I. INTRODUCTION

Top Previous Next References

In this review, we describe the maturational changes occurring from the fetal stage to adulthood in the structural, functional, and regulatory processes at work in the central respiratory network. This network can be defined as a set of neurons located in the medulla and the cervical spinal cord that exhibit rhythmic electrical activity in phase with respiration. Some of these neurons are respiratory motoneurons (Mns), and special attention is paid in this review to those controlling the upper airways (especially the hypoglossal Mns, or XII-Mns, innervating the tongue) and the diaphragm (the phrenic Mns, or Phr-Mns). The other components include medullary and spinal interneurons. In this review, we focus in particular on those located in the medulla, which form the so-called respiratory centers, i.e., the part of the network that produces the rhythmic central drive received by the respiratory Mns.

In vivo studies on neonatal ventilatory activity are first reviewed. Obvious developmental changes have been described at the overall level which are likely to originate, at least partly, from central maturation, but almost no information is available so far about the changes occurring at the level of the respiratory medullary neurons. The maturation of the mechanisms underlying respiratory rhythmogenesis is then documented based on in vitro experiments performed on neonatal rodents (for review, see Ref. 350), whereas studies on the adult central respiratory network have been carried out for a long time on cats (for review, see Ref. 35). Problems arise here due to the interspecies and age-related differences (these are thoroughly discussed in sect. III). Because rat and cat adult respiratory networks may be different, there is little point in attempting to make deductions about maturational changes on the basis of comparisons between data on neonatal rats and adult cats. Fortunately, the discrepancies between the data on adult cats and neonatal rats have led to increasing attention being paid to the adult rodent respiratory network. We therefore first present the in vitro results obtained on newborn rodents (and even fetuses) and then those obtained on more mature animals, with special reference to rodents. Several experimental data support the idea that a special set of neurons in the rostral ventrolateral medulla (RVLM) may be respiratory bursting pacemaker neurons acting as primary respiratory neurons in neonates as well as in adults.

Ventilatory movements, which have been previously reviewed (190, 291), are complex motor acts that involve several types of muscles. To be efficient, the contractions of these muscles must be perfectly coordinated, and this is achieved at the level of the medulla as well as at that of the Mns that integrate central and peripheral inputs. We therefore report here on the morphological and electrophysiological maturation of the respiratory Mns.

Some of the mechanisms whereby respiratory activity is adapted to the physiological conditions are then analyzed at the levels of both the medullary respiratory centers and the respiratory Mns. Among the numerous modulatory substances involved, some of those that have been particularly well documented (serotonin, catecholamines, neuropeptides) have been described in greater detail. It has emerged that serotonin (5-HT) may play a crucial role; this is one of the first neuromediators to be expressed during ontogeny (43, 257, 265, 382), it is involved in nervous maturational processes (482), and it modulates the activity of the central network by acting at various levels (89, 181, 283, 293, 301, 302, 320), even at the fetal stage (86).

Most of the latest results have been obtained under in vitro conditions, and their physiological validity still remains to be confirmed. If the in vitro data turn out to be as reliable as they seem to be at present, we will be able to state with certainty that the respiratory bursting pacemaker neurons of the RVLM constitute the kernel of both the neonatal and adult respiratory networks. Meanwhile, however, we can work on the thought-provoking hypothesis that thanks to the plasticity of the respiratory network, several mechanisms may possibly be responsible for respiratory rhythmicity, depending on the state of the network.

	II. IN VIVO NEONATAL VENTILATORY PATTERN

Top Previous Next References

The in vivo data reported below illustrate some of the main developmental changes occurring in the ventilatory pattern from birth to adulthood. These data provide an overall picture of the changes in respiratory movements, in terms of their frequency, the characteristics of the respiratory cycle, and those of the electromyogram (EMG) discharge pattern, for example, and little information is available so far on the unitary activity of neonatal respiratory medullary neurons and Mns in vivo.

A.  Main Characteristics of the In Vivo Neonatal Respiratory Pattern

1.  Paradoxical neonatal respiratory movements

One of the first points worth mentioning is the immaturity of the rib cage effectors. In infants, a paradoxical inward movement of the rib cage has been commonly observed during inspiration (74, 229, 312). The muscular effort necessary for a normal tidal volume during these paradoxical movements is four times greater than during normal respiratory movements (152). The paradoxical inward deflection of the rib cage may be due to the combined effects of the highly compliant rib cage and the small intercostal muscle tone versus the development of the maximal diaphragmatic force. The low rigidity of the rib cage may be due to relatively immature segmental reflex loops and spindle afferents (424). The maturation of the proprioceptive reflex loops has been mainly studied in lumbosacral Mns (140, 141). There exist intercostal-diaphragmatic reflexes in adult cats (77, 392, 426), young kittens (456), and newborn infants (230), but abolishing proprioceptive afferents does not affect the intercostal muscle activity until postnatal day 14 (P14) (104, 424). Given the abundance of the spindle afferents within the various rib cage muscles, the inward respiratory movements observed may be due to a deficient proprioceptive regulation of the lateral intercostal muscles (98, 99, 217, 246, 267), but this does not rule out the possibility that the immaturity of the central command may be also involved. Decerebration of kittens induced a flexor activity in all the thoracic muscles (435, 480), instead of the classical extensor rigidity observed in adult cats that results from a -loop central control (103). The flexor rigidity induced by decerebration persists until P40 and suggests an abnormal central control on Mn activity.

One of the other main characteristics of the neonatal respiratory activity is its irregularity. Unpredictable spontaneous changes in respiration occur frequently in newborn animals and infants (see Fig. 1A), consisting of a mixture of episodes such as apnea, bradypnea, eupnea, and tachypnea. These episodes have been reported to occur in unanesthetized neonates of various species ranging in size from mice to infants (72, 124, 160, 312). Respiratory frequency at birth, often inversely correlated with the body weight, is slow in preterm infants that present an exacerbated respiratory variability. These infants breathe periodically or show prolonged apneas lasting for more than 20 s (398) along with bradycardia, which is rarely observed in full-term infants (208). The central apneas are often preceded by a sigh or accompanied by body movements (481) and are more frequent during active sleep and in the awake state, while quiet sleep is associated with more regular respiration (74, 361). The variability of the neonatal central rhythm may originate from 1) the intrinsic properties of the respiratory rhythm generator and/or 2) peripheral and central afferents acting on the rhythm generator. The frequency of occurrence of central apneas decreases with age, and in full-term babies, they tend to be replaced by obstructive apneas similar to those observed in adults (105). Central apneas have been reported to occur in unanesthetized decerebrate kittens (266) and in anesthetized kittens and rabbit pups (266, 460). Anesthesia stabilizes the respiratory activity and decreases the occurrence of central apneas. This improvement under anesthesia may be due to a depression of the afferent messages impinging on the respiratory rhythm generator.

View larger version (25K): [in this window] [in a new window]   Fig. 1. Some developmental characteristics of neonatal respiration. In A, variability of preterm infant respiration. Top and bottom traces, raw and integrated diaphragmatic electromyogram (EMG) discharges, respectively. Note spontaneous transitions from eupnea to tachypnea (A1) and bradypnea (A2). In B, discharge pattern of adult cat phrenic nerve (B1) and diaphragmatic EMG of preterm infant (B2) and newborn rat (B3; 2 successive inspirations). In neonates, inspiratory bursts are abrupt, are of short duration, and show low-frequency oscillation within discharge.

The particular respiratory response of neonates to hypoxia should be mentioned here, although it will be dealt with in greater detail in sections VA2 and VIA. In adults, hypoxia induces sustained hyperventilation, whereas this facilitatory response is very weak, and transient in neonates, and is followed by a marked respiratory depression (42, 208, 404). The initial facilitatory response disappears completely in cold environments (58).

The respiratory cycle differs between adults and neonates, since the inspiratory time (TI) is shorter and the expiratory time (TE) is longer in neonates than in adults. In decerebrate cats (266), for example, the adult TI and TE are ~1.30 and 1.45 s, respectively, with a TI/TE of 0.90, whereas the neonatal TI and TE are ~0.76 and 2.18 s, respectively, with a TI/TE of 0.35 on P7. The TI/TE increases with age, reaching 0.41 on P14, 0.74 on P21, and an almost adult value (0.88) at 5 wk of age. Similar low ratios have been reported in the case of unanesthetized neonatal mice and rabbits (312) and human infants (173). The lengthening of TI with age is one of the most conspicuous developmental changes; it has been observed even in the opossum, where TI at birth is ~100 ms, whereas the adult values are four to eight times longer (119). In neonates, most of the ventilatory adaptations involve TE changes, whereas TI remains quite constant. The decrease in respiratory frequency induced by either hypoxia or the deepening of anesthesia results from an increase in TE, whereas TI remains unchanged (401, 460).

Diaphragmatic and phrenic (Phr) nerve recordings have shown the existence of developmental changes (268, 448). In the adult cat, the amplitude of the Phr discharge increases slowly during inspiration (Fig. 1B1), due to the increasing discharge frequency of the individual Phr-Mns and the delayed recruitment of Mns throughout the inspiratory phase (191). In the kitten, this increasing pattern of discharge is no longer observed; the onset and offset of the discharge are abrupt, and the Phr discharge recorded at low speed resembles a gasp (268). In the preterm infant and the decerebrate newborn rat, a similar pattern was observed (Fig. 1, B2 and B3). The central mechanisms governing the recruitment of individual Phr-Mns during inspiration seem to differ between the newborn and the adult cat (182, 191, 268).

In the adult cat, high-frequency oscillations have been observed within the Phr burst, as well as within the recurrent, hypoglossal, and medullary respiratory neuron discharges (65, 68, 281). These oscillations occur at a frequency of ~60 Hz in the adult and are present within the neonatal Phr discharge, but their frequency is lower, i.e., ~20 Hz in kittens (268), newborn rats (Fig. 1B3), newborn dogs (448), and even in preterm human infants (Ref. 100 and Fig. 1B2). Because the Phr oscillations are thought to reflect medullary synaptic processes, the neonatal low-frequency oscillations suggest the existence of differences between neonate and adult medullary respiratory networks.

B.  Afferent Regulation of Neonatal Respiratory Activity

1.  Pulmonary afferent regulation

Among the respiratory afferent systems, the most thoroughly studied so far is the vagal pulmonary afferent system, since Hering-Breuer's demonstration in 1868 (178) that lung inflation and deflation inhibits and facilitates inspiration, respectively. Activating the vagal pulmonary stretch receptors by applying lung inflation inhibits the central inspiratory command (3, 358, 483, 484), while bivagotomy slows down the respiratory frequency, mainly via a TI increase, and increases the amplitude of the inspiratory discharge in all adult mammals but humans, where this regulation does not seem to occur (398, 401). In adult cats, the vagal pulmonary afferents play an even more crucial role after pneumotaxic center lesion (66, 67), since after bivagotomy, the normal breathing pattern is replaced by an apneustic breathing pattern characterized by very long inspirations (inspiratory cramps with TI >5 s) interrupted by very short expirations.

At birth, vagal respiratory regulation exists since respiratory rhythm is depressed by vagotomy in kittens (101, 102) and newborn rats (320, 439), but it differs from the adult pattern. Vagotomy performed on decerebrate kittens slows down the respiratory rate, lengthens TI, and reinforces the inspiratory activity, as in adults, but these inspiratory changes are quite moderate (102). The decrease in respiratory frequency is mainly due to lengthening of TE and is associated with a distinct enhancement of expiratory muscle activity (Fig. 2A). Vagotomy in kittens increases TE and TI 10- and 4-fold, respectively, compared with 1.2- and 2-fold, respectively, in adult cats. This paradoxical expiratory facilitation gradually disappears with age, and the adult pattern of response to bivagotomy is reached by about P21 (102, 460). In decerebrate kittens with pneumotaxic center lesion, bivagotomy does not result in the adult apneustic breathing pattern (102, 460) but in very long "active expiratory apneas" (>1 min) interrupted by short inspirations (102). The adult apneustic pattern is not observed until P21. The predominance of expiratory activity in neonates can be said to be part of the general motor behavior, which is mainly characterized by the activation of the flexor muscle group to which the expiratory muscles belong. This is suggested by the predominance of TE versus TI during eupnea, by the flexor activity of expiratory muscles observed during decerebration rigidity, and by the expiratory cramps elicited after vagal and pontine afferent eliminations. Thus removing the upper central structures and vagal afferents makes the expression of the neonatal flexor (expiratory) activity possible.

View larger version (20K): [in this window] [in a new window]   Fig. 2. Respiratory vagal regulation in neonates. In A, effect of vagotomy in decerebrate postnatal day 1 (P1) kitten. Top and bottom traces, integrated and raw EMG discharges, respectively, of third inspiratory interchondral muscle before (A1) and after bivagotomy (A2). Bivagotomy decreases respiratory rate, mainly by increasing expiratory duration, and induces recruitment of previously silent expiratory triangularis sternum muscle of third intercostal space. In B, Hering-Breuer deflation reflex (between arrows) in P2 (B1) and P21 (B2) decerebrate kittens. Top and bottom traces, diaphragmatic EMG discharge and tracheal pressure changes, respectively. Pulmonary deflation activates inspiratory discharges of P2 and P21 neonates, but inspiratory activation is weak on P2.

Overinflating the lungs of neonatal cats and rabbits elicits an inspiratory inhibition (75, 101, 102, 460), and vagal inspiratory inhibition is also observed in vitro in newborn rats (193, 319). In decerebrate kittens, tracheal occlusion performed at the very end of inspiration to maintain the lungs in the inflated status lengthens TE and delays the occurrence of the next inspiration. Conversely, tracheal occlusion at the very end of expiration reinforces the next inspiratory discharge in decerebrate kittens (102) and newborn rabbits (75). These pulmonary responses are stronger in neonates than in adults, since the adult lungs have to be overinflated or overdeflated to elicit any clear-cut respiratory changes. In human infants, the pulmonary vagal regulation of respiration is weak or absent, depending on age. The inflating reflex is weak at 32 wk of postconceptional age; it peaks at 38 wk and declines thereafter (38). It may be absent in full-term babies during the first few days of life (73). Airway occlusion at the end of expiration may strengthen the next inspiration in preterm and term infants (343), but this was not the case in full-term infants during quiet sleep (450). At birth, infants behave no differently from adult humans, in whom the inflating reflex is not present (398).

Although lung deflation induces inspiratory activation, the sustained reinforcement of inspiratory activity observed in adult cats is not present in kittens (Ref. 101 and Fig. 2B1). The adult response appears around P21 (Fig. 2B2). The pulmonary stretch receptor discharges seem to be similar in neonatal and adult cats (424, 269), although few low-threshold receptors are found on P3 (424), but receptors of both types are present on P7 (269). In neonates, the receptors are mainly located in the intrathoracic trachea, and their maximum firing rate is slightly lower than in adults (269). Despite their late myelinization in the vagus nerve (267) and their low sensitivity, the neonatal pulmonary stretch receptors are functional at birth. Therefore, the existence of vagal inspiratory inhibition observed in vivo in neonatal cats, rabbits, and rats (75, 101, 102, 460) compared with the lack of sustained inspiratory reinforcement observed during deflation at birth suggests a functional immaturity of the vagal system of regulation, which may be due to both the neonatal rib cage compliance and an immature central network.

The sneeze reflex is one of the main reflexes triggered by nasal mucosa stimulation. In adult cats, it is characterized by a preparatory inspiratory phase (long and intense inspiratory activity), followed by a brief expiratory burst with a rapid onset (233, 474, 476). In kittens, mechanical stimulation applied to the nasal mucosa elicits markedly different sneeze responses depending on age. At birth, the response consists almost entirely of expiratory efforts, which are longer, occur after longer delays than in adults, and are not preceded by any inspiratory preparation (232, 476). In addition, the aspiration reflex cannot be elicited in newborn kittens, whereas in adult cats, it can be evoked by stimulating the inferior nasal meatus and consists of a sudden increase in inspiratory activity that is not followed by an expiratory effort (473). Both sneeze and aspiration reflexes show the adult pattern by P21. Although the nasal afferents might not be completely mature at birth, their immaturity alone does not explain their ability to activate expiration and their inability to activate inspiration during trigeminal reflexes. Moreover, a study on c-fos expression during the sneeze reflex has provided evidence that the maturation of sneezing depends on the development of the central relays necessary for peripheral inputs to be integrated by the neurons involved in respiratory control (270). Therefore, a lack of central connectivity associated with the immaturity of the central pattern generator seems to be a plausible explanation (475).

Respiratory changes induced by activating buccal trigeminal afferents have been tested in the kitten (473, 224). In anesthetized and decerebrate kittens from P0 to P3-5, applying tiny amounts of water to the mouths induces long apneas (>60 s) and only rarely swallowings. Swallowing coexists with long apneas up to P15 but, on P21, the adult pattern appears, in which swallowing disturbs only one to two respiratory movements. Both electrical stimulation applied to the central end of the lingual nerve and mechanical oral stimulation lead to a strong reinforcement of XII-Mn and expiratory muscle activities during apnea. Serotonin is likely to be involved in the oral trigeminal-induced apnea, since its application to the fourth ventricle in decerebrate kittens induced similar apneas (223, 224), whereas intraperitoneal application of 5-HT antagonist suppressed the oral trigeminal-induced apneas (Duron, unpublished results).

It has been known for a long time that friction applied to infants at birth activates inspiration, and activating cutaneous afferents in newborn kittens greatly facilitates respiration. Nonnociceptive cutaneous stimulation applied to the neck and back reinforces the frequency and amplitude of respiratory movements in decerebrate kittens (266) and preterm infants (100). The fact that the activation of nonnociceptive cutaneous afferents efficiently stimulates inspiration even under anesthesia indicates that a highly potent reflex is involved. After P14, this response becomes very difficult to trigger under anesthesia, but it persists in decerebrate kittens until becoming inconsistent, as in adult cats (266). More intense cutaneous stimulation, consisting for example of pinching the ventral face of the neck, thorax, and abdomen, inhibits inspiration, elicits apnea and bradypnea, and concomitantly activates the XII-Mns and the expiratory muscles (224). This might possibly correlate with asymmetrical effects of the dorsal and ventral cutaneous afferents. Finally, activating cutaneous afferents by applying electrical stimulation to the saphenous nerve of kittens increases the respiratory frequency, and this effect persists for a long time after the end of the stimulation (459). The long-lasting nature of this effect shows the importance of nonspecific systems in the maintenance of neonatal breathing.

	III. MATURATIONAL CHANGES IN RESPIRATORY RHYTHMOGENESIS

Top Previous Next References

In this section, we focus on the mechanisms through which the respiratory network generates a rhythmic command at different ages. We first report in vitro experiments performed on the neonatal rodent respiratory centers, and then those performed on fetuses using similar in vitro approaches, before reviewing the data available on adult mammals, with a special emphasis on rodents. In the adult cat brain stem, three main groups of respiratory neurons have been identified (for review, see Ref. 35), namely, two medullary groups, the ventral and dorsal respiratory groups (VRG and DRG, respectively), and a third pontine group, forming the pneumotaxic center. The DRG and the pneumotaxic center seem likely to play an important role in respiratory rhythmogenesis in the adult cat (35, 66, 67), but whether they are of any functional importance is questionable in the adult rat (149, 193, 284, 496).

A.  Newborn Respiratory Rhythm Generator

1.  In vitro neonatal rodent brain stem-spinal cord preparations

In 1984, Suzue (449) reported that it was possible to record rhythmic bursts in the cervical roots giving rise to Phr nerve, as well as in the XII roots, after isolating the medulla and the cervical spinal cord of neonatal rats in vitro (Fig. 3). These bursts recur in vitro for several hours, but their low frequency (5-10/min), the low TI/TE, and the unusual decreasing pattern of Phr bursts are not in agreement with what was generally recognized at that time as constituting respiratory activity in the adult anesthetized cat (35, 191). In addition, after elimination of pontine and vagal afferents, the neonatal in vitro respiratory network does not produce the apneustic pattern observed in adult cats after bivagotomy and elimination of the pons (35, 66, 67). Despite Suzue's arguments (449) supporting the idea that these in vitro Phr bursts are of respiratory origin, few scientists were convinced at that time that these preparations were actually suitable for respiratory network studies. It has been demonstrated by now, however, that the in vitro Phr bursts 1) originate from the medulla, since they were abolished after upper cervical transection but persist after elimination of the pons (192, 449); 2) induce rhythmic inflation movements of the chest that are accompanied by synchronous diaphragmatic and thoracic EMG discharges (192, 449); 3) occur concomitantly with the firing of neurons located in the VRG area (87, 155, 345-356, and see Fig. 3A); and 4) are sensitive to pH, CO₂ levels (294), and vagal afferent activation (277, 319).

View larger version (28K): [in this window] [in a new window]   Fig. 3. Rostral ventrolateral medulla (RVLM), a crucial area for in vitro neonatal respiratory rhythm. Top left, schematic arrangement showing in vitro brain stem-cervical cord preparation of neonatal rat (ventral surface upward) and in vitro phrenic bursts. Black dot indicates location of RVLM area where inspiratory neurons can be recorded (A, from top to bottom, intracellular activity of a RVLM inspiratory neuron, integrated and raw phrenic discharges). In RVLM, electrical stimulation and drug applications (performed at arrow in BD via a multi-barreled micropipette) affect occurrence of integrated phrenic bursts. A single shock applied in RVLM on-switches immediately a premature phrenic burst (B) while norepinephrine (C) and serotonin (D) applications in RVLM decrease and increase respiratory frequency, respectively.

Numerous scientists are therefore now using Suzue's model or similar models. Three different approaches using different experimental materials can be distinguished: 1) Suzue's isolated in vitro brain stem-spinal cord preparation from neonatal rats (192-194, 283, 285, 293-295, 345-356), mice (180, 181), and even opossums (117, 326, 500); 2) in vitro brain stem slices in which only the most indispensable part of the network is preserved (366-369, 383-385, 436-439); and 3) in vitro-like perfused brain stem of maturing mice and guinea pigs (306-310, 363-365). With all these in vitro approaches, there are several unique advantages, such as the possibility of changing the extracellular medium, directly stimulating, destroying, or recording from different regions, performing patch-clamp recordings under the most favorable conditions, and even using optical imaging procedures (355).

A restricted area of the RVLM was quickly found to be crucial for respiratory rhythmogenesis in vitro (16, 17, 87, 114-116, 345-356, 436-439). This finding was first suggested by electrical stimulation (see Fig. 3B) and lesion experiments. A single electrical shock delivered to the RVLM during mid-expiration can almost induce a similar Phr burst to that which would normally have occurred 2-3 s later. Electrolytic lesions performed in the same area abolished the respiratory rhythmicity, while local drug applications (5-HT, substance P, catecholamine, and high-K⁺ medium) almost immediately induced respiratory frequency changes (89, 115, 116, 437; see Fig. 3, C and D). In the histological checks performed to locate the efficient sites, no well-defined nucleus was observed but a diffuse region extending from the caudal end of the facial nucleus to the rostral end of the nucleus ambiguus. This area is referred to as the RVLM, although other names have been proposed, such as the pre-Bötzinger complex (PBC) (437). Upon applying the neonatal rat in vitro approach to the neonatal mouse, the neonatal mouse respiratory network was found to also remain functional in vitro and electrical stimulation and lesion of the RVLM induced and abolished inspiration, respectively (180).

The patterns of distribution and the activity of the medullary neurons firing in phase with the Phr bursts have been studied in these neonatal preparations. Various research groups have classified the medullary neurons as belonging to several types on the basis of their discharge patterns (16, 17, 345-356, 363-369, 391, 436-439). No consensus has yet been reached as to how many types of respiratory neurons actually exist, but we describe here one of the most thorough attempts to answer this question (345-356).

Neurons that are depolarized and fire several hundreds of milliseconds before the Phr bursts have been called preinspiratory neurons (Pre-I). Some Pre-I neurons fire throughout inspiration (20%), but most of them are silenced for 200-400 ms during the Phr bursts and then fire again for a while at the beginning of expiration. Their silence during inspiration has been attributed to an active inhibition, but releasing the inhibition does not affect their ability to fire rhythmic bursts of spikes (see Fig. 8 of Ref. 350). The firing of Pre-I neurons has been studied under normal artificial cerebrospinal fluid (aCSF) and under modified aCSF containing low Ca²⁺ and high Mg²⁺ (87, 214, 345-348). The modified aCSF was found to be an efficient means of abolishing synaptic transmission based on the suppression of the Phr bursts observed and the silencing or the low tonic frequency of most of the recorded neurons (87, 348). However, one-half of the Pre-I neurons still showed some bursting firing ability under low-Ca²⁺ high-Mg²⁺ aCSF. These neurons, which continue to rhythmically deliver bursts of spikes when the synaptic connections have been blocked, may be inspiratory bursting pacemaker neurons (Fig. 4). When two simultaneously recorded Pre-I are subjected to low-Ca²⁺ high-Mg²⁺ aCSF, they both continue to produce bursting firing, but the bursts of spikes become desynchronized; they are resynchronized when normal aCSF is restored. The Pre-I neurons are preferentially located in the RVLM, where they extend rostrocaudally from the caudal end of the facial nucleus to the rostral end of the nucleus ambiguus. Therefore, the induction of inspiration by RVLM stimulation may result from the massive synchronous activation of the Pre-I bursting pacemaker cells. Among the neurons exhibiting a "classical" inspiratory discharge (non-Pre-I), some have also been classified in the respiratory bursting pacemaker group because they continue to fire in bursts under low-Ca²⁺ high-Mg²⁺ aCSF (87, 214). Intracellular recordings have confirmed that some inspiratory neurons respond to intracellularly injected depolarizing current by increasing the frequency of their bursts of spikes (437). Here again, these neurons are likely to be inspiratory bursting pacemaker neurons playing a major role in respiratory rhythmogenesis within the central respiratory network. The Pre-I pattern of discharge therefore does not seem to be a reliable criterion for determining pacemaker properties, since cells other than Pre-I can also be respiratory pacemaker cells.

View larger version (26K): [in this window] [in a new window]   Fig. 4. RVLM inspiratory bursting pacemaker neuron of in vitro neonatal rat preparations. AC, top and bottom traces: integrated phrenic bursts and extracellular activity of RVLM inspiratory neuron, respectively. Replacing normal artificial cerebrospinal fluid (aCSF) (A) by a low-Ca2+ high-Mg2+ aCSF first depresses (B, 5 min) and thereafter abolishes (C, 20 min) phrenic bursts, but the neuron retains an almost unchanged bursting firing. This neuron was classified as RVLM inspiratory bursting pacemaker neuron.

In addition to these Pre-I neurons, three other subtypes of inspiratory neurons have been distinguished on the basis of the synaptic events occurring during expiration, but "there were intermediate subtypes" (350). The type I neurons receive numerous excitatory postsynaptic potentials before and also after the Phr burst, when the Pre-I cells are firing, and therefore type I neurons might be driven by the Pre-I cells. Some type I neurons start firing well before the Phr bursts, as Pre-I neurons do, but do not stop during inspiration. Type II neurons fire during the Phr burst, and no discernible synaptic events occur just before or after the Phr bursts. The type III neurons also fire during the Phr bursts, but hyperpolarizing events occur just before and after the Phr bursts. Most of the inspiratory neuron types have a decreasing discharge frequency, and their distribution overlaps that of the Pre-I neurons, although they extend more caudally within the VRG. Finally, expiratory neurons have been encountered in the VRG (436, 439) and subdivided in two further groups, which either fire tonically throughout expiration or are activated only late in expiration. These neurons are actively inhibited during Phr bursts, possibly by Pre-I neurons and inspiratory neurons (17, 436, 439).

As mentioned in section III, the cat dorsal medulla contains densely packed inspiratory neurons constituting the DRG (35). Surprisingly, inspiratory neurons have never been said to be encountered so far in the dorsal medulla in any of these in vitro studies, and lesions of the rodent neonatal dorsal medulla do not affect the respiratory rhythmogenesis (193, 366, 367).

The details of the discharge pattern of the respiratory bursting pacemaker neurons (Pre-I type or not) are a relatively subsidiary issue. First, Onimaru et al. (350) recognized that "there were intermediate subtypes" (350). Second, the criteria used for neuronal classification have not been clearly defined (350). The distinction between some type I neurons, which start to fire well before the Phr bursts, and the Pre-I neurons, which also start very early and do not stop firing during inspiration (20% of the Pre-I pool), is not very clear. Finally, the neuronal pattern of discharge is too dependent on the experimental conditions (ionic aCSF composition, the catecholaminergic and serotonergic nuclei preserved at dissection). The inspiratory pause displayed by 80% of the Pre-I neurons results from an active inhibition exerted by caudal medulla that is released when the caudal medulla is lesioned and cooled (349, 350). The Pre-I pattern therefore is unlikely to occur in rhythmic slice preparations including only part of the caudal medulla. Although worthy attempts have been made by several teams to classify the respiratory neurons, we are not convinced that the discrepancies between the results should be emphasized, since they may simply reflect differences between the experimental conditions used.

The interconnections between the various groups of neonatal respiratory neurons have been analyzed by performing cross-correlation analysis. The existence of mono- or paucisynaptic excitatory connections and shared inputs has been suggested by cross-correlograms between pairs of inspiratory neurons (219, 356). With the use of the spike-triggered averaging methods (219, 354, 356), it has been established that Pre-I neurons excite inspiratory neurons and that Pre-I, type I, and type II inspiratory neurons project to the spinal cord, whereas the type III inspiratory neurons may be propriobulbar interneurons (see Ref. 35). When the excitatory amino acids mediating these interactions accumulate as the result of blocked reuptake processes, the respiratory frequency increases via the activation of non-N-methyl-D-aspartate (NMDA) receptors (154).

Some other noteworthy findings in addition to those showing the possible existence of respiratory bursting pacemaker cells within the medullary respiratory network are those demonstrating that Cl-mediated inhibition is not essential to neonatal respiratory rhythm generation (123, 349, 367-369, 384, 427). All these results obtained on in vitro neonatal rodent preparations but also those obtained in rhythmic slices from rather mature mice are consistent with the idea that respiratory rhythmogenesis does not depend on Cl-mediated inhibition within the central respiratory network (386, 388) and that the RVLM respiratory bursting pacemaker neurons may constitute a trigger-clock imposing its rhythm on the rest of the network. However, the functional relevance of inhibition might be more important in vivo than in vitro (388).

From 1970 to 1990, fetal breathing movements were studied in the pregnant ovine. The occurrence of these movements was first suggested by the abdominal movements observed in human and were subsequently confirmed by direct intrauterine observations of the fetus and EMG recordings. Because this question has already been extensively reviewed (165-167, 208, 209, 227, 228, 398-404), only a few in utero studies are mentioned in section IIIB1.

During early gestation, both human and ovine fetuses first produce nonrespiratory movements involving a tonic activation of the thoracic muscles and the diaphragm (208, 209, 403). Fetal breathing movements become frequent but episodic at around one-third of gestation, although fetal breathing movements can be triggered in response to cutaneous stimulation as early as 40 days of gestation in exteriorized sheep fetus. With increasing age, the tonic activity-to-fetal breathing movement time ratio decreases, but the fetal breathing movements are still episodic. On the basis of the use of a window technique (399, 403), it has been observed that fetal breathing movements, general movements, and swallowing activities occur during rapid eye movement (REM) sleep and, conversely, that they are absent during quiet sleep, except for the generalized tonic discharge occurring from time to time; the question as to whether fetal breathing movements occur in states other than REM sleep has been settled with the statement that "wakefulness is not part of fetal life" (400). Fetal breathing movements may be modulated by several endogenous substances as, for example, prostaglandin E₂ (471, 327) and 5-HT (125).

Although fetal respiratory movements were known to occur in utero, little is known about the central command governing these movements, despite some noteworthy efforts in this direction (51, 205). This topic is difficult to investigate because of the inaccessibility of the fetus in utero and its great sensitivity to surgical trauma and anesthesia. In one of the latest reviews published on this subject (209), the authors concluded that "much descriptive work ... has been published, but there is a dearth of information regarding the central respiratory rhythmogenesis in the fetus." Furthermore, it was suggested that "in the future, more direct recording will be required to advance our knowledge significantly" because the few studies published on the respiratory centers have been mainly descriptive and not very informative, even allowing for the fact that they were particularly difficult to perform (51, 205).

Two different groups have simultaneously adapted the in vitro newborn rat brain stem-spinal cord preparation to the fetal rat and have published very similar results (85, 155).

In the fetal rat, brain stem-spinal cord preparations show two different patterns of easily distinguishable rhythmic activity. The first one takes the form of a long-lasting burst of spikes (several seconds) occurring from time to time in all the spinal roots, including the Phr ones. This activity might correspond to the tonic activity previously reported in the fetal lamb. The second type of activity consists of rhythmic bursts of spikes occurring in both the Phr and XII roots with a short duration (<1 s) and a recurrent frequency of 5-10/min. The rhythmic Phr bursts are likely to be respiratory in nature, since they occur concomitantly with respiratory chest movements when the rib cage is kept attached to the spinal cord, and they originate from the medulla since they are abolished by upper cervical transection (85). They are never observed in fetal rat preparations on gestational day 15 (E15), are occasionally seen on E16 (82), and are present in almost all preparations at older ages. In young fetuses, the inspiratory bursts often have weak amplitudes and short durations. Their amplitude and duration increase with age, and by E20, they resemble the neonatal pattern.

In comparison with the in vitro rhythmic activity recorded in newborn rat preparations, the fetal rat respiratory activity is more variable (Fig. 5A). It consists of intermingled short and long cycles including bursts of variable amplitude, although the mean respiratory frequency is in the neonatal range (85, 87, 90). The variability decreases with age from E16 to E20-21, when it almost reaches the stability of the newborn. To estimate the variability, a coefficient of variation has been defined as the SD-to-mean ratio of the cycle duration (or amplitude) measured during 60 successive respiratory cycles. The coefficient of variability ranges between 0.1 and 0.2 in neonatal rats, but it is more than twice as high in E18 fetuses, whereas intermediate values are reached by E20. The fetal rat preparation has been adapted recently to the fetal mouse, and similar results have been obtained, which confirm the stabilization of the fetal respiratory output observed with age (Hilaire, unpublished data). The inability of the isolated central rhythm generator to build up a stable motor command may be viewed as a sign of immaturity, and the stabilization with age of the motor output as a sign of maturation (57, 85, 87, 90).

View larger version (26K): [in this window] [in a new window]   Fig. 5. Variability as a characteristic of in vitro fetal mouse respiratory activity. Top: schematic arrangement showing procedures for studying in vitro fetal and neonatal rodent respiratory activities. A and B: integrated phrenic bursts recorded in E18 fetal and P2 neonatal preparations, respectively. Histograms show distribution of amplitude of phrenic bursts (expressed as percent of mean) and respiratory cycle duration (in s) for 60 successive cycles. Fetal activity is highly variable while neonatal activity is stable. (From G. Hilaire and C. Bou, unpublished data.)

In E20 fetal rat preparations, as in newborn preparations, single electrical shock stimulation applied to the RVLM results almost immediately in an unexpected Phr burst with a similar shape and amplitude to that which would normally have occurred later. This response has never been observed in E18 fetuses, however, where RVLM stimulation either has no effect or elicits a diffuse facilitation when large stimulating currents are applied (87). Electrolytic lesions performed in the RVLM abolishes respiratory rhythmogenesis in E20-21 and newborn preparations but has no effect in young E18 fetuses. Upon exploring the RVLM area with extracellular electrodes, respiratory neurons were observed at all developmental ages. When a low-Ca²⁺ high-Mg²⁺ aCSF was used to block the synaptic connections within the preparation, the rhythmic Phr bursts disappeared, and most of the RVLM respiratory neurons lost their bursting behavior. In fetal E20 preparations displaying a fairly stable rhythm, ~10% of the RVLM neurons retained a rhythmic bursting activity under low-Ca²⁺ high-Mg²⁺ aCSF (Fig. 6C). These neurons may be respiratory bursting pacemakers, likely those mentioned above in newborn rats. Surprisingly, none of these neurons has been encountered at E18, when the rhythm is unstable. It has therefore been postulated that the variable respiratory rhythm present in the young fetus may reflect an immature network, whereas in late fetal and newborn preparations, the pacemaker properties observed in some RVLM respiratory neurons may be at least partly responsible for the stable activity of the network (85, 87, 90). The variable E18 fetal rhythm may either result from the wiring properties of the medullary network or be imposed by some other neuronal networks, such as that described in the chick embryo (61, 127, 128).

View larger version (20K): [in this window] [in a new window]   Fig. 6. Fetal E18 respiratory neurons and E20 inspiratory bursting pacemaker neuron. In AC, top and bottom traces refer to extracellular activity of respiratory neurons and integrated phrenic bursts, respectively. A and B: recordings of two different medullary inspiratory neurons in same in vitro E18 fetal rat preparation; at this developmental age, some medullary inspiratory neurons may stop firing during a phrenic burst (A) and fire without a concomitant phrenic burst (B). Respiratory variability illustrated in Fig. 5 may therefore be of central origin. C: RVLM inspiratory neuron in an in vitro E20 fetal rat preparation; phrenic bursts are abolished when replacing normal aCSF (C1) by a low-Ca2+ high-Mg2+ aCSF (C2; 17 min), but inspiratory neuron continues to burst. This type of inspiratory bursting pacemaker neuron was not found in E18 fetal rat preparations.

For a long time, what was known about the adult medullary respiratory centers was based on data obtained in the cat, and only recently, data have been published on the adult rodent respiratory network. The numerous studies performed during the last few decades on the adult cat medullary network have been reviewed recently (35). Tremendous efforts have been made to locate and classify the respiratory neurons. The DRG, which is located in the dorsal part of the brain stem, contains mainly inspiratory neurons projecting toward the spinal cord, where they excite the Phr-Mns (for reviews, see Refs. 190, 291). The VRG, which is located in the vicinity of the nucleus ambiguus, is a more heterogeneous structure containing both inspiratory and expiratory cells, which have been identified as cranial Mns, bulbospinal neurons, and central interneurons. Several models have been proposed to explain the respiratory rhythmogenesis, and all are based entirely on the excitatory and the inhibitory connections within the network. The authors of one of the latest reviews on the subject (35) concluded that "brain stem respiratory activity results from the sequential activation of at least six neuronal populations, leading to two fast phase-switching transitions and three relatively long respiratory phases. Each process is conditioned by the previous one and initiates the next." The possible role of the RVLM in respiratory rhythmogenesis is described in a subchapter entitled "Other structures," beginning by "Medullary structures other than the DRG and VRG have been proposed as playing a role in the generation and control of breathing." By now, a consensus has almost been reached as to the crucial role played by the RVLM (30, 96, 122, 143, 354, 394, 396, 410-412).

The first reports on the adult rat respiratory network suggested that the medullary respiratory neurons might be located in the VRG and DRG groups (202, 413). The existence of the DRG has been questioned, however (118, 344, 491), and the results of both in vitro neonatal brain stem preparations and in vivo adult experiments have shown that it may not in fact exist in the rat (193). Intracellular soundings performed in the dorsomedial medulla show the existence of respiratory activity originating from axons but not somas (496, 497). Although respiratory activities have again been reported in the rat dorsal medulla (76, 457), the existence of the DRG in rats seems unlikely, and, if a rat DRG exists, it will certainly be of less functional significance than in the cat (457). In the mouse, the existence of a DRG seems also unlikely, at least in vitro, since ablation of the dorsal part of the medulla had no effect on respiratory activity. The data on newborn and adult rodents obtained in vivo and in vitro are therefore consistent with the idea that the DRG contributes little if at all to respiratory rhythmogenesis in these mammals.

Within the VRG, all the neuronal types previously encountered in the case of the cat were found to be present in the rat, namely, cranial Mns, respiratory bulbospinal neurons, and central interneurons (107-111, 496-498). Here again, several attempts have been made to classify these neurons in terms of the location of their soma, their pattern of discharge, and their axonal projection. The axons of the inspiratory bulbospinal neurons project to either the ipsilateral side (22/28), the contralateral side (16/28), or both (10/15; Ref. 456). Studies using cross-correlation techniques have confirmed that a substantial bilateral excitation is exerted on the Phr-Mns (95) by VRG bulbospinal neurons and have suggested that the drive to the intercostal Mns may be relayed by interneurons (455, 456). Anatomic results have been published that are in complete agreement with these electrophysiological findings. Injecting different tracers into both left and right Phr nuclei leads to the double labeling of a substantial number of bulbospinal neurons in the rostral part of the nucleus ambiguus, but only a few in the dorsal part of the medulla, while single-labeled neurons are distributed throughout the rostrocaudal extent of the medulla (376). The bulbospinal neurons are located more ventrolaterally than the laryngeal Mns in the rostral part of the nucleus ambiguus, and they are located around the laryngeal Mns at the more caudal level of the nucleus (377, 378). Propriobulbar, bulbospinal, and vagopharyngeal neurons are unevenly distributed along the rostrocaudal axis of the VRG (333, 334). The brain stem afferent connections to VRG neurons and the possible interconnections within the respiratory centers have been also studied using neuroanatomic approaches (297, 332). Pontine efferent projections from the VRG reached the perifacial areas, the parabrachial and Kolliker-Fuse nuclei, as well as the catecholaminergic A5 area (146).

The question now arises whether the RVLM plays the same crucial role in adult rodent respiratory rhythmogenesis. In an in vitro-like perfused mouse brain stem preparation (363-365), the activity of respiratory neurons displaying the Pre-I type firing was recorded in the RVLM. These neurons started to fire before the beginning of inspiration, continued to fire bursts of spikes when the synaptic connections had been blocked, and fired rhythmically when all the other respiratory neurons had been silenced (central apnea induced by stimulating C vagal fibers). The firing behavior of Pre-I neurons was assessed after switching to low-Ca²⁺ high-Mg²⁺ perfusate to block the synaptic transmission. In the absence of all synaptic transmission, seven of eight Pre-I neurons continued to discharge rhythmically and four of eight still presented a bursting activity 25 min after the synaptic blockade. Unfortunately, the single Pre-I bursting discharge shown during synaptic blockade (Fig. 9 from Ref. 365) was not very convincing, since the Pre-I pattern of discharge that recurred under synaptic blockade is very different from the control pattern (burst duration 5-10 s during synaptic blockade versus <1 s during control conditions). It may therefore be safer to wait for further confirmation of the existence of respiratory bursting pacemaker neurons in adult mice rather than making any hasty conclusions on this point. Nevertheless, these data suggest that respiratory bursting pacemaker neurons may exist in the mature mouse, although this does not rule out the possibility that other circuitries and mechanisms may be involved in the adult respiratory rhythmogenesis.

On the one hand, it has been reported that strychnine does not block the respiratory rhythm in neonatal mice until P15, but amino acid inhibition becomes essential later however (175, 366-369). On the other hand, data obtained on brain stem transverse slices from P0 to P22 mice (383-385) have shown that inhibitory synaptic events are not essential to respiratory rhythmogenesis at birth. Their functional relevance increases with age, but they never become essential to respiratory rhythmogenesis. Blocking glycine receptors by strychnine did not stop rhythmicity but increased the burst frequency at all ages. Strychnine applied at concentrations between 1 and 50 µM did not abolish the respiratory rhythm in either neonatal and adult mice, although the expiratory neuronal activity was profoundly disturbed in adult slices. Although the inspiratory neuronal pattern of firing remained phasic, the "expiratory neurones became almost tonic, suggesting that these neurones are not essential for generating rhythmic inspiratory activity" (384). The discrepancies existing between the two rodent approaches concerning the role of inhibitory synaptic events in respiratory rhythmogenesis may be due to either technical differences (tilted versus transverse slices) or the actual developmental age. -Aminobutyric acid innervation in the rat nucleus ambiguus increases from the fetal to the early postnatal age and then decreases (134, 135, 488). In humans, a similar pattern of development has been observed in the cerebrospinal fluid GABA levels, which are correlated with the conceptual age but not with the birth age (148). The role of inhibitory mechanisms in adult respiratory rhythmogenesis is still an open question, but these mechanisms do participate in shaping the respiratory neuron discharge pattern (421).

The authors of in vivo studies have reported that inspiration lengthens with age, whereas the duration of the in vitro inspiratory bursts has been said to remain constant or to increase from birth to P15 (366-369, 383-385). In vivo, the inspiratory duration may be affected by inputs from both the lungs and the pneumotaxic center (66, 67). Upon suppressing both inputs in anesthetized cats, the eupneic pattern is replaced by an apneustic pattern characterized by very long inspirations interrupted by short expirations. The cat pontine mechanisms involve NMDA receptors, since an apneustic pattern is induced by the NMDA channel blocker MK-801 (131, 423) and a high density of NMDA binding sites is reported in the pneumotaxic center (415). In the anesthetized, vagotomized adult rat, the existence of apneustic breathing has been questioned, since, on the one hand, MK-801 lengthens inspiration and only occasionally induces short periods of apneusis (285), but on the other hand, local application of MK-801 to the pons induces apneusis (142). MK-801-induced apneustic breathing has been reported to occur in vagotomized Sprague-Dawley rats but not in Wistar rats (70). There seem to exist both interspecies differences and a problem as to how apneusis should be defined: inspirations are characteristically long (~5 s) during apneusis in cats (132) and guinea pigs (56) but are rather short during the so-called apneustic periods in rats and mice (~1.9, 0.5, 1.5, and 1.6 s in Swiss mice, BALB/c mice, Wistar rats, and Sprague-Dawley rats, respectively, from Table 1 of Ref. 56). In addition, apneusis can occur under anesthesia, but during wakefulness, an eupneic pattern is maintained (56), and apneusis can be elicited in the adult guinea pig under in vivo but not in vitro conditions (306). It is worth noting that the possibility of evoking apneustic breathing in kittens may be questioned (460, 102, 423). In neonatal and fetal rodents, the possibility that pontine NMDA receptors may be involved in the offset of inspiratory seems unlikely, since 1) the in vitro medullary network generates a respiratory rhythm in the absence of pulmonary and pontine afferents, 2) applying MK-801 to neonatal rat brain stem and medullary slices does not induce apneusis (145, 154), and 3) transgenic mice lacking functional NMDA receptors show a respiratory activity in vitro (121, 144). The NMDA receptors may therefore contribute to the inspiratory offset under some circumstances (anesthetized and vagotomized adults) but are unlikely to be a crucial factor from this point of view.

Many in vivo results obtained on the adult cat and published in the past can now be viewed as suggesting the possible involvement of the RVLM in adult cat respiratory rhythmicity. The first transection experiments on adult cats showed that eliminating the rostral medulla abolishes rhythmic respiratory movements (478, 258). In the 1970s, respiratory changes induced at the ventral surface of the rostral medulla were taken to originate only from the chemoceptive areas (256, 282, 420). At the beginning of the 1980s, unilateral transections or reversible procaine blocks in the vicinity of the facial nuclei (the rostral limit of the RVLM) were found to alter the respiratory rhythm (157, 158). Unilateral cooling of the RVLM to 20°C completely abolished inspiratory rhythmicity, which gave rise to the idea of calling the ventral medullary area extending from the caudal part of the facial nucleus to the rostral end of the nucleus ambiguus the "apnea region" (45).

More recently, the cat RVLM was found to contain respiratory neurons that can be activated from both the DRG and VRG; this finding suggests that "RVLM neurons are component of medullary loops operating in the control of breathing" (40). In this area, pharmacological and electrical stimulation both induce respiratory changes (41). The patterns of respiratory neuronal activity have been examined in a circumscribed area of the cat RVLM, defined as the PBC and located caudal to the retrofacial nucleus and rostral to the lateral reticular nucleus (69, 422). This area appears to be a specialized subregion of the adult cat VRG corresponding to the neonatal rat RVLM. It has been proposed that the PBC remains essential to rhythm generation throughout postnatal development, since blocking the synaptic transmission in the PBC of adult cats induced central apnea (387, 396). In the adult cat, recordings from the activity of inspiratory neurons located in the PBC which displayed a Pre-I pattern of discharge showed that these neurons project to the contralateral VRG and are involved in vestibular and superior laryngeal nerve reflexes as well as in swallowing and vomiting (499). Inhibitory mechanisms play an important role in respiratory rhythmicity in the adult cat (421) and bicuculline and strychnine applied within the PBC drastically affected rhythmic Phr activity (373). These data again argue in favor of the PBC being a crucial region for respiratory rhythm generation, but also indicate that inhibitory mechanisms play an important, if not essential, role in respiratory rhythmogenesis in the adult cat in vivo.

	IV. MATURATION OF THE RESPIRATORY MOTONEURONS

Top Previous Next References

The fact that fetal respiratory movements occur early in prenatal life means that the respiratory Mns show a rhythmic activity long before birth. On the one hand, this early sustained activity may affect their development. On the other hand, they are bound to encounter the same maturational problems as other Mns, namely, cell migration, cell death, as well as morphological and electrophysiological maturation. Mouse XII-, vagal- and Phr-Mns differ in their homeodomain protein expression, and their fate is partly defined by the Pax6 homeobox gene that is expressed by undifferentiated cells in the ventral region of the neural tube (113).

The tongue contains intrinsic and extrinsic muscles that serve complex motor functions such as chewing, suckling, swallowing, coughing, vomiting, vocalizing, and breathing. The intrinsic muscles, which are devoid of bony attachments, constitute the body of the tongue. The extrinsic muscles, which are attached to bones, act as either tongue protruders (genioglossus) or retractors (hyoglossus and styloglossus). The Mns innervating tongue muscles are located in the dorsal part of the brain stem, within the XII nucleus along with genioglossal Mns located in the XII ventral subnucleus and hyoglossal and styloglossal Mns located in the dorsal subnucleus. Transynaptic tracing studies on adult rats have shown that the protrudor and retractor Mns do not have the same central drivers but share a common drive from the raphe nuclei (92).

Aside from the tongue, the remaining muscles in the upper respiratory tracts (larynx, pharynx oral and nasal cavities) are innervated by Mns located in the nucleus ambiguus. A bilateral location of the laryngeal Mns was first suggested (255, 362), but this has not been confirmed. Contrary to what occurs in the cat and the rabbit, the rat Mns innervating the cricothyroid and the posterior cricoarytenoid muscles are not arranged in distinct subgroups (377), but the Mns innervating the oropharyngeal muscles have a viscerotopic pattern of organization (334).

The complete pattern of adult rat XII nucleus organization is present quite early in development, by P2 (440). Most of the XII-Mns are generated largely at E10-11 (11, 12) and are present in a well-defined nucleus by E15. Mutation in Pax6 homeobox gene profoundly perturbs XII-Mn maturation and results in an apparent transformation of XII-Mns into vagal Mns (113). Cell death, which is a widespread developmental phenomenon (177, 204, 339), occurs prenatally in the case of XII-Mns, with a loss of 35% of the Mns between E16 and E19 and no significant decrease in the Mn number subsequently, which remains around 3,300 at P1 (135). Studies on the postnatal development of XII-Mns have shown that they undergo considerable changes in the dendritic organization but that the somata seem to remain quite unaffected (33, 328-331, 336, 463, 464). At birth, the dendritic tree is well extended, and its branching complexity decreases (loss of terminal branches) concomitantly with the lengthening of the remaining branches observed during the two first postnatal weeks, but the overall area of the XII-Mns (soma and dendritic trees) does not change. New dendritic branches then regenerate during the third week, while the diameter of all the existing branches increases, so that the dendritic area increases twofold.

The development of the nucleus ambiguus Mns has also been studied in the fetal rat (12, 134). By E14, a peak has been observed in the neuroblast number. They are densely clustered in the dorsal brain stem, and the ventrolateral brain stem shows no apparent nucleus ambiguus. The cells then migrate out of the proliferative zone into discrete, less dense neuronal centers, and an ill-defined compact division of the nucleus ambiguus can be seen by E15. By E17, the nucleus ambiguus is fully formed and contains around 2,700 Mns. Their number decreases significantly by E19 and E20 (2,000 and 1,600, respectively) up to P1 (~1,300), when the adult mean number is reached (1,400).

Cell death therefore occurs in the respiratory cranial Mns only during prenatal life, with no further postnatal loss in tongue, larynx, and pharynx Mns. The fate of the facial Mns is different. Their number decreases drastically from 5,000 at E15 to 3,000 at P1 and continues to decline to 2,000 after birth until P10 (19). Cell death occurs at a higher rate (~70%) and continues longer (up to P10) among the facial Mns than among the XII and nucleus ambiguus Mns. As far as the spinal Mns are concerned, conflicting results have been published, and cell death is said either to occur only at the early fetal ages or to continue during the first postnatal week (245, 337, 357, 406).

During the first two weeks of life, the electrophysiological properties of XII-Mns undergo considerable changes: the input membrane resistance and membrane time constant decrease by ~50%, whereas the rheobase current increases by 100% (328-330, 335, 464). Although it might be expected that the electrophysiological changes originate from developmental morphological changes, it has been reported that the total surface area of XII-Mns does not increase significantly during this period (328) and that no strong correlation exists between membrane resistance and total surface area (331). It has therefore been hypothesized that the electrophysiological changes may reflect synaptic input changes (330, 335).

Anatomic and electrotonic couplings between XII-Mns exist before P10 and seem to subsequently disappear after (273). Because electrotonic coupling may contribute to the synchronization between Mns, the loss of coupling beyond P18 may at least partly explain why the XII burst shape changes from a decreasing one at birth to a bell-shaped one in older mice (384). Several currents, such as current and hyperpolarizing-activated inward current (I_h), Ca²⁺ have been studied in neonatal XII-Mns (360, 461-463). The I_h found in neonatal XII-Mns, based on the occurrence of a depolarizing sag during the injection of hyperpolarizing current, seems to be three times smaller in neonatal than in the adult XII-Mns, and this may contribute to the XII pattern of discharge.

In isolated in vitro perinatal brain stem-spinal cord preparations and neonatal brain stem slices, XII-Mns are rhythmically depolarized and fire during inspiration. In vitro, as in vivo, the XII inspiratory bursts occur before the Phr bursts (304), and the in vitro XII burst shape often decreases during inspiration, like that of the in vivo cat (428). In mouse transverse medulla slices, also rhythmic respiratory discharges continued in XII-Mns and medullary neurons (383-385). During the first 3 wk of life, the cycle length and the XII burst duration do not change significantly, but the XII burst shape changes from a decreasing one (P1P9) to a bell-shaped one (P10P22). Furthermore, the strength of the coupling between XII and medullary cell bursts decreases from younger mice (P0P4: ratio 1/1) to older mice (P5P18; ratio 1/3-4). In perfused guinea pig brain stems (307, 310), the above results reported on mice were not confirmed; the XII burst shape could be either decreasing, increasing, or bell shaped, and it was not specified whether uncoupling between XII and medullary respiratory cells occurred. This was never observed in neonatal rat preparations.

Glutamate excitatory effects are mediated via two broad classes of ionotrophic receptors, NMDA (slow kinetics, highly permeable to Ca²⁺) and non-NMDA (fast kinetics, weakly permeable to Ca²⁺) receptors, and these two types of receptors are colocalized at the same synaptic locations on the XII-Mn membranes, but some sites contain only NMDA receptors (338). Non-NMDA receptors mediate the transmission of the respiratory drive to XII-Mns (145). Several data suggest that XII and Phr-Mns are not driven by the same medullary respiratory neurons (298, 304, 428). The XII central drive may originate from respiratory pre-Mns located in the reticular formation lateral to the XII nucleus (91, 92), where electrical stimulation elicits short-latency glutamatergic excitatory postsynaptic potentials in XII-Mns, the amplitude of which may be modulated by adenosine and 5-HT (29, 433, 434). However, no respiratory active neurons have yet been detected in this area. In the neonatal rat and the adult cat, the inspiratory XII-Mns are silent during expiration, and their repolarization is a passive process, involving no inhibitory effects during expiration (486). -Aminobutyric acid, a key inhibitory neurotransmitter in the central nervous system, operates via two classes of receptors, GABA_A and GABA_B. In the E16 fetal rat, a very small number of GABA-positive fibers is present into both the XII nucleus and nucleus ambiguus. By E20, a dramatic increase in GABAergic innervation occurs (134, 135). At birth, GABA_A receptors are present in most brain areas, especially in brain stem regions that rely more on GABA_A receptors in early postnatal life than at more mature stages; their density increases from birth to P5, peaks at P10, and then decreases (488). In infants, the GABA levels in the cerebrospinal fluid show a comparable pattern during development (148). In the adult rat, the lack of effect of the GABA_B antagonist on XII-Mn activity suggests that no permanent GABA_B inhibition is exerted directly on XII-Mns, although the receptor is present (342).

The diaphragm is the main inspiratory muscle, and the maturational changes in its structural and functional organization have been studied in cats (429-431), rats (262), and baboons (272, 274). The premature and full-term baboon diaphragm contains a larger number of highly oxidative fibers, whereas the adult diaphragm is composed of a mixture of oxidative (type I) and glycolytic (type IIB) fibers (274). The neonatal diaphragm contracts and relaxes more slowly, develops less force, and is more resistant to fatigue than in adults (274).

In the fetal rat, the diaphragm is fully innervated by E18, and at birth, ~80% of the Phr axons are already myelinated (267). Phrenic motoneurons are generated in the ventricular zone by E11-12, while the cervical roots leading to the Phr nerve start to be formed (12). A leading group of "pioneering" Phr axons migrates, reaching the diaphragm by E13. The diaphragm and the growing Phr nerve descend together toward the thoracic level by E14, and intramuscular branching commences clearly correlated with myotube formation, which indicates the existence of intimate regulatory signaling mechanisms between the developing muscle and nerve (9, 10). A Pax6 homeobox gene mutation induces a 45% decrease in the total number of Mns at C₃-C₄ cervical segments by E12 which encompasses the Phr-Mns and those in the median column (113). The morphological changes undergone by Phr-Mns with development have been studied by performing retrograde labeling with carbocyanine dye, DiI, from E13E14 to birth (10