maturational processes and functions (Lauder and Krebs, 1978; Vogt, 1982; Lauder et al., 1985; Whitaker-Azmitia, 1991; Hamon and Gozlan, 1993; Lauder, 1993; Hoyer et al., 1994; Peroutka, 1994; Levitt et al., 1997). To determine whether 5-HT affects the maturation of the respiratory network, we compared the respiratory activity at birth between two strains of mice that have different perinatal endogenous 5-HT levels, the wild C3H/HeJ strain (C3H) and the transgenic Tg8 strain. The latter was obtained from C3H (Cases et al., 1995) by performing insertional inactivation on the gene encoding monoamine oxidase A (MAOA), the main enzyme that degrades 5-HT. MAOA deficiency results in brain 5-HT levels that are several times greater than in normal C3H mice (Lajard et al., 1999).
In both strains, we measured the 5-HT levels at E18 and analyzed thein vitro respiratory activity between E20 and postnatal day 5 (P5), the responses to exogenous 5-HT, the electrophysiological properties and dendritic features of PhMns at P0–P3, and the expression of 5-HT1AR at P7. Several Tg8 anomalies were observed, including failure to develop a stable respiratory rhythm, lack of respiratory response to 5-HT1AR activation, and an abnormal morphology of PhMn dendrites. Because these abnormalities were prevented by prenatal pharmacological treatments affecting the 5-HT system, they probably originated from maturational disturbances attributable to excess 5-HT, which suggests that 5-HT is actually involved in the maturation of the respiratory network.
MATERIALS AND METHODS
HPLC experiments. HPLC measurements of 5-HT levels were performed on the brainstem and cervical cord of exteriorized fetuses dissected out from pentobarbitone-anesthetized (60 mg/kg) timed-pregnant C3H and Tg8 mice on gestational day 18 (Lajard et al., 1999). Results are expressed as picograms per milligram of tissue.
Electrophysiological experiments. Experiments were performedin vitro on the en bloc medulla–spinal cord preparation of anesthetized fetal and neonatal C3H and Tg8 mice (Di Pasquale et al., 1992; Hilaire et al., 1997a,b). Briefly, the medulla and spinal cord were isolated and superfused with artificial CSF (aCSF) gassed with 95% O2 plus 5% CO2 and maintained at 26.5 ± 1°C. The activity of the C4 phrenic roots was recorded with suction electrodes and integrated. To estimate the variability of the phrenic activity, the cycle duration and the amplitude of integrated phrenic bursts were measured during 60 successive cycles, and coefficients of variability (CVd) were defined as the SD/mean ratio.
Whole-cell recordings were performed using an Axoclamp 2A amplifier (Axon Instruments, Foster City, CA) and 6–8 MΩpatch microelectrodes (Di Pasquale et al., 1996). The neurons selected had a 3–5 GΩ seal, a stable resting membrane potential, and a serial resistance of <30 MΩ. The following parameters were measured between phrenic bursts: (1) the resting potential, (2) the spike amplitude and duration at half amplitude, (3) the membrane input resistance (from voltage changes elicited by a stepped 1 sec negative pulses), (4) the firing threshold (the absolute membrane potential at the onset of the first action potential elicited by a 1 sec depolarizing pulse), and (5) the slope of the I–Frelationship when 1 sec depolarizing pulses with a stepped intensityI induce a train of spikes at the frequencyF.
Analysis of phrenic motoneuron morphology. PhMns were recorded with pipettes containing biocytin for at least 1 hr (resting potential clamped at −70 mV). The cervical cord was fixed with 4% paraformaldehyde in 0.1 m sodium phosphate buffer, pH 7.4 (48 hr). Frontal sections (150 μm) were obtained on a vibroslicer (Campden Instruments, Loughborough, UK) and collected in chilled 0.1 m sodium PBS (NaPBS) pH 7.4. Sections were rinsed in NaPBS, treated with Triton X-100 (1% in NaPBS; Sigma, St. Louis, MO), and incubated in avidin-biotin-HRP complex (ABC Elite kit; Vector Laboratories, Burlingame, CA). After several rinses, they were reacted with diaminobenzidine (DAB kit; Vector Laboratories). The sections were mounted onto gelatin-coated slides, dehydrated (100% ethanol), cleared (xylene), and coverslipped with DePex. Neurons were observed under aZeiss (Oberkochen, Germany) microscope (Axiophot) and drawn under a camera lucida. Each PhMn generally appeared on one to two successive slices in the case of C3H and often on three slices in that of Tg8. Drawings were analyzed with NIH Image software to define the area of the soma, the number, and length of the dendrites of various orders, and the total extent of the dendritic tree. Dendritic spines (thin dendrites <5 μm in length) and swollen varicosities were counted under a Zeiss microscope.
Pharmacological experiments. As described previously (Monteau et al., 1994; Hilaire et al., 1997b), the following drugs [Sigma and Research Biochemicals (Natick, MA)] were diluted in aCSF and applied to the preparations for 4 (5-HT and 5-HTR agonists), 10 (antagonists), and 20 [l-tryptophan (l-Trp)] min: 5-HT and 8-hydroxy-2-(di-n-propyl-amino) tetralin (8-OH-DPAT) as 5-HT1AR agonists; methoxyphenyl)-4-[4- (2-phthalimido)butyl] piperazine hydrobromide (NAN190) and 4-iodo-N-[2-[4-(methoxyphenyl)-1-piperazinyl]ethyl]-N-2-pyridinyl-benzamide hydrochloride (p-MPPI) as 5-HT1AR antagonists; 5-methoxy 3(1,2,3,6-tetrahydro 4-pyridinyl) 1-H-indol-succinate (RU24969 from Hoechst Marion Roussel, Frankfurt am Main, Germany) as 5-HT1BR agonist;S(−)-1-(1H-indol-4-yloxy)-3-[(1-methylethy)amino]-2-propanol[(−)pindolol] as 5-HT1BR antagonist; 2,5-dimethoxy-4-iodoamphetamine hydrochloride (DOI) and αCH3–5-HT as 5-HT2AR agonists; and {trans,4-([3Z)3-(2-dimethylaminoethyl) oxyimino-3(2-fluorophenyl)propen1-yl] phenol hemifumarate} (SR46349B from Sanofi-Recherche, Montpellier, France) as 5-HT2AR antagonist. In some experiments, the recording chamber was divided by placing a barrier at the level of the first cervical root, so as to be able to apply drugs to either the medulla or the spinal cord. The control phrenic burst frequency was measured for at least 5 min before any drug application; the frequency under aCSF-containing drug was expressed as a percentage of the control value (100%).
To test the effects of short-term treatment on PhMn morphology, 5-HT (200 μm) and SR46349B (80 μm) were added to aCSF. The aCSF-containing drug (either 5-HT alone or 5-HT plus SR46349B) was applied instead of normal aCSF while a given PhMn was being recorded with a pipette containing biocytin. To test the effects of long-term treatment, SR46349B and DOI were injected subcutaneously (0.3 ml) to timed-pregnant mice (50 mg/kg per 24 hr) from E18 to delivery; sham experiments were performed with 0.3 ml of saline (n = 2). The 5-HT synthesis inhibitorp-chlorophenylalanine (PCPA) (Research Biochemicals) was injected into Tg8 (n = 3) and C3H (n = 2) pregnant mice (300 mg/kg per 24 hr) from E18 to delivery and into Tg8 neonates from birth up to and including P2 (n = 3).
Statistics. Results were expressed as the mean ± SEM in all the experiments, except for those on the PhMn morphology, which were expressed as the median ± quartile deviation to avoid having to make any assumptions about the normality of the distribution. Pharmacological experiments were repeated on several preparations to evaluate any differences existing between control aCSF and drug-containing aCSF, which were taken to be significant atp < 0.05 (paired t test). In biochemical and electrophysiological experiments, differences between the mean C3H and Tg8 values were taken to be significant at p < 0.05 (unpaired t test). Regarding the PhMn morphology, the working hypothesis that C3H and Tg8 PhMns were different was tested by comparing the data (significance level of p < 0.05) by performing nonparametric tests, i.e., the Mann–Whitney Utest for two independent groups. The working hypothesis that the pharmacological treatments affected the PhMn morphology was evaluated (p < 0.05) by performing an ANOVA-1 test on ranks on the three groups (C3H, Tg8, and a treated group), i.e., the Kruskal–Wallis test followed by Dunnett's test (SigmaStat Software).
Autoradiographic experiments. 5-HT1AR binding experiments were performed on cryostat-prepared, 25-μm-thick brainstem and cervical cord sections (Boulenguez et al., 1991) from four sets of 7-d-old C3H and Tg8 neonates. Three sets consisted of one C3H and one Tg8 pup, and the fourth set comprised two C3H neonates from the same litter and one Tg8 neonate. The brain tissue from each set was processed and analyzed simultaneously throughout the experiment, from freezing to quantitative autoradiography. Serial frontal sections were collected on gelatin-coated slides, which were stored at −80°C before being incubated in a specific 5-HT1AR radioligand, [3H]8-OH-DPAT (specific activity, 127 Ci/mmol, 2 nm in Krebs' medium, at 25°C for 1 hr; NEN, Boston, MA). After quick washing and drying, sections from similar regions of both strains were exposed on the same sheet of film for 2.5–3 months ([3H]Hyperfilm; Amersham Pharmacia Biotech, Arlington Heights, IL). Quantitative analysis of the autoradiograms was performed with a computer image analysis device [PowerMacIntosh (Apple Computers, Cupertino, CA) and the Biolab software program]. Histograms were drawn, and statistical analysis (unpaired t test) was performed using the Prism 2.0 software program (GraphPad, San Diego, CA). The mean value of the specific binding recorded in each structure was expressed in femtomoles per milligram of tissue (±SEM), based on nsections analyzed from three to five individuals of each strain.
RESULTS
Very high 5-HT concentrations occur in Tg8 fetuses at E18
The 5-HT levels are only 4- and 1.5-fold higher than normal in brainstem and cervical cord from Tg8 fetuses at E20, whereas they are 8- and 3.5-fold higher in Tg8 neonates at P5 (Lajard et al., 1999). The increase in endogenous 5-HT levels in Tg8 mice from E20 to P5 raises the question of 5-HT levels at E18 because a major step in respiratory network maturation occurs between E18 and E20. We therefore measured the 5-HT levels in E18 fetuses of C3H and Tg8 mice. They were 4.8-fold higher than normal in Tg8 brainstem (986 ± 92 vs 204 ± 85 pg/mg, mean ± SEM, n = 4) and 6.3-fold higher in Tg8 cervical cord (729 ± 203 vs 115 ± 95 pg/mg). During maturation from E18 to P5, Tg8 PhMns and medullary neurons are actually exposed to very high 5-HT levels.
Greater variability of phrenic burst cycle durations and integrated phrenic burst amplitudes in Tg8 neonatal preparations
In both C3H and Tg8 medulla–cervical cord preparations (Fig.1A), the isolated respiratory centers produced phrenic bursts at similar frequency between E20 and P5 (8.10 ± 0.25 cycles/min,n = 34), but the phrenic bursts were more variable in Tg8 preparations in terms of the cycle duration and amplitude (Fig.1B). The CVd at E20 was found to be slightly greater in Tg8 fetuses, but this difference was not significant. Although CVd then decreased with age in both strains (Fig.1C,D), the mean CVd was greater in Tg8 than in C3H neonates at P0–P5 (Fig. 1E). The amplitudes of the integrated phrenic bursts evolved in a similar way, showing a high coefficient of variability in fetuses of both strains at E20 (0.42 ± 0.03, n = 8). This coefficient then decreased at birth but was greater in Tg8 than in C3H neonates at P0–P2 (0.38 ± 0.04, n = 21 vs 0.24 ± 0.04, n= 18) and P3–P5 (0.26 ± 0.03, n = 9 vs 0.17 ± 0.03, n = 11).
MAOA deficiency increases both 5-HT and norepinephrine levels (Cases et al., 1995). To determine whether the abnormal variability of the Tg8 phrenic bursts is attributable to 5-HT excess, we administered the 5-HT synthesis inhibitor PCPA, either from P0 to P3 or from E18 to birth. Postnatal PCPA treatment did not significantly affect the mean CVd in Tg8 preparations at P3–P5 (0.27 ± 0.05, n = 3) compared with control Tg8 neonates (0.30 ± 0.04,n = 9), whereas prenatal PCPA treatment significantly decreased the CVd (Fig. 1F). Tg8 neonates born from PCPA-treated dams showed more stable cycle durations at P0–P2, with lower CVd (0.30 ± 0.06, n = 4) than Tg8 neonates born from control dams (0.46 ± 0.03, n = 21). The fact that the CVd of these prenatally treated Tg8 neonates was comparable with that of C3H neonates (0.30 ± 0.03,n = 18) suggests that 5-HT excess is responsible for the CVd alterations observed in Tg8 preparations. In addition, the effects of prenatal PCPA indicates that the greater variability of the cycle durations observed in Tg8 neonates is of developmental origin. C3H neonates born from PCPA-treated dams showed very variable cycle durations at P0–P2, and their CVd were higher (0.71 ± 0. 09,n = 5) than those of C3H neonates born from control dams (0.30 ± 0.03, n = 18). The changes in the neonatal respiratory pattern observed in response to both increases and decreases in the 5-HT levels during the prenatal period show that 5-HT plays a crucial role in the maturation of the central respiratory network.
We then investigated whether the Tg8 variability may be mediated by the 5-HT2AR. To mimic in C3H fetuses the intense activation of the 5-HT2AR that probably occurs in Tg8 fetuses because of the high 5-HT levels, we administered the 5-HT2AR agonist DOI to pregnant C3H dams from E18 to delivery. Conversely, to prevent activation of the 5-HT2AR in the Tg8 fetuses, we applied the 5-HT2AR antagonist SR46349B to Tg8 dams.
Prenatal DOI treatment of C3H dams (Fig. 1F) significantly increased the mean CVd in P0–P2 neonates (0.43 ± 0.03, n = 4) compared with control C3H neonates (0.30 ± 0.03, n = 18), whereas postnatal DOI treatment to C3H pups from birth to P2 did not affect the mean CVd at P3–P5 (0.22 ± 0.05, n = 5) compared with control C3H pups (0.19 ± 0.02, n = 11). Prenatal 5-HT2AR antagonist SR46349B treatment of Tg8 dams (Fig. 1F) significantly lowered the mean CVd in P0–P2 neonates (0.38 ± 0.05, n = 5) compared with control Tg8 neonates (0.46 ± 0.03, n = 21). Therefore, 5-HT2AR are likely contributors to the deleterious effects of 5-HT excess.
Abnormal 5-HT modulation of respiratory activity in medulla–spinal cord preparations of Tg8 neonates
As reported in rats and OF1 mice pups (Morin et al., 1991a;Lindsay and Feldman, 1993; Di Pasquale et al., 1994a, 1997; Monteau et al., 1994; Hilaire et al., 1997b), replacing normal aCSF by aCSF containing 25 μm 5-HT (4 min) in P0–P3 C3H preparations (n = 16) increases the burst frequency by 23 ± 5%, induces a tonic discharge in the phrenic roots, and depresses the amplitude of the inspiratory bursts (Fig.2A1), whereas 5-HT neither accelerated the frequency nor depressed the phrenic burst amplitude in Tg8 mice (n = 10) but still induced a tonic discharge (Fig. 2A2).
Exogenous 5-HT and phrenic burst frequency
In P0–P3 C3H neonates, a dose-dependent increase in the burst frequency was elicited by applying 5-HT to the medulla alone (Fig.2B). This increase probably resulted from the activation of 5-HT1AR (Fig. 2C), because it was prevented by a pretreatment with the 5-HT1AR antagonist NAN190 (n = 7) but not with the 5-HT2AR antagonist SR46349B (n = 5). The increase was mimicked by the 5-HT1AR agonist 8-OH-DPAT (34 ± 8%,n = 8) but not by the 5-HT2AR agonists αC3H-5-HT (n = 5) and DOI (n= 5). The increases in the burst frequency induced by 5-HT may reflect a permanent facilitatory modulation of the rhythm generator by endogenous 5-HT (Fig. 2D); the burst frequency decreased after blocking the 5-HT1AR with an antagonist, namely NAN190 (17 ± 8%, n = 9) orp-MPPI (34 ± 11%, n = 7), but increased after activating 5-HT synthesis with the precursorl-tryptophan (17 ± 10%, n= 7).
In P0–P3 Tg8 neonates, 5-HT did not affect the burst frequency, even at high concentrations (Fig. 2B). Furthermore, the 5-HT1AR agonist 8-OH-DPAT, the two antagonists NAN190 (n = 9) and p-MPPI (n= 5) nor the 5-HT precursor l-tryptophan (n = 7) affected the Tg8 burst frequency (Fig.2D). The 5-HT modulation of burst frequency was therefore lacking. However, depleting the endogenous 5-HT levels by applying PCPA treatments to Tg8 dams (from E18 to birth,n = 3) or neonates (from birth to P2, n= 3) restored the 5-HT facilitatory effects; the Tg8 burst frequency increased by 25 ± 2% (n = 6) under 25 μm 5-HT (Fig. 2D, black bar).
Exogenous 5-HT and tonic discharges in phrenic roots
5-HT also induced a tonic discharge in the phrenic roots with a 2–3 min latency in ∼70% of both C3H and Tg8 P0–P3 neonates (Fig.2A1,2A2). The tonic discharge reflected the activation of cervical receptors, because it could be elicited by applying 5-HT to the cervical cord alone. In addition, the receptors involved were probably of the 5-HT2Atype, because the discharge was elicited by 25 μm of either the 5-HT2AR agonist αCH3–5-HT (n = 3 of 5) or DOI (n = 3 of 5) but not by the 5-HT1AR agonist 8-OH-DPAT (30 μm, n = 0 of 5). Lastly, the tonic discharge was prevented by pretreatment with the 5-HT2AR antagonist SR46349B (40 μm, n = 0 of 5) but not with the 5-HT1AR antagonist NAN190 (40 μm, n = 5 of 7).
Exogenous 5-HT and depression of phrenic burst amplitude
During the 5-HT-induced tonic discharge, the amplitude of the inspiratory bursts decreased greatly in C3H neonates, and the bursts were hardly discernible in the overall phrenic root recordings during the last minute of 5-HT application (Fig. 2A1). This effect did not involve either 5-HT1AR or 5-HT2AR, because it still occurred after antagonist treatments with 20 μm NAN190 (n = 5) and 40 μm SR46349B (n = 5), respectively. This amplitude decrease might be mediated by cervical 5-HT1BR, because the burst amplitude was depressed by applying the agonist RU24969 (n = 5) to the cervical cord (Fig.2E). Furthermore, the 5-HT-induced amplitude depression was prevented by pretreatment with 40 μm 5-HT1BR antagonist (−)pindolol (n = 5). In contrast, in Tg8 neonates, the inspiratory bursts remained unchanged during 5-HT application (25 μm, n = 10) (Fig.2A2), and their amplitude was not affected by the 5-HT1BR agonist RU24969 (n = 5).
Abnormally low expression of 5-HT1A receptors in the brainstem and cervical cord of Tg8 neonates
A comparative study on [3H]8-OH-DPAT binding in the brainstem and cervical cord was performed on C3H and Tg8 neonates at P7. Three distinct regions of the medulla were analyzed, namely the spinal trigeminal nucleus (SpV), the hypoglossal area, and the ambiguus area. The 5-HT1AR density was also determined in the dorsal and ventral horns of the cervical cord. In C3H pups, the highest densities were recorded in the SpV nucleus (32.9 ± 0.6 fmol/mg tissue) and slightly lower values in the dorsal horn (23.9 ± 0.3 fmol/mg) (Fig. 3A); in the hypoglossal, ambiguus and ventral horn areas, the densities were approximately one-third of those measured in the SpV. In Tg8 pups, the 5-HT1AR densities decreased by >52% in all four areas tested compared with the C3H pups (Fig. 3B–E). In the ventral horn of the cervical cord in which the PhMns are located, the 5-HT1AR density was only ∼5 fmol/mg tissue in Tg8 pups. The abnormally low level of expression of the 5-HT1AR observed in the mutant strain may therefore at least partly explain the loss of the respiratory frequency response to exogenous 5-HT via 5-HT1AR.
Electrophysiological study on cervical neurons in C3H and Tg8 neonates
At the C4 level, the phrenic nucleus was explored in C3H and Tg8 preparations at P0–P3 using whole-cell patch-clamp micropipettes; 186 neurons were tested and classified as either PhMns (n = 74) or interneurons (n = 112).
Membrane properties of phrenic motoneurons
Neurons were defined as being PhMns if they were antidromically activated by C4 ventral root stimulation (Fig.4A) and if they fired during the phrenic bursts (Fig. 4B). Among the 74 PhMns recorded from, 54 were selected on the basis of the excellent quality of the recordings (seal of >3 GΩ, overshooting potentials, and stable recording). The 23 C3H and 31 Tg8 PhMns tested did not differ significantly in their resting potentials during expiration (−68 ± 1 mV), spike amplitudes (74 ± 2 mV), spike half durations (1.02 ± 0.03 msec), or firing thresholds (−51 ± 1 mV). When driven above the firing threshold by 1 sec depolarizing pulses, the PhMns fired a sustained burst of potentials (Fig.4C), and the relationship between the intensity (I, in picoamperes) of the depolarizing pulses and the frequency (F, in Hertz) of the potentials was identical in both strains (I–F slope, 0.15 ± 0.01 pA/Hz). The membrane input resistance R was calculated from the membrane voltage changes V elicited by 1 sec hyperpolarizing pulses with an increasing stepped intensity I (Fig.4D) in line with the relationship r =V/I. The membrane input resistance of PhMns was found to be larger in C3H (94 ± 10 MΩ) than in Tg8 (68 ± 7 MΩ) neonates (Fig. 4E).
Respiratory discharges of phrenic motoneurons
In both strains, the PhMns were silent during expiration, depolarized by the central inspiratory drive (10–20 mV), and fired a sustained burst of potentials during inspiration. The amplitudes of the inspiratory depolarizations imposed by the central drivers and the resulting spike patterns were stable in C3H PhMns. In Tg8 PhMns, on the contrary, the spike discharge varied from one inspiration to the next (Fig. 4F); a Tg8 PhMn could fire normally, deliver only a few action potentials, or fire a long, sustained burst of potentials. Because PhMns are known to be monosynaptically controlled by their central drivers (Monteau and Hilaire, 1991), the variability of the Tg8 PhMn discharges may have originated from variations in (1) the central drive they received, (2) its transmission to PhMns, or (3) the PhMn responsiveness.
In both C3H and Tg8 neonates, PhMns produced stable bursts of potentials in response to repetitive depolarizing pulses applied during expiration with a constant intensity. The Tg8 burst variability was therefore not attributable to variations in PhMn responsiveness. When PhMns were silenced by a hyperpolarization imposed on their membrane (-120 mV) so that the inspiratory depolarizations evoked by the central drivers failed to reach the firing threshold, the inspiratory depolarizations occurred without any superimposed spiking. They were stable from one inspiration to the next in C3H but variable in Tg8 PhMns. The fact that the time course of the synaptic drive potential was found to mirror the underlying synaptic drive current measured using the voltage-clamp technique (Liu et al., 1990) means that the synaptic drive current received by the PhMns was variable in Tg8 neonates.
In addition, the spiking variations in a given Tg8 PhMn perfectly reflected the amplitude variations of the integrated phrenic bursts (Fig. 4F). This means that all of the Tg8 PhMns showed the same concomitant variations in their patterns of activity. If some failure of the synaptic transmission to PhMns had been responsible for the changes in burst amplitude, it would have been simultaneous in all the PhMns, but this possibility seems rather unlikely. The variability of the Tg8 discharges therefore originated from the variability of the central drive impinging on the PhMns. At birth, the Tg8 respiratory centers were able to generate neither a stable rhythm nor a respiratory command with a stable amplitude.
Scarcity of respiratory interneurons in the Tg8 phrenic nucleus
The neurons in the C4 ventral horn that were not antidromically activated by C4 ventral root stimulation were classified as cervical interneurons. They had less highly polarized resting potentials (47 ± 1 mV) and threefold higher membrane input resistances (364 ± 32 MΩ) than PhMns. They were further subclassified into “respiratory” and “nonrespiratory” interneurons on the basis of their discharge patterns. In C3H neonates, 49 interneurons were identified, 12 of which (18% of the sample) displayed a respiration-modulated activity. In Tg8 neonates, 62 of the 63 interneurons tested were nonrespiratory. Only one interneuron showed respiration-modulated activity (1.6%). The difference between the frequency of occurrence of respiratory interneurons in C3H versus Tg8 was significant (χ2 = 8.960,p = 0.003). There are several possible reasons why we observed very few interneurons firing with a respiratory modulation in the cervical cord of Tg8 mice. First, the presence of a sampling bias or some unknown technical cause cannot be completely ruled out. Second, the cervical respiratory interneurons exist in Tg8 mice, but they are located elsewhere within the cervical cord, probably at a more dorsal site, because we extensively explored the ventral horn. Third, these cervical interneurons may exist and be in the right place but may be rare because only a few of them have been contacted by the central respiratory drivers. This may mean either that the axons of the central drivers may project far from their normal place or that competition may exist between the cervical targets, i.e., the endings of the axon central drivers that should have contacted the cervical interneurons have in fact mainly contacted the abnormally numerous dendrites of PhMns in Tg8 instead of the cervical interneurons. Although negative results should always be treated with caution, the lack of Tg8 respiratory cervical interneurons observed here probably reflected their scarcity in the cervical ventral horn, which argues in favor of the idea that there exists an abnormal pattern of wiring between the medullary centers and their cervical targets.
The phrenic motoneuron morphology differs between C3H and Tg8 neonates
A total number of 22 PhMns (10 in C3H and 12 in Tg8 preparations at P0–P3) were stained with biocytin. Their soma were located in the same region of the C4 ventral horn, and their frontal area was ∼300 μm2 (262 ± 28 μm2 and 283 ± 32 μm2, for C3H and Tg8 PhMns, respectively). C3H and Tg8 PhMns showed obvious morphological differences in the pattern and extent of their dendritic trees (Fig.5). In C3H PhMns, most of the primary dendrites extended in two main directions (dorsolateral and ventromedial), forming a bipolar tree, although one to two primary short dendrites occasionally extended for a short distance in the dorsomedial direction, sometimes giving rise to a few secondary dendrites (Fig. 5A). In Tg8 PhMns, the bipolar dendritic tree was replaced by a multipolar tree (Fig. 5B). The primary dendrites extended in all possible directions and gave rise to numerous distal dendrites. The length and number of proximal dendrites were in the same range in C3H and Tg8 pups; six to seven primary-order dendrites and 11–12 second-order dendrites were detected in both strains, each of which extended for a distance of 25–40 μm. However, the Mann–Whitney U test for two independent groups showed that the number of distal dendrites (∼40 μm in length) was significantly greater in Tg8 (18 ± 6, 15 ± 6, and 13 ± 7 for dendrites of orders three, four, and five, respectively) than in C3H neonates (8 ± 3, 6 ± 3, and 1 ± 5). The total extent of the dendritic tree was therefore greater in Tg8 than in C3H PhMns (Fig. 5E); this modification is in agreement with the observed reduction in the membrane input resistance. Both C3H and Tg8 PhMns possessed dendritic spines (Fig. 5C) and swollen varicosities (Fig. 5D), but these features were rare in C3H and much more frequent in Tg8 (Fig.5F,G).
5-HT excess, 5-HT2A receptors, and phrenic motoneuron morphology
Pharmacological treatments were applied to fetuses and neonates to determine whether the abnormal morphology observed in Tg8 PhMns was attributable to the 5-HT excess. First, to mimic in C3H neonates the postnatal 5-HT excess present in the Tg8 strain, C3H preparations at P0–P3 were superfused with aCSF containing 200 μm 5-HT for ∼2 hr. The 5-HT-treated C3H PhMns were then injected with biocytin, and their morphology was compared with that of the control C3H PhMns. Among the seven stained PhMns, five still showed a bipolar dendritic tree, but two others had dendrites running in unexpected directions (Fig. 6A,arrow). The Kruskal–Wallis test and Dunnett's method were used to compare the three groups (treated C3H PhMns vs C3H and Tg8 control PhMns). The number and length of the proximal dendrites were not affected by the 5-HT treatment, but the number of order 4 dendrites was significantly larger, reaching the Tg8 range (15 ± 4). The total extent of the dendritic trees of the 5-HT-treated PhMns was slightly larger than the C3H control value, but the difference was not significant (Fig. 6B). The frequency of occurrence of dendritic spines and varicosities was much greater in the treated C3H PhMns than in the control C3H PhMns, however, reaching similar levels to those recorded in Tg8 PhMns (Fig.6C,D).
In a second set of experiments, C3H preparations were pretreated with the 5-HT2AR antagonist SR43349B (80 μm) for 1 hr, and the PhMns were then stained under aCSF containing both 5-HT (200 μm) and the 5-HT2AR antagonist SR46349B (80 μm). None of the four stained PhMns were morphologically affected by 5-HT application; the pattern of the dendritic trees, the total dendritic extent, nor the frequency of occurrence of the dendritic spines and varicosities differed from the control values.
In the third set of experiments, we mimicked in C3H fetuses the overactivation of the 5-HT2AR occurring in Tg8 fetuses. Pregnant C3H mice were treated with the 5-HT2AR agonist DOI from E18 to birth. Of four PhMns stained at P0, only one PhMn still showed a typical C3H dendritic tree, whereas the other three PhMns showed a multipolar-shaped dendritic tree (Fig. 7A). The total dendritic extent was greater than that of the control C3H PhMns and was in the Tg8 range (Fig. 7C). The number of distal dendrites differed from the C3H control value and was in the Tg8 range (25 ± 4, 25 ± 4, and 16 ± 7 for dendrites of orders 3, 4 and 5, respectively). The frequency of occurrence of dendritic spines and varicosities was larger than the C3H control value but similar to the Tg8 one (Fig. 7D,E). On the other hand, pregnant Tg8 dams were treated with the antagonist SR46349B from gestational day E18 to birth to prevent overactivation of the 5-HT2AR. At P0, none of the six stained PhMns displayed the classical Tg8 multipolar dendritic tree, but all showed a bipolar tree (Fig. 7B). They still exhibited, however, an extensive dendritic tree, a similar number of distal dendrites, and a similar rate of occurrence of the dendritic spines typical of Tg8 PhMns (Fig. 7C,D). Only the rate of occurrence of the varicosities was significantly lower than in the control Tg8 (Fig.7E). In neonates born from two pregnant Tg8 dams treated with saline, the morphology of the biocytin-labeled PhMns (n = 2) was similar to that of the control Tg8 PhMns.
DISCUSSION
To establish whether 5-HT plays a role in respiratory network maturation, comparisons were made between the respiratory activity of control C3H and MAOA-deficient Tg8 neonates characterized by prenatal 5-HT excess. At birth, three main anomalies were noted in the Tg8 respiratory network: its inability to develop a stable rhythm, a defective 5-HT modulation, and an abnormal PhMn morphology. In Tg8 neonates, MAOA deficiency affects both the noradrenaline and 5-HT levels, but the increase in the noradrenaline levels is very weak compared with the 5-HT excess (Cases et al., 1995). Although the possible involvement of noradrenaline cannot be completely ruled out, it therefore seems likely that the maturational abnormalities observed here originated mainly from 5-HT excess, because prenatal treatments affecting 5-HT function and 5-HT2AR activity affected the occurrence of these abnormalities.
Variability of respiratory activity and abnormal wiring of respiratory network in Tg8 neonates
In vitro, the rat respiratory network begins at E15–E16 to generate a rhythmic activity that is very variable at first but stabilizes with age. This stabilization occurred in wild C3H neonates, but in mutant Tg8 neonates, both the rhythm and amplitude of the respiratory activity are abnormally variable. The inability of the Tg8 centers to develop a stable activity several days after birth is likely to result from 5-HT2AR overactivation by 5-HT excess during the prenatal period; prenatal treatment of Tg8 fetuses with either PCPA or 5-HT2AR antagonist SR46349B stabilized the respiratory activity at birth, whereas prenatal treatment of C3H fetuses with 5-HT2AR agonist DOI increased the variability.
The variability observed in preparations from which the periphery has been eliminated may be attributable only to central factors (Di Pasquale et al., 1996). The Tg8 rhythm variability may therefore result from maturational deficits affecting the medullary respiratory network, via the membrane properties of some of its elements and/or the wiring between its different elements. The maturation of the PhMn membrane properties was not drastically affected in Tg8 mice, but more severe maturational disorders may have affected the inspiratory bursting pacemaker cells, which form the kernel of the medullary network (for review, see Hilaire and Duron, 1999). In addition, the presence of an abnormal wiring of the Tg8 medullary network cannot be ruled out, because some features suggesting abnormal wiring were observed at the cervical level. First, the fact that few respiratory interneurons were detected within the Tg8 cervical cord although these neurons have been commonly encountered in the cord of fetal and neonatal rats (Di Pasquale et al., 1996) and C3H mice may have resulted from an abnormal pattern of wiring between the medullary centers and their cervical targets. Second, PhMn dendritic trees differ greatly between C3H and Tg8 neonates. The bipolar dendritic tree of a C3H PhMn extends mostly in the two directions in which the axons of its central drivers run (Monteau and Hilaire, 1991, their Fig. 13). The Tg8 PhMn dendrites extend in all directions, which suggests that either the axons of their medullary drivers are not in the right place or the Tg8 PhMns receive other inputs that C3H PhMns do not receive. Moreover, the fact that dendritic spines known to establish synaptic contacts (Saito et al., 1992; Dailey et al., 1994) show a greater density again argues for the existence of an abnormally large number of synaptic inputs to the Tg8 PhMns. These arguments are all indirect ones, however, and it is not yet possible to draw any definite conclusions. Whether Tg8 rhythm variability reflects a transient or permanent dysfunction of the Tg8 respiratory network is another open question. The mouse respiratory network is thought to be fully mature by P15–P21 (Hilaire and Duron, 1999), but the respiratory network cannot be studied at this age using the present in vitro approach.
Abnormal regulation of respiratory activity by 5-HT and its receptors in Tg8 neonates
In neonatal mice, endogenous 5-HT exerts a facilitatory modulation on the respiratory rhythm generator via medullary 5-HT1AR (Hilaire et al., 1997b). This modulation exists in C3H neonates but is lacking in Tg8. The absence of 5-HT facilitatory effects in Tg8 may have resulted from the reduced expression of 5-HT1AR in the Tg8 medulla (binding data), probably attributable to 5-HT excess. This idea agrees with the finding that the 5-HT facilitatory effects were restored in Tg8 pups after 5-HT level normalization by PCPA prenatal and postnatal treatments.
At the cervical level, a dense 5-HT innervation from the raphe nuclei establishes synaptic contacts with the PhMns (Holtman et al., 1990;Pilowsky et al., 1990; Ellenberger et al., 1992), and activation of 5-HT2AR induces a tonic discharge in OF1 mouse pups (Hilaire et al., 1997b), as well as in both C3H and Tg8 pups.
Activating cervical presynaptic 5-HT1BR depresses the amplitude of the integrated inspiratory PhMn bursts (Lindsay and Feldman, 1993; Di Pasquale et al., 1997; Hilaire et al., 1997b). This depressive effect of 5-HT is present in C3H but lacking in Tg8 neonates.
The absence of both 5-HT1AR facilitatory effects on the rhythm generator and 5-HT1BR depressive effects on the burst amplitude in Tg8 mice contrasts with the persistence of the 5-HT2AR-mediated tonic discharge. This apparent discrepancy may be attributable to the fact that 5-HT1R and 5-HT2R react in different ways to the 5-HT excess. Prenatal treatment with 5-HT agents may abolish 5-HT1R but not 5-HT2AR responses (Shemer et al., 1991), although desensitization or downregulation of 5-HT2AR may occur after agonist activation (McKenna et al., 1989; Ivins and Molinoff, 1991; Roth et al., 1995), particularly in adult Tg8 mice (Shih et al., 1999).
Role of 5-HT2A receptors in some effects of 5-HT excess
The 5-HT2AR subtype is a good candidate for mediating the effects of 5-HT excess on the PhMn, because these receptors are present on their membrane (Morin et al., 1991b;Ellenberger et al., 1992; Lindsay and Feldman, 1993). In addition, various treatments affecting the 5-HT2AR induced PhMn morphological changes in both strains.
Short postnatal application of 5-HT to C3H preparations occasionally induced the development of one to two primary dendrites in unexpected directions, and this effect was prevented by 5-HT2AR antagonist treatment. It seems quite remarkable that such a brief 5-HT application can induce dendritic outgrowth within 2–3 hr. On the one hand, this result should be treated with caution because of the smallness of the sample size and the shortness of the latency. On the other hand, dorsal horn neurons may undergo dendritic reshaping a few minutes after intense activation of substance P receptors (Mantyh et al., 1995). In addition, morphological changes in the adult rat phrenic nucleus have been reported to develop within 2 hr of cervical hemisection (Sperry and Goshgarian, 1993), and 5-HT may be involved in these changes (Hadley et al., 1999; Zhou and Goshgarian, 1999). Last, the effects of short 5-HT applications are completely in agreement with those observed in fetuses after applying long-term treatments to the dams. Long-term DOI activation of the 5-HT2AR in fetuses increased the total dendritic extent and induced multiple branching of distal dendrites in neonates. Conversely, preventing the 5-HT2AR activation in Tg8 fetuses with the antagonist SR46349B affected the overall pattern of the dendritic tree, leading to Tg8 PhMns with large-sized arborizations, resembling C3H PhMns. These results strongly suggest that overactivation of 5-HT2AR affects the shaping of the PhMn dendritic tree.
The number of dendritic spines in C3H PhMns was smaller than in Tg8 PhMns but increased in response to both short postnatal (5-HT) and long prenatal (DOI) treatments activating the 5-HT2AR. In Tg8 fetuses, however, prenatal treatment with 5-HT2AR antagonist did not decrease the number of spines at birth. This may reflect 5-HT neurotrophic effects mediated via non-5-HT2AR.
The numerous varicosities formed by PhMn dendrites in Tg8 neonates resemble the swollen varicosities characteristic of 5-HT neurons. They may result from the 5-HT uptake and storage, because maturing neurons in Tg8 mice may transiently express the 5-HT transporter (Lebrand et al., 1996; Cases et al., 1998). These varicosities also closely resemble the histological profiles of internalized receptors (Mantyh et al., 1995) and may result from 5-HT2AR internalization in response to their overactivation by excess 5-HT. Rapid agonist-induced internalization of 5-HT2AR has been reported to occur in a clonal cell line, which was blocked by 5-HT2AR antagonist (Berry et al., 1996). These data are in line with the reduced number of varicosities present in Tg8 PhMns at birth after prenatal 5-HT2AR antagonist treatment and the increase in their number that occurs in C3H neonates after prenatal agonist treatment.
5-HT is involved in maturational processes, because it modulates cell division and differentiation, growth, and synaptogenesis (Lauder, 1993). High 5-HT levels affect the maturation of the thalamocortical (Cases et al., 1996) and retinal (Mooney et al., 1998) projections, the number of neurites present in hippocampal cells (Chubakov et al., 1986), the pattern of dendritic branching, and the number of spines occurring in dentate granule cells (Yan et al., 1997a,b). 5-HT1AR are thought to be responsible for these cellular changes (Sikich et al., 1990; Yan et al., 1997a,b). Our results suggest that 5-HT2AR contribute to shaping the PhMn morphology, but other receptor types might also be involved. The existence of multiple modes of control (Levitt et al., 1997) would explain why prenatal treatment with 5-HT2AR antagonist restored a bipolar dendritic tree in Tg8 PhMns without affecting either the total dendritic length or the number of dendritic spines. Depending on the cell type (interneuron vs motoneuron), the function of the structure (hippocampus vs ventral horn of the spinal cord), and the types and densities of the receptors locally expressed (low 5-HT1AR density in the cervical ventral horn) (Fig. 3A), different mechanisms might be involved in modulating the morphology of maturing cells.
In conclusion, the present results suggest that 5-HT actually plays an important role in respiratory network maturation. Genetic 5-HT excess in Tg8 neonates and PCPA-induced 5-HT depletion in C3H neonates both affect the ability of the respiratory centers to develop a stable respiratory activity at birth. Disturbances affecting 5-HT metabolism during pregnancy might therefore have deleterious effects on the maturation of the respiratory center, resulting in respiratory dysfunction at birth. The fetal 5-HT metabolism can be disturbed by numerous factors, such as stress-inducing situations (Peters, 1989), food restrictions (Hernandez et al., 1989), antidepressive drugs, etc. It is noteworthy that defective cardiorespiratory and sleep control, which are both functions modulated by 5-HT, might partly explain sudden infant death syndrome, the victims of which have been found to show abnormally high tryptophan and 5-HIAA CSF levels (Caroff et al., 1992;Cann-Moisan et al., 1999).
Footnotes
This research was supported by the Centre National de la Recherche Scientifique and the Institut Fédératif de Recherche des Sciences du Cerveau. C.B.-F. was the recipient of a three year fellowship from the Ministère de l'Enseignement, Recherche et Technologies while working on this project. We acknowledge the excellent contribution of Annie Christolomme to binding analysis, Marie Gardette to the figures, and Dr. Jessica Blanc for English revision.
Correspondence should be addressed to Gérard Hilaire, Centre National de la Recherche Scientifique, Unité Propre de Recherche 9011, Laboratoire Neurobiologie et Mouvement, 31 chemin Joseph-Aiguier, 13402 Marseille Cedex 20, France. E-mail: hilaire{at}dpm.cnrs-mrs.fr.