Asymmetric neuromodulation in the respiratory network contributes to rhythm and pattern generation

Figure 1. Identifying the pre-BötC neuronal firing patterns that underlie the three-phase respiratory motor pattern. (A) Reconstruction of an MEA positioned in the pre-BötC in situ in a representative experiment. (B) A representative recording of the three-phase respiratory motor pattern on hypoglossal (HNA), phrenic (PNA) and vagal (VNA) nerves in concert with an ensemble of pre-BötC neurons. (C) Cycle-triggered histograms of the pre-BötC neurons shown in panel (A) include neurons that spike in late-expiration (E2), inspiration (I) and post-inspiration (PI). ϕ: phase; rad: radians. (D) Cycle-triggered histograms of all recorded pre-BötC neurons before clustering. (E) K-means clustering identifies 14 classes of pre-BötC firing patterns that underlie the three-phase respiratory motor pattern. (F) Reconstructed locations of a subset of neurons suggest that pre-BötC neuronal types are spatially mixed. The rostral direction is to the right of the plot. μm: micrometers.

Figure 2. Training a Hopfield model to encode the firing patterns of pre-BötC neuronal clusters. (A) The centroids of pre-BötC neuronal clusters were used as a basis to determine the sequential firing patterns to be encoded in the model. The respiratory cycle was discretized into eight sequential states to account for the brief firing patterns of the Pre-I, Post-I B and I-D populations. I: inspiratory; E: expiratory. (B) Each pre-BötC cluster was represented by 5 neurons in the model. Their training vectors were taken as +1 when the cluster fired at a high-frequency or −1 when the cluster was silent or firing at a low-frequency. (C,D) The resultant fast- (C) and slow- (D) synaptic connectivity of the Hopfield network after training to encode the sequential state vectors using Hebbian learning rules. wFast: fast synaptic weight; wSlow: slow synaptic weight. (E) As expected, the model generated the learned sequential firing patterns that underlie the three-phase respiratory motor pattern in the pre-BötC. Black: tonic or respiratory-modulated; Purple: inspiratory; Red: post-inspiratory; Orange: late-expiratory; Neuron No.: Neuron number; t.s.: timesteps.

Figure 3. Neuromodulation of the respiratory rhythm in silico. (A) The sequential firing pattern produced by the network at baseline. T: period; t.s.: timesteps. (B) Increasing slow synaptic inhibition in the model evokes an increase in the frequency of the respiratory rhythm without any change in the sequential firing pattern of the network. T: period; t.s.: timesteps. (C) Increasing slow excitation in the model evokes a collapse of the respiratory rhythm characterized by a reduction in the number of active units whose remaining activity was either tonic or fast bursting. t.s.: timesteps.

Figure 4. The 5HT1aR agonist 8-OH DPAT increases the frequency of the respiratory rhythm without changing the firing pattern of pre-BötC ensembles. (A) Systemic administration of 8-OH DPAT evoked an increase in the frequency of the respiratory rhythm as observed in PNA and VNA that was associated with a reconfiguration of pre-BötC ensemble activity. In this representative experiment six pre-BötC neurons maintained their original firing patterns, but at a higher frequency. In addition, five pre-BötC neurons became silent, and six pre-BötC neurons were activated. Left panel: baseline; right panel: after systemic 1 μM DPAT. (B) The frequency of the respiratory rhythm was significantly increased after systemic application of 8-OH DPAT. * p < 0.05. F: frequency; bpm: breaths per minute. (C) K-means clustering of all recorded pre-BötC neurons at baseline and after systemic 8-OH DPAT identified many of the pre-BötC neuronal types previously observed. ϕ: phase; rad: radians. (D) All clusters except the Aug-E population were present at similar ratios at baseline (circles) and after systemic 8-OH DPAT (crosses) suggesting that despite the reconfiguration of pre-BötC ensemble activity, the…

Figure 5. The reconfiguration of pre-BötC ensemble activity after opioid-induced respiratory depression is consistent with model simulations after increasing slow-excitation. (A) Systemic fentanyl administration evokes a collapse of the respiratory rhythm on phrenic (PNA), hypoglossal (HNA) and vagal (VNA) nerves. (B) Consistent with the model simulations of uniformly increasing slow excitation, systemic fentanyl administration was associated with a reduction in the size of pre-BötC ensembles and spared tonic (gray), fast- (purple) and slow- (green) bursting firing patterns. Please note that these raster plots reflect neuronal ensembles compiled from all experiments to better visualize the firing patterns expressed after fentanyl-evoked opioid-induced persistent apnea. As such, the ordering of units post-fentanyl exposure (right panel) do not have a one-to-one correspondance with the neuronal rasters at baseline. Further, to aid in the visualization of baseline firing patterns, we have plotted pink bars to indicate the inspiratory period for each pre-BötC ensemble. (C) Systemic fentanyl administration significantly reduced the number of active neurons in pre-BötC ensembles. ***p < 0.001.…

Figure 6. Population coding of the transitions between respiratory phases in the model and in the pre-BötC in situ. (A) Because the recall of the next sequential state involves a slight overlap of sequential assemblies, the model generates brief peaks in the population firing rate at each of the three transitions between respiratory phases when the adjacent state vectors are most distant. Black: tonic or respiratory-modulated; Purple: inspiratory; Red: post-inspiratory; Orange: late-expiratory; Pop. Rate: population firing rate in spikes/timestep. (B) Consistent with the model, pre-BötC ensemble activity is associated with a population firing rate (Pop. Rate, green trace) that also peaks at or near the transitions between respiratory phases. (C) There was a significant cross-correlation between the population firing rate and the three-phase respiratory pattern of vagal nerve activity (VNA). This cross-correlation was considered significant if it was greater than the 99.9%-ile of a bootstrap distribution in which the inter-spike intervals for each unit were shuffled before computing the shuffled ensemble’s population firing rate and its correlation with VNA. White bar: original…