Different RRNs fire at different times during inspiration and expiration
Eupneic breathing is highly stereotyped and consists of two major phases—inspiration and expiration (Fig. 32-5A, B). During the inspiratory phase, phrenic nerve output to the diaphragm gradually increases in activity during 0.5 to 2 seconds and then declines precipitously at the onset of expiration (see Fig. 32-5C). The ramp increase in activity helps to ensure a smooth increase in lung volume. During the expiratory phase, the phrenic nerve is inactive, except—in some cases—for a brief burst at the onset. Some physiologists consider this “postinspiratory” burst to represent its own phase.
FIGURE 32-5 Neural activity during the respiratory cycle. The activity of RRNs in the medulla (examples of which are shown in D and E) leads to the phasic activity of the phrenic nerve (C) and other respiratory nerves, which produces airflow (B), causing lung volume to change (A). ENG, electroneurogram; Exp, expiration; FRC, functional residual capacity; Insp, inspiration; TV, tidal volume; Vm, membrane potential. N32-9
Underlying the activity of the phrenic nerve—and the other motor nerves supplying the muscles of inspiration and expiration—is a spectrum of firing patterns of RRNs located within the DRG and VRG (see above). RRNs can be broadly classified as inspiratory or expiratory, but each class includes many subtypes, based on how their firing patterns correlate with the respiratory cycle. Figure 32-5D and E shows two such patterns. N32-9 Each subtype presumably plays a unique role in generating and shaping respiratory output—that is, the activity of the nerves to each respiratory muscle. RRNs may be further subclassified on the basis of their responses to afferent inputs, such as lung inflation and changes in arterial .
Neural Activity During the Respiratory Cycle
Contributed by George Richerson, Emile Boulpaep, Walter Boron
Some RRNs fire mostly during early inspiration, others evenly throughout in inspiration, and still others during late expiration or evenly throughout expiration. Some fire during both inspiration and expiration, but with a peak of activity during one of the two phases of ventilation.
eFigure 32-1 is an expansion of Figure 32-5. eFigure 32-1 includes four additional panels: E, F, G, and I. Panel E in Figure 32-5 is the same as panel H in eFigure 32-1.
EFIGURE 32-1 Neural activity during the respiratory cycle. The activity of RRNs in the medulla (examples of which are shown in D to I) leads to the phasic activity of the phrenic nerve (C) and other respiratory nerves, which produces airflow (B), causing lung volume to change (A). ENG, electroneurogram; Exp, expiration; FRC, functional residual capacity; Insp, inspiration; TV, tidal volume; Vm, membrane potential.
As shown in Figure 32-5C—as well as eFigure 32-1C—the activity of the phrenic nerve varies in a characteristic way during the phases of the respiratory cycle. During the inspiratory phase, phrenic nerve output to the diaphragm gradually increases in activity over 0.5 to 2 seconds, followed by a precipitous decline at the onset of expiration. The ramp increase in activity helps to ensure a smooth increase in lung volume. Although not evident in eFigure 32-1C, during the first expiratory phase (E1, or the postinspiratory phase), there may be a paradoxical burst of activity in the phrenic nerve. This burst offsets the large elastic recoil of the lungs at the end of inspiration. This elastic recoil would tend to collapse the lungs rapidly if the tone of the inspiratory muscles decreased too abruptly.
During the second expiratory phase (E2), the phrenic nerve is inactive. During a quiet expiration, which is normally a passive event, the accessory muscles of expiration are also inactive. However, during a forced expiration, the accessory muscles of expiration (e.g., internal intercostals, abdominal muscles) become active during E2.
Underlying the activity of the phrenic nerve—and the other motor nerves supplying the muscles of inspiration and expiration—is a spectrum of firing patterns of different RRNs throughout the medulla, some of which are shown in eFigure 32-1D–I. RRNs can be broadly classified as inspiratory or expiratory, but for each class there are many subtypes of RRNs, each presumably with different functions in generating and shaping the “respiratory output”—that is, the activity of the nerves to each of the respiratory muscles.
The tracings in eFigure 32-1 are idealized drawings. These recordings would have been made in a preparation such as the brain of a cat with an extracellular electrode recording the activity of a particular RRN. A cuff electrode on the phrenic nerve would have simultaneously recorded the activity of the phrenic nerve. All the panels in eFigure 32-1 relate to the condition of eupnea. In the following descriptions, the letters refer to the panel designations:
A. Lung volume. Note that inspiration is shorter than expiration.
B. Airflow. Airflow into the lung is represented by a downward deflection in the record. Note that the air flow is zero when lung volume is not changing (e.g., at the transition from inspiration to expiration).
C. Phrenic-nerve activity. The panel shows a typical waveform of phrenic nerve output during eupnea.
D. An RRN with inspiratory-ramp activity. Note that the neuron fires during inspiration with a gradually increasing firing rate.
E. An RRN with early-burst activity. The neuron fires during early inspiration and then with a decrementing firing rate.
F. An RRN with constant inspiratory activity. The neuron fires with a relatively constant rate during inspiration.
G. An RRN with late-onset inspiratory activity. The neuron is silent except for a short burst late in inspiration. This neuron could be part of an “off-switch” that terminates inspiration.
H. An RRN with early-expiratory activity. The neuron fires during E1, immediately after inspiration (postinspiratory), and then the firing rate decrements.
I. An RRN with expiratory-ramp activity. The firing rate gradually increases during E2.
The various groups of respiratory neurons in the brainstem consist of networks of neurons with firing patterns such as those outlined in the previous passage. Although the precise role each type of RRN plays in generating and shaping the respiratory motor output is unknown, it is instructive to consider some of the neurons that make up, for example, the DRG. In the DRG, most RRNs (>90%) are inspiratory, and three of the most common of these are known as Iα, Iβ, and P neurons.
Iβ neurons (the I stands for “inspiratory”) fire with a ramp pattern (see eFig. 32-1D), presumably because they are driven by the central inspiratory activity that drives inspiration. In addition, they receive sensory input directly from pulmonary stretch receptors; lung inflation stimulates Iβ neurons. Many Iβ neurons do not send their axons to the spinal cord, but instead are interneurons whose axons synapse locally within the DRG. Thus, Iβ neurons probably integrate sensory information from pulmonary stretch receptors and may play a key role in some respiratory reflexes, such as the Hering-Breuer reflex (see p. 706).
P neurons, unlike the Iβ neurons, are not driven by the central inspiratory activity. However, they do receive direct input from pulmonary stretch receptors, so that their activity simply tracks changes in lung volume. Thus, like the Iβ neurons, the P neurons presumably are important for sensory integration.
Iα neurons, like Iβ neurons, fire with a ramp pattern (see eFig. 32-1D). Input from other DRG neurons tends to inhibit Iα neurons with increases in lung inflation. Iα neurons are premotor neurons; their axons synapse on phrenic motor neurons and external intercostal motor neurons. That is, the firing of an Iα neuron indirectly tends to cause a muscle of inspiration to contract. Thus, it is not surprising that the firing pattern of the Iα neurons is similar to that of the phrenic nerve. Collaterals of Iα neurons also branch off and synapse on neurons in the VRG.
The firing patterns of RRNs depend on the ion channels in their membranes and the synaptic inputs they receive
What are the mechanisms for generating so many types of activity in RRNs? For example, if the RRN is a premotor neuron, its firing pattern must be appropriate for driving a motor neuron, such as one in the phrenic nerve nucleus. Two complementary mechanisms appear to contribute to the firing patterns necessary for the neuron to do its job. (1) The intrinsic membrane properties of RRNs—the complement and distribution of ion channels present in a neuron—influence the firing pattern of that neuron. (2) The synaptic input—excitatory postsynaptic potentials (EPSPs) and inhibitory postsynaptic potentials (IPSPs)—changes with an appropriate pattern during the respiratory cycle and thereby generates a specific firing pattern.
Intrinsic Membrane Properties
Many neurons in respiratory nuclei have intrinsic membrane properties that influence the types of firing patterns they are able to produce. For example, many DRG neurons have a K+ current called a transient A-type current (see p. 193). If we first hyperpolarize and then depolarize such a neuron, it begins firing action potentials, but only after a delay. The hyperpolarization removes the inactivation of the A-type current, and the subsequent depolarization transiently activates the A-type K+ current and transiently slows depolarization of the membrane and inhibits generation of action potentials (see Fig. 7-18C). If the A-type current is large, the neuron cannot begin to fire until after the A-type current sufficiently inactivates (see Fig. 7-18D). The delay in firing of a neuron with A-type current can explain why some RRNs fire only late during inspiration, even though they receive EPSPs continuously during inspiration. As we will see, other neurons have pacemaker properties due to their complement of ion channels, which allows them to fire action bursts of potentials spontaneously without synaptic input. N32-10
Dorsal Respiratory Group Neuron with Ca2+-Activated K+ Current
Contributed by George Richerson, Emile Boulpaep, Walter Boron
A second type of neuron in the DRG has a large amount of Ca2+-activated K+ current (see p. 196). After the neuron fires several action potentials, the accumulation of Ca2+ inside the cell slowly activates this KCachannel, hyperpolarizing the cell and limiting the firing rate. Thus, suddenly depolarizing such a neuron triggers a burst of firing that tapers off as KCa channels activate. The decrease in firing rate, called spike-frequency adaptation, mimics the early-burst activity of some RRNs. N32-9 Like the A-type current, KCa channels could cause these neurons to have an appropriate firing pattern, even in the absence of rigidly patterned synaptic input.
The most obvious explanation for a neuron's having a specific firing pattern is that it receives excitatory synaptic input when it is supposed to fire action potentials and receives inhibitory synaptic input when it is quiet. Indeed, some RRNs fire only during early inspiration (see Fig. 32-5E), when they receive strong excitatory synaptic input. These early inspiratory neurons also inhibit late-onset inspiratory neurons, and vice versa (Fig. 32-6). N32-11 As a result of this reciprocal inhibition, only one of the two subtypes of inspiratory neurons can be maximally active at a time.
FIGURE 32-6 Patterned synaptic input: reciprocal inhibition. Because of reciprocal inhibition between the early-burst neuron and the late-onset inspiratory neuron, only one can be maximally active at a time. Vm, membrane potential.
Contributed by George Richerson
Independent of intrinsic membrane properties, the patterned synaptic input that many RRNs receive from other respiratory neurons can also influence the firing pattern. For example, the early-burst activity of the neuron whose firing pattern is shown in the upper panel of Figure 32-6 parallels the strong excitatory synaptic input that the neuron receives during the early part of inspiration as well as the inhibitory synaptic input that it receives during expiration.
Some firing patterns of RRNs result in part from inhibitory connections between neurons. For example, the early-burst inspiratory neurons described in the preceding paragraph appear to inhibit late-onset inspiratory neurons (N32-9, specifically eFig. 32-1G), and vice versa. As a result of this reciprocal inhibition, only one of the two subtypes of inspiratory neurons can be maximally active at any one time.
In addition to patterned synaptic input from RRNs, respiratory neurons also receive input from other neuronal systems, which allows the respiratory system to respond appropriately to challenges and to be integrated with many different brain functions. Some of the neurotransmitters released onto respiratory neurons—both by RRNs and by neurons from other regions of the CNS—are listed in eTable 32-1.
Neurotransmitters Released onto Respiratory Neurons
Gamma aminobutyric acid (GABA)
Serotonin (5-hydroxytryptamine, 5-HT)
Thyrotropin-releasing hormone (TRH)
Corticotropin-releasing hormone (CRH)
In summary, both intrinsic membrane properties and patterned synaptic input can cause RRNs to fire in their characteristic patterns. In theory, either of these properties might be sufficient to ensure that these neurons have the right firing pattern. The presence of both properties illustrates a general concept that appears to be important in critical biologic systems: redundancy. For those body systems critical for life support, use of multiple mechanisms may ensure that normal functioning will occur despite severe perturbations.
In addition to receiving synaptic input from RRNs that occurs rhythmically, in phase with breathing, respiratory neurons also receive input from other neuronal systems. This input can either interrupt regular breathing or control the level of ventilation, allowing the respiratory system to respond appropriately to challenges and to be integrated with many different brain functions.
Pacemaker properties and synaptic interactions may both contribute to the generation of the respiratory rhythm
Perhaps the most important questions that still need to be answered about the neural control of breathing revolve around the mechanism and identity (or location) of the respiratory CPG. We address the mechanism in this section, and which cells are involved in the next.
Two general theories have evolved for the mechanism of the respiratory CPG. The first posits that subsets of neurons have pacemaker activity; the second, that synaptic interactions create the rhythm.
Some cells have ion channels that endow them with pacemaker properties (see pp. 397–398). For example, isolated cardiac myocytes produce rhythmic activity via “pacemaker currents” (see p. 489). Some neurons have similar pacemaker activity and repeatedly fire one spike at a time (i.e., beating pacemakers). Other pacemaker neurons repeatedly generate bursts of spikes (i.e., bursting pacemakers).
The first evidence for pacemaker activity in a mammalian respiratory nucleus came from the laboratory of Peter Getting. In brain slices from the guinea pig, putative premotor respiratory neurons in the NTS fire erratically at a low rate, with no apparent respiratory rhythm. However, adding thyrotropin-releasing hormone (TRH; see p. 1014) to the bathing medium causes these neurons to generate bursts of action potentials (Fig. 32-7A), similar to inspiratory bursts in RRNs of the DRG during eupnea in an intact animal (see Fig. 32-7B). Indeed, axons from the medullary raphé nuclei (see pp. 312–313) project to the NTS, where they release TRH (as well as serotonin and substance P) and thereby could induce bursting pacemaker activity in NTS neurons. This pacemaker activity might contribute to generation of the respiratory rhythm or augment respiratory output generated within another site. Pacemaker activity is also present in neurons within the preBötC (see p. 708), where serotonin (see Fig. 13-8B)—the main neurotransmitter released by the medullary raphé neurons noted above—can induce pacemaker activity. N32-12
FIGURE 32-7 Pacemaker activity in RRNs. ENG, electroneurogram; ΔVm, membrane potential difference. (A, Data from Dekin MS, Richerson GB, Getting PA: Thyrotropin-releasing hormone induces rhythmic bursting in neurons of the nucleus tractus solitarii. Science 229:67, 1985; B, data from Richerson GB, Getting PA: Maintenance of complex neural function during perfusion of the mammalian brain. Brain Res 409:128, 1987.)
Ability of Serotonin to Induce Pacemaker Activity in the Pre-Bötzinger Complex
Contributed by George Richerson, Walter Boron
eFigure 32-2 from an article by Ptak and colleagues shows the results of an experiment on an inspiratory preBötC neuron (see p. 708). This neuron has no intrinsic bursting activity. The bursting that we see in panel A is due to synaptic inputs.
EFIGURE 32-2 Serotonin converts some non–intrinsically bursting inspiratory preBötC neurons into intrinsically bursting cells. (From Ptak K, Yamanishi T, Aungst J, et al: Raphé neurons stimulate respiratory circuit activity by multiple mechanisms via endogenously released serotonin and substance P. J Neurosci 29:3720–3737, 2009, with permission.)
As seen in the left half of panel B, blocking these synaptic inputs with extracellular Cd2+ ions eliminates all activity. In the right half of panel B—still in the presence of Cd2+—we see that graded depolarizing currents of 15 pA and then 20 pA produce tonic spiking, but not rhythmic bursting.
In panel C, the addition of serotonin (5-HT) to the Cd2+ now causes rhythmic bursting during depolarization. Note that an increased level of depolarizing current (from 15 pA to 20 pA) in the right portion of this panel increases the frequency of this bursting.
In panel D, washing out the serotonin eliminates the bursting activity.
Neural circuits without pacemaker neurons also can generate rhythmic output (see pp. 397–398). Indeed, synaptic connections within and between the DRG and VRG establish neural circuits and generate EPSPs and IPSPs onto each other with a timing that could explain the oscillatory behavior during the respiratory cycle (see Fig. 32-6). Computational neurobiologists have proposed a variety of pure network models N32-13 of respiratory rhythm generation. According to these models, the CPG that produces the respiratory rhythm depends solely on synaptic connections between subtypes of RRNs and not at all on the pacemaker activity of individual neurons. Thus, the rhythm would be an emergent property of the network.
The Central Pattern Generator as an “Emergent Property”
Contributed by George Richerson, Emile Boulpaep, Walter Boron
One of the network models of respiratory rhythm generation includes three key elements for generating normal inspiratory output: (1) “central inspiratory activity,” (2) an integrator, and (3) an inspiratory off-switch. A group of neurons in the DRG would generate central inspiratory activity that would in turn cause a steady increase in firing of premotor neurons (such as the Iα neurons discussed in N32-9) that drive inspiratory motor neurons and the Iβ interneurons (also discussed in N32-9). The Iβ interneurons also receive input from the pulmonary stretch receptors. Thus, the Iβ interneurons “integrate” information from both the command to the inspiratory muscles and the resultant lung inflation. When the integrator determines that the inspiration has proceeded far enough, it triggers the inspiratory “off-switch” neurons, the late-onset inspiratory neurons in eFigure 32-1G. According to the model, the activated inspiratory off-switch neurons cause inspiration to stop by resetting central inspiratory activity, which allows expiration to take over. Expiration is passive and ends when central inspiratory activity builds up enough to initiate firing of inspiratory neurons again.
According to the model, neurons of the so-called pneumotaxic center in the pons can also trigger the inspiratory off-switch. This hypothesis fits with experimental data, because interrupting input from both the pneumotaxic center and the pulmonary stretch receptors (i.e., input carried by vagal afferents) results in loss of the off-switch and thus in apneusis (see p. 702).
One of the difficulties with network models is that the exact identity of each of the components of the model is not certain.
One of the difficulties with network models of breathing is that not all of the neurons within the respiratory network are known. Network models also must take into account the presence of the rich complement of intrinsic membrane properties that exist in the component neurons, but not all of these are fully characterized either. As a result, pure network models must be very complex to explain all aspects of the normal respiratory rhythm. Supplementing network models with intrinsic membrane properties (e.g., pacemaker activity) allows the models to be simpler. Moreover, from an evolutionary perspective, it is reasonable to infer that pacemaker cells may have driven primitive respiratory systems. In higher organisms, both pacemaker activity and synaptic interactions appear to be involved in generating the normal respiratory rhythm.
The respiratory CPG for eupnea could reside in a single site or in multiple sites, or could emerge from a complex network
Where is the respiratory CPG for eupnea? In 1851, Flourens proposed a noeud vital [vital node] in the medulla, a small region that is the sole site producing respiratory output. In 1909, the neuroanatomist Santiago Ramón y Cajal reported that neurons in the NTS receive afferents from pulmonary stretch receptors and that NTS neurons project directly to the phrenic motor nucleus. He concluded that the NTS is the site of respiratory rhythm generation. However, today—in spite of considerable progress—the question of where in the medulla the CPG is located remains controversial. Here we discuss three major proposals for the location of the respiratory CPG.
Toshihiko Suzue first introduced the in vitro brainstem/spinal cord preparation from the neonatal rat, and Jeffrey Smith and Jack Feldman further developed this preparation to study the generation of respiratory output. In this preparation, a small region in the rostral VRG—the preBötC (see Fig. 32-4)—generates rhythmic motor output in the phrenic nerve and hypoglossal nerve (CN XII, which innervates the tongue, an accessory muscle of inspiration). Destroying the preBötC in the isolated brainstem causes respiratory output to cease. In a slice preparation, the preBötC generates rhythmic bursts of activity that one can record from the hypoglossal nerve rootlets (Fig. 32-8). These bursts require the continuous release of serotonin from raphé neurons onto neurons of the preBötC. These experiments have led to the hypothesis that the preBötC is the site of the respiratory CPG. N32-14
FIGURE 32-8 Possible role of preBötC as the respiratory CPG. The recordings were made in a brain slice containing the preBötC, a piece of the hypoglossal nucleus, some rootlets of the hypoglossal nerve, and serotonergic neurons. (Data from Smith JC, Ellenberger HH, Ballanyi K, et al: Pre-Bötzinger complex: A brainstem region that may generate respiratory rhythm in mammals. Science 254:726, 1991.)
Site of the Respiratory Central Pattern Generator
Contributed by George Richerson, Emile Boulpaep, Walter Boron
Given the long history of the field and the many investigators working on different aspects of the problem, any conclusion about the precise location of a respiratory CPG is bound to be controversial.
In the case of the preBötC (see p. 706), some critics believe that the pattern of respiratory output generated resembles gasping. In addition, the conditions normally used to study the activity are artificial. In addition, destroying the preBötC in adult rats leads to ataxic breathing, and yet the animals survive. However, most investigators agree that the preBötC plays an important role in breathing, and defining the mechanisms of rhythm generation in the preBötC will be important for understanding normal breathing. For example, it is now clear that there are pacemaker neurons within the preBötC. Some of these pacemaker neurons are inhibited by hypoxia, whereas others are unaffected. This observation provides a possible explanation for why the pattern of breathing can change from eupnea to gasping when the oxygen levels in the brain become low, such as from cardiac arrest.
Distributed Oscillator Models
A different theory is that there is more than one CPG, any one of which could take over the job of generating the respiratory rhythm, depending on the conditions. As already noted, TRH can induce bursting pacemaker activity in DRG neurons. Moreover, a group of neurons in the parafacial respiratory group near the ventral surface of the medulla can also produce rhythmic activity. One interpretation is that various groups of respiratory neurons are latent CPGs and that the location of the dominant CPG can shift during different behaviors. A second interpretation is that the presence of rhythmicity in multiple areas represents redundancy, ensuring that respiratory output does not fail. This design would be analogous to initiation of the heartbeat, in which many cells within the sinoatrial node can be the first to fire, and cells in other parts of the heart can take over if the sinoatrial node fails (see p. 489). In this model, some respiratory neurons with intrinsic oscillatory behavior would contribute to rhythm generation only under unusual conditions (e.g., hypoxia, anesthesia, early in development); under these conditions, the neurons might generate abnormal or pathological respiratory output (e.g., gasping). A third interpretation is that only one CPG exists for eupnea, and that other regions augment the rhythm (and make it more robust) but are unable to generate a rhythm on their own.
Emergent Property Model
The most common early explanation for generation of the respiratory rhythm (e.g., that proposed by Lumsden; see p. 702) is that no individual region of the DRG or VRG is sufficient to generate the rhythm but that many of them are necessary. A normal rhythm would require the component neurons in multiple brainstem regions to be “wired up” in a specific way. Some still believe that this view is essentially correct, arguing that none of the individual regions proposed to contain the CPG can produce a pattern of activity with all of the features described during eupnea.
None of the models we have discussed is universally accepted, and some of their elements are not mutually exclusive. The challenges in testing these hypotheses include the complexity of the CNS even at the level of the medulla, the technical difficulty in studying neurons of the mammalian brainstem under conditions that replicate exactly those present in vivo, and the large number of nonrespiratory neurons in the medulla. Yet, because it is primitive both ontogenetically and phylogenetically, the respiratory CPG will probably prove to be far easier to define than most other mammalian neural networks, such as those responsible for consciousness or memory.