Reconfiguration of the pontomedullary respiratory network: a computational modeling study with coordinated in vivo experiments

Abstract

A large body of data suggests that the pontine respiratory group (PRG) is involved in respiratory phase-switching and the reconfiguration of the brain stem respiratory network. However, connectivity between the PRG and ventral respiratory column (VRC) in computational models has been largely ad hoc. We developed a network model with PRG-VRC connectivity inferred from coordinated in vivo experiments. Neurons were modeled in the "integrate-and-fire" style; some neurons had pacemaker properties derived from the model of Breen et al. We recapitulated earlier modeling results, including reproduction of activity profiles of different respiratory neurons and motor outputs, and their changes under different conditions (vagotomy, pontine lesions, etc.). The model also reproduced characteristic changes in neuronal and motor patterns observed in vivo during fictive cough and during hypoxia in non-rapid eye movement sleep. Our simulations suggested possible mechanisms for respiratory pattern reorganization during these behaviors. The model predicted that network- and pacemaker-generated rhythms could be co-expressed during the transition from gasping to eupnea, producing a combined "burst-ramp" pattern of phrenic discharges. To test this prediction, phrenic activity and multiple single neuron spike trains were monitored in vagotomized, decerebrate, immobilized, thoracotomized, and artificially ventilated cats during hypoxia and recovery. In most experiments, phrenic discharge patterns during recovery from hypoxia were similar to those predicted by the model. We conclude that under certain conditions, e.g., during recovery from severe brain hypoxia, components of a distributed network activity present during eupnea can be co-expressed with gasp patterns generated by a distinct, functionally "simplified" mechanism.

FIG. 1.

Schematic of the initial model of the brain stem respiratory network. To facilitate the tracing of pathways, regional connections are color-coded and dots are used to mark branch points of divergent projections. Both evidence-based (see text) and more speculative functional connections are represented in the model (see key). Model parameters for cell properties and connections are detailed in Tables A1–A4 of the appendix. Circled numbers and dashed lines in this and subsequent model diagrams label specific simulated perturbations applied. See text for details.

FIG. 2.

A and B: discharge patterns of individual respiratory neurons from each simulated population and network region (labels on the left) and the integrated population traces for expiratory and phrenic motoneurons and lung volume receptors (3 bottom traces) during “eupneic” respiratory rhythm (A) and following disconnection of lung volume or slowly adapting PSR feedback (B,“vagotomy”—perturbation 1—circled here and in Fig. 1), which produced an increase in the amplitude and duration of phrenic discharge. C: the effect of rostral pons “removal” (inactivation of populations representing the rostral pons; perturbation 2—circled here and in Fig. 1) resulted in an apneustic pattern with prolonged and irregular inspiratory phase durations. D: complete removal of the pons (perturbation 3—circled here and in Fig. 1) led to a gasping-like pattern. See text for details.

FIG. 3.

A: schematic of network module illustrating inactivation of the I-Driver population by inhibition (perturbation 4—circled) and subsequent application of external excitatory drive to both I-Aug and I-Dec populations (perturbation 5—circled). B: elimination of I-Driver activity terminated both network oscillations and phrenic activity. Subsequent increased excitation of the I-Aug and I-Dec populations (an additional excitatory drive to these populations) re-established network oscillations and a normal rhythmic phrenic output. Vertical dashed line indicates truncation of period without rhythm.

FIG. 4.

Extended model of the pontomedullary respiratory network incorporating connections inferred from results of coordinated in vivo experiments (Segers et al. 2008). Model parameters are detailed in Tables A5 and A6 of the appendix. A subset of the E-Aug-late population had a higher threshold and was designated E-Aug-late-HT. Same labeling conventions as in Fig. 1; key indicates speculative connections that were proposed in the initial model and confirmed (yellow triangles overlaid with blue squares), connections carried over from the initial model that were based on preliminary results of the work presented in Segers et al. (2008), and “new” connections suggested by later results in the study of Segers et al. (2008). See text for details.

FIG. 5.

Removal of slowly adapting PSR feedback in the extended model (perturbation 1 “vagotomy”; vertical dashed line) produced an increase in the amplitude and duration of phrenic discharges.

FIG. 6.

Firing rate histograms from simultaneously recorded pontine respiratory group neurons during fictive cough elicited by mechanical stimulation of the intrathoracic trachea in a decerebrate neuromuscularly blocked cat. A: traces show control respiratory-modulated discharge patterns [3 inspiratory (I) and 3 inspiratory-expiratory (IE) phase-spanning neurons] and altered rates during the cough motor pattern defined by altered phrenic (inspiratory) and lumbar (expiratory) motoneuron activity. B: firing rates of 3 pontine respiratory group (PRG) expiratory neurons during fictive cough. Data in A and B are modified from Shannon et al. (2004a) and used with permission.

FIG. 7.

Simulation with the extended model produced cough-like motor patterns. The simulated activities of phrenic, lumbar, and laryngeal motoneurons and functionally antecedent VRC and pontine neuron populations were similar to those observed in vivo. A fiber population consisting of 100 fibers, each with 100 excitatory synaptic terminals (see legend, Table A3 of the appendix) with a synaptic strength of 0.02 was used to represent cough receptor excitation; each fiber had a firing probability of 0.05 at each simulation time step. This fiber population excited a second-order “cough” neuron population (Fig. 4); see Table A7 (appendix) for properties of the cough population, Table A8 (appendix) for details of connections with other populations, and the text for further details.

FIG. 8.

Modeling effects of hypoxia on the respiratory motor pattern during non-random eye movement (NREM) sleep. A: changes in the integrated profiles of activity of different respiratory neurons during hypoxia (red traces) vs. normoxia (blue) from a coordinated in vivo study by Lovering et al. (2006) and used with permission. B: activity profiles in different medullary model populations closely reproduced changes found during hypoxia in NREM sleep. Hypoxia was simulated by incorporation of additional source of excitatory drive. C: schematic shows excitatory inputs that simulated a “hypoxic drive” to the indicated VRC populations (perturbation 3). A fiber population consisting of 100 fibers, each with 100 excitatory synaptic terminals with a synaptic strength of 0.025 was used to represent chemoreceptor drive; each fiber had a firing probability of 0.05 at each simulation time step. This fiber population, in turn, excited a second-order “Hypoxia” effect neuron population; see Table A7 (appendix) for properties of this population and Table A8 (appendix) for details of connections with other populations represented in C. See text for further details.

FIG. 9.

Gasping-like pattern (in both A and B) was produced by PRG removal (perturbation 4—circled; see also in Fig. 4). The I-Driver population in this case operates in the endogenous pacemaker mode and drives respiratory oscillations in the network and motor output. Membrane potential traces (independently scaled) of a single I-Driver neuron and a subset of other representative respiratory neurons are shown in both A and B. A: subsequent suppression of ventral respiratory column (VRC) expiratory neuron populations (vertical dashed line) did not suppress endogenous gasping-like activity in the I-Driver population and phrenic motor output. B: 2nd simulation run with initial activity the same as in A. Inhibition of I-Driver population (vertical dashed line) eliminated the phrenic gasp pattern; however, interactions among remaining active VRC populations were sufficient to generate rhythmic activity within the network.

FIG. 10.

Model prediction of possible co-expression of gasp-augmenting burst and eupneic-like ramp patterns with progressive alteration in tonic excitation of VRC populations. A, left: gasp-like phrenic activity pattern and traces of representative membrane potentials from single representative neurons after PRG removal and reduced excitation of E-Aug-BS,VRC IE, I-Aug, I-Dec, E-Aug-late, E-Aug-early, E-Dec-P, and E-Dec-T populations. A “burst–ramp” type phrenic motor pattern emerged with the onset of tonic re-excitation of the VRC network populations (red dashed line), together with additional I-Aug excitation mediated by an excitatory “Biasing” population (see Table A7, appendix for parameters). B: detail from A shows a “burst–ramp” type pattern during an inspiratory phase. C: summary of a prediction from the model showing that 2 active rhythm-generating mechanisms can be simultaneously expressed during recovery from hypoxic gasping. See text for details.

FIG. 11.

In vivo VRC neuronal activity profiles during control, hypoxic gasping, and recovery. A–F: firing rates of 10 simultaneously recorded VRC neurons and efferent phrenic nerve activity during the prestimulus control period (A), hypoxia-induced gasp-like activity (B), and re-oxygenation (C–F). G: control respiratory cycle-triggered histograms for neurons in A; 37 cycles averaged. One inspiratory neuron had a relatively uniform firing rate during the inspiratory phase and was designated “I-plateau” (I-Plat); H: integrated phrenic nerve activity profiles detail control, gasping, and a return to eupneic-like phrenic patterns with superimposed augmented bursts. Pattern with dashed line ellipse is similar to the phrenic activity profile observed in model simulations as shown in Fig. 10. See text for details.