Ventrolateral respiratory column (VRC) circuits that modulate breathing in response to changes in central chemoreceptor drive are incompletely understood. We employed multielectrode arrays and spike train correlation methods to test predictions of the hypothesis that pre-Bötzinger complex (pre-BötC) and retrotrapezoid nucleus/parafacial (RTN-pF) circuits cooperate in chemoreceptor-evoked tuning of ventral respiratory group (VRG) inspiratory neurons. Central chemoreceptors were selectively stimulated by injections of CO2-saturated saline into the vertebral artery in seven decerebrate, vagotomized, neuromuscularly blocked, and artificially ventilated cats. Among sampled neurons in the Bötzinger complex (BötC)-to-VRG region, 70% (161 of 231) had a significant change in firing rate after chemoreceptor stimulation, as did 70% (101 of 144) of the RTN-pF neurons. Other responsive neurons (24 BötC-VRG; 11 RTN-pF) had a change in the depth of respiratory modulation without a significant change in average firing rate. Seventy BötC-VRG chemoresponsive neurons triggered 189 offset-feature correlograms (96 peaks; 93 troughs) with at least one responsive BötC-VRG cell. Functional input from at least one RTN-pF cell could be inferred for 45 BötC-VRG neurons (19%). Eleven RTN-pF cells were correlated with more than one BötC-VRG target neuron, providing evidence for divergent connectivity. Thirty-seven RTN-pF neurons, 24 of which were chemoresponsive, were correlated with at least one chemoresponsive BötC-VRG neuron. Correlation linkage maps and spike-triggered averages of phrenic nerve signals suggest transmission of chemoreceptor drive via a multipath network architecture: RTN-pF modulation of pre-BötC-VRG rostral-to-caudal excitatory inspiratory neuron chains is tuned by feedforward and recurrent inhibition from other inspiratory neurons and from “tonic” expiratory neurons.
- inspiratory and expiratory neurons
- pre-Bötzinger complex
- respiratory network
- spike train
central chemoreceptors monitor brain CO2/pH and, with their chemoresponsive follower neurons, provide an essential component of the drive to breathe. The retrotrapezoid nucleus/parafacial region of the brain stem (RTN-pF) contains central chemoreceptors (Abbott et al. 2009; Gourine et al. 2010; Guyenet et al. 2010) and is a rostral extension of the medullary ventrolateral respiratory column (VRC), a network containing circuits essential for generating and modulating the motor pattern for breathing (Onimaru et al. 2008; Smith et al. 2009). The VRC extends caudally through the Bötzinger complex (BötC), largely composed of inhibitory expiratory neurons (Fedorko et al. 1989; Jiang and Lipski 1990; Lindsey et al. 1989; Merrill and Fedorko 1984), and the pre-Bötzinger complex (pre-BötC), a core “compartment” for inspiratory rhythm generation (Smith et al. 1991) that may also have a CO2/pH chemosensory function (Koizumi et al. 2010; Nattie 2001; Peever et al. 2001; Solomon et al. 2000). The most caudal region of the VRC is the ventral respiratory group (VRG) with its premotoneuron and motoneuron populations. The VRG includes bulbospinal neurons that drive phrenic motoneurons innervating the diaphragm (Bianchi et al. 1995; Lois et al. 2009) and may also contain chemoreceptors (Nattie and Li 1996).
Although axonal projections of RTN-pF neurons with properties characteristic of chemoreceptors have been traced to every segment of the VRC (Abbott et al. 2009), the organization of chemoreceptor reflex circuits within the VRC remains incompletely understood (Guyenet et al. 2010). In a recent study, we employed multielectrode arrays and spike train analysis to test hypotheses on network mechanisms for the respiratory modulation and central chemoreceptor-evoked responses of RTN-pF neurons (Ott et al. 2011). Here, we report complementary results from that work. Correlational linkages support a new network model for chemoreceptor-mediated tuning of pre-BötC-VRG circuits and respiratory drive. The model incorporates tonic columnar expiratory neurons in a multipath architecture.
General Methods and Surgical Preparation
All experiments were performed according to protocols approved by the University of South Florida's Institutional Animal Care and Use Committee with strict adherence to all Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC), National Institutes of Health, and National Research Council guidelines.
As detailed descriptions of the methods have recently been published (Nuding et al. 2009b; Ott et al. 2011; Segers et al. 2008), only a brief review is included here. Data were obtained from seven adult cats (2.9–4.3 kg) of either sex. Animals were initially anesthetized with 5% isoflurane mixed with air and maintained with 0.5–3.0% isoflurane until decerebration. Arterial blood pressure, end-tidal CO2, and tracheal pressure were monitored continuously; arterial Po2, Pco2, and pH were measured periodically. The left and right vago-sympathetic nerve trunks were isolated in the neck and sectioned to remove vagal sensory feedback from pulmonary stretch receptors. A concentric catheter was inserted into the left axillary artery and advanced to the bifurcation of the vertebral artery (Nuding et al. 2009b); preceding branches of the axillary artery were ligated (Kuwana and Natsui 1987). At the end of each experiment, animals were overdosed with Beuthanasia (0.97 mg/kg; Schering-Plough Animal Health) and perfused with a 10% neutral-buffered formalin solution.
Efferent phrenic nerve activity—used as an indication of respiratory drive, to assess stimulus effectiveness, and to identify the phases of breathing—was monitored together with signals from two multielectrode arrays with individually adjustable high-impedance tungsten microelectrodes (1-μm tip diameter; 10–12 MΩ). A 24-electrode array (4 × 6 arrangement) was placed in the rostral region of the medulla to monitor RTN-pF neurons, and a 32-electrode array (2 × 16) was placed in the region of the BötC-VRG. Electrode placement was guided by anatomical landmarks (obex, brain stem midline), appropriate stereotaxic coordinates derived from Berman (1968), and the results of previous studies (Baekey et al. 2004; Connelly et al. 1992; Schwarzacher et al. 1995). Stereotaxic coordinates of recording sites were mapped into the three-dimensional space of a computer-based brain stem atlas derived from The Brain Stem of the Cat: A Cytoarchitectonic Atlas with Stereotaxic Coordinates (Berman 1968) with permission of the University of Wisconsin Press, as described in Segers et al. (2008).
Stimulation of Central Chemoreceptors
Control neuronal activity was recorded for a 30-min period before any stimuli were presented. Central chemoreceptors were then selectively stimulated by 30-s injections of 1.0 ml of a CO2-saturated 0.9% saline solution into the vertebral artery (Arita et al. 1988a, 1988b; Nuding et al. 2009b; Ott et al. 2011). The stimulus protocol included at least five trials separated by 4.5-min intervals to allow phrenic nerve activity to return to prestimulus levels. Stimulus effectiveness was confirmed by measures of the peak amplitude of the integrated phrenic nerve signal; effective reflexes were identified by a change >2 SD in the peak integrated phrenic nerve amplitude from the mean of prestimulus values (Nuding et al. 2009b). Control saline injections did not evoke significant changes in phrenic nerve frequency or amplitude.
Postexperimental Processing and Data Analysis
Classification of neuronal responses.
Signals from single neurons were isolated with interactive spike sorting software (O'Connor et al. 2005). Methods used to evaluate and classify neuronal responses have been described previously (for details see Fig. 3C in Ott et al. 2011). Briefly, we compared neuronal firing rates during a 90-s “response” window (30-s stimulus injection plus 60 s postinjection) with those during the immediately preceding 90 s of “control” in order to measure significant firing rate changes. Note that window duration extended beyond the injection period to ensure that responses with varying time lags would be detected. Results reported are averages of at least five trials. Cumulative sum histograms (Ellaway 1978) were calculated from peristimulus time histograms. Changes in activity that exceeded 3 SD (Davey et al. 1986) were confirmed with a bootstrap-based statistical method (as described in Nuding et al. 2009b); the P value threshold (significance level) was set by controlling the false discovery rate to a level of 0.05 (Benjamini and Hochberg 1995). Responses were classified into one of five response categories: increase (↑), decrease (↓), biphasic response [increase-decrease (↑↓) or decrease-increase (↓↑)], or no change (→).
Neurons were also assessed for significant changes in the depth of respiratory modulation [i.e., “rate ratio” (↕)]; the rate ratio is a measure of cross-phase modulation. This parameter was evaluated by dividing each respiratory cycle within the stimulus response period and its corresponding control period into 20 slices and measuring the mean firing rate in a 7-slice-wide window for each of the 20 possible starting locations of the window, yielding 20 numbers for each cycle. This test measured the ratio of the maximum to the minimum mean firing rate. The bootstrap method was used to evaluate whether or not there was a significant change in the degree of modulation.
Respiratory modulation of firing rates.
Normalized respiratory cycle-triggered histograms (CTHs) were constructed from the 30-min control period recordings for each neuron (Cohen 1968). Neurons were classified as respiratory modulated if either of two complementary statistical tests (Morris et al. 1996a; Netick and Orem 1981; Orem and Netick 1982) rejected the null hypothesis (P < 0.05); neurons with no preferred phase of maximum activity were considered non-respiratory modulated (NRM). Respiratory-modulated neurons were classified as inspiratory (I), expiratory (E), or phase-spanning (IE or EI) according to the part of the cycle during which the cell was most active (Cohen 1968). If the peak firing rate occurred during the first or second half of the phase, I and E cells were further classified as decrementing (Dec) or augmenting (Aug), respectively. Neurons were additionally designated as phasic (P), if their firing probability was essentially zero during any part of the respiratory cycle, or tonic (T) otherwise (Morris et al. 1996b).
Cross-correlation histograms (CCHs) were calculated by using the entire recording for all pairs of simultaneously monitored neurons. The goal of this approach is to define simple circuit models that reproduce experimentally observed features (Aertsen and Gerstein 1985; Kirkwood 1979; Moore et al. 1970; Ostojic et al. 2009). For example, central peaks or troughs can be simply interpreted as evidence of a shared input that influences both cells' firing rates with similar or opposite effects, respectively. An offset peak with a positive time lag suggests excitation of the target neuron by the trigger (or reference) cell, whereas an offset trough is evidence for an inhibitory process. All offset feature data are presented with a positive time lag.
Peak or trough features were identified as departures ≥3 SD from the mean of shift-predictor control correlograms calculated using 20 respiratory cycles at a time with all possible shifts of these cycles (Nuding et al. 2009a). Detectability indices (equal to the ratio of the maximum amplitude of feature departure from background activity divided by the SD of the correlogram noise) were calculated (Aertsen and Gerstein 1985; Melssen and Epping 1987). Autocorrelation histograms aided interpretation of the cross-correlogram features (Moore et al. 1970; Perkel et al. 1967a, 1967b). Correlation linkage maps for groups of simultaneously monitored neurons were generated (Segers et al. 2008).
The spike times of 186 neurons in 5 recordings were used to generate triggered averages of full-wave rectified contralateral phrenic nerve signals (see, e.g., Shannon et al. 2000). Averages were screened for short-time scale characteristics in the efferent signal time-locked to the trigger events. Peaks and troughs in the averages were evaluated for significance (P < 0.05) with an adaptation of the “multiple fragment” statistical analysis method (Poliakov and Schieber 1998) applied to all bins of the average. A spike-triggered average (STA) was calculated for each pair of adjacent even-numbered respiratory cycles, bin by bin. Control averages were calculated from the same two cycles by taking trigger events from one cycle and the analog signal from the other. Using signals from every other cycle minimized the interference of local serial correlations on the statistics. A two-sided Wilcoxon signed-rank test with a Bonferroni correction for multiple testing was used to determine whether the mean of each STA bin significantly differed from the mean of the corresponding control STA bin. Signals from odd-numbered cycles were similarly evaluated. Features were classified as significant if reported as such for either data subset.
This work was part of a larger study on VRC network organization and central chemoreceptor reflex circuits. Complementary results on functional connections for shaping the respiratory modulation and chemoreceptor-evoked responses of RTN-pF neurons have been reported (Ott et al. 2011). The present data were acquired during eight recording sessions in seven animals. A 32-electrode array monitored neurons (n = 231) within the BötC-to-VRG domain, extending from 1.5 mm caudal to 5.1 mm rostral to the obex, 3.3 to 4.6 mm lateral to the midline, and 2.6 to 5.8 mm below the dorsal surface of the medulla. A 24-electrode array recorded RTN-pF neurons (n = 144) in the region extending from 5.0 to 7.0 mm rostral to the obex, 1.4 to 4.2 mm lateral to the midline, and 3.3 to 7.8 mm below the dorsal surface of the medulla.
Table 1 shows the numbers of neurons found in each region grouped by category of respiratory modulation. The majority of neurons in the BötC-VRG and RTN-pF domains were respiratory modulated (79% and 66%, respectively). During the initial control recording period, a greater proportion of respiratory-modulated BötC-VRG neurons had a phasic discharge pattern (62%), whereas most respiratory-modulated RTN-pF neurons (94%) were tonic with some activity throughout the respiratory cycle. The respiratory CTHs in Fig. 1A illustrate the variety of discharge patterns relative to phrenic nerve activity found in one simultaneously recorded group of neurons in the pre-BötC-VRG region; firing patterns are shown for several I-Aug-P and I-Dec-P neurons, two tonic expiratory neurons [cells 813 (E-Dec-T) and 815 (E-Aug-T)], an E-Aug-P neuron (820), and an E-Dec-P cell (848).
Responses to Central Chemoreceptor Stimulation
Firing rate histograms generated from the activity of neurons shown in Fig. 1A (marked by colored boxes around each cell's identification code) and three additional I-Dec-P cells during one of five central chemoreceptor challenges are shown in Fig. 1B together with integrated efferent phrenic activity and arterial blood pressure. The 30-s stimulus injection period is delineated by the horizontal line at the bottom; the respiratory modulation and response of each neuron (arrow) based on analysis of all five stimulus trials are indicated to the left of each trace. Cells with similar respiratory-modulated discharge patterns had diverse responses. The maximum firing rate represented in each histogram is shown to the right. Recording sites mapped to stereotaxic coordinates (Fig. 1C) include color-coded responses of the corresponding neurons (see key). The plots in Fig. 1D show peak firing rates per respiratory cycle for paired control (black traces) and 90-s response evaluation periods (red traces) for three of five central chemoreceptor stimulus trials. The representative traces for neurons 808 (Fig. 1D1) and 820 (Fig. 1D2) show significant (P < 0.05) peak rate increases and decreases, respectively.
Seventy percent of neurons in both the BötC-VRG and RTN-pF responded to central chemoreceptor stimulation with a significant change in firing rate (Table 2). The more common direction of change differed for neurons in the two regions, however: 69% of rate changers in the BötC-VRG responded with an increased firing rate, whereas the rates of 61% of those in the RTN-pF initially decreased. There was no significant change in firing rate in response to chemoreceptor stimulation for cell 822; however, this cell's depth of respiratory modulation increased (denoted by double-headed arrow, ↕; see methods). We observed examples of neurons from both regions (24 BötC-VRG; 11 RTN-pF) that had this significant change in the rate ratio metric unaccompanied by a significant change in firing rate relative to the control. These neurons were also considered responsive. Some cells (57 RTN-pF; 125 BötC-VRG) responded with an alteration in depth of respiratory modulation as well as a change in firing rate; these cases were classified according to the rate change, not the rate ratio.
Functional Connectivity Among BötC-VRG Neurons
Cross-correlation analysis was applied to 3,831 pairs of BötC-VRG neurons. Table 3 summarizes the total number of central and offset peaks and troughs. BötC-VRG neuron pairs are grouped according to the respiratory discharge pattern of the trigger (left) and target neurons (top). Overall, 161 different BötC-VRG neurons (70%) had short-time scale spike train correlations with at least one other BötC-VRG neuron. A total of 146 BötC-VRG neurons (63%) were elements of BötC-VRG neuron pairs with correlogram central peaks (n = 269) or troughs (n = 103) indicative of a shared input of like or opposite sign, respectively. Seventy-six neurons triggered 202 cross-correlograms with other BötC-VRG target neurons that featured offset peaks or troughs, correlational signatures of excitation or inhibition (104 peaks, 98 troughs; Table 3). Forty-five of these reference neurons were correlated with more than one target neuron, a result consistent with local functional divergence. The majority of correlations were among chemoresponsive BötC-VRG neurons: When this same analysis is limited to the 2,619 pairs composed of chemoresponsive BötC-VRG cells, 70 BötC-VRG neurons triggered 189 offset-feature correlograms (96 peaks and 93 troughs) with at least one other chemoresponsive BötC-VRG cell (Table 4); 42 cells were correlated with more than one target neuron.
Correlational Linkages of Chemoresponsive Pre-BötC-VRG Neurons
Cross-correlograms (Fig. 2A) from the group of neurons represented in Fig. 1 had offset peaks (Fig. 2A, 1–3), offset troughs (Fig. 2A, 5–7 and 9–18), and central peaks (Fig. 2A, 4, 8–11, and 19); note that some correlograms had both central peak and offset trough features. The correlation linkage map (Fig. 2C) shows a compilation of pairwise correlations for the group of neurons and provides a framework to aid visualization of the detected relationships and correlogram feature sets. In this and subsequent maps, each large “sphere” represents a neuron with the corresponding identification number, color-coded respiratory modulation pattern, and chemoreceptor-evoked response (arrow). Inspiratory neurons are represented in the left region of the map; inspiratory neurons that triggered correlograms with positive-lag offset troughs are distinguished by an alternate color (pink). Tonic and phasic expiratory neurons are represented in the right column. Small white and black circles at the ends of the numerically labeled lines between spheres represent offset peaks or troughs, respectively, in the correspondingly numbered correlogram. The curved gray lines with filled circles at both ends indicate central correlogram peaks.
Correlations among inspiratory neurons represented by the green spheres included a “chain” of successive offset peaks extending from pre-BötC region inspiratory neurons 831, 802, and 808 to more caudal neurons 851, 822, and 817; cells 822 and 817 then converge upon 821. The central peak feature (Fig. 2A, 4) for the correlogram calculated for neurons 802 and 808 indicates a shared coordinating influence. Offset-peak correlograms triggered by neuron 802 indicate additional links with inspiratory target neurons 853 and 842 (pink spheres). Each of these neurons, in turn, triggered correlograms with offset troughs, as did two other putative inhibitory inspiratory neurons, 826 and 812 (Fig. 2A, 5 and 6). We note that cell 853 was (recurrently) linked with an offset trough to inspiratory neuron 802 (Fig. 2A, 7) and also correlated with E-Aug-P neuron 820 (blue sphere). The central peak feature for 853 and 812 (Fig. 2A, 8) indicates short-time scale coordination of the spike activity by a shared influence. Neuron 842 triggered offset-trough correlograms with E-Dec-T (yellow) target neuron 813 (Fig. 2A, 14), as did inspiratory neuron 826 (Fig. 2A, 12), which was also linked to E-Dec-P neuron 848 (Fig. 2A, 13).
Correlograms triggered by inspiratory neuron 816 (Fig. 2A, 9–11) had offset troughs adjacent to asymmetric central peaks (target neurons 826, 812, and 851). These features, coupled with the neurons' common phase of peak firing and responses to chemoreceptor stimulation, suggest the operation of shared influences upon the neurons and a sequential inhibitory chain (e.g., 816-to-812-to-802) for modulating the pre-BötC node of the excitatory inspiratory neuron chain.
Tonic expiratory neuron 815 (yellow) was tightly coordinated with E-Aug-P neuron 820 (Fig. 2A, 19). Both neurons had augmenting activity patterns as the expiratory phase developed (Fig. 1A), and both responded to central chemoreceptor stimulation with transient reductions in firing rates. The two neurons also had divergent offset-trough functional connections with inspiratory neurons distributed throughout the sampled region (e.g., Fig. 2A, 15–18). The gray “substrate” under sections of the linkage map highlights the distributed associations of cell 815. The chemoreceptor-evoked reduction in the firing rate of this tonic neuron, most notably during the inspiratory phase (Fig. 1B), is consistent with disinhibitory influence upon the linked inspiratory neurons.
Results from STA of the full-wave rectified contralateral phrenic nerve signal were consistent with hypotheses suggested by other relationships represented in the linkage map. The average triggered by inspiratory neuron 802 revealed an offset peak (arrow, Fig. 2Ba) superimposed upon a broader central trough. The peak supports the excitatory chain hypothesis; the broader trough is consistent with the influence of a recurrent inhibitory circuit also driven in part by neuron 802. The offset trough (arrow, Fig. 2Bb) in the average triggered by tonic expiratory neuron 815 is consistent with functional inhibition of the phrenic motoneurons or antecedent elements of the excitatory inspiratory neuron chain. Overall, 42 of 186 STAs contained characteristics indicative of a cell's influence upon phrenic motor nerve activity. The majority of these involved BötC-VRG neurons: Significant features in STAs calculated for 39 of 113 BötC-VRG cells had central (n = 23) or offset (n = 4) peaks or troughs or combinations of these features (n = 12). An effect upon phrenic activity could be inferred for only 3 of 73 evaluated RTN-pF neurons; in each case, the STA contained a central trough. Further examples of STAs are shown in subsequent figures.
Distributed Linkages of Tonic Expiratory Neurons
Results from a second animal provide additional evidence for widespread actions of tonic expiratory neurons upon pre-BötC-VRG inspiratory neurons. Figure 3A shows the recording sites and chemoreceptor-evoked responses of a group of 14 simultaneously monitored neurons. CTHs from 4 of 12 inspiratory neurons and 2 tonic expiratory neurons are shown in Fig. 3B. We note that, as for the data shown in Fig. 1, the tonic expiratory neurons (879, 847 in Fig. 3A) were recorded at sites well caudal to the coordinates of the pre-BötC. The average peak firing rates of both expiratory neurons increased in response to central chemoreceptor stimulation, as did their depth of respiratory modulation, reflecting in part a reduced firing rate during the inspiratory phase. Cross-correlogram 20 (Fig. 3C) for pre-BötC inspiratory neuron 810 and caudal target inspiratory neuron 862 featured an offset peak. Correlograms triggered by neuron 862 with target cells 857 and 814 revealed additional offset peaks represented in the correlation linkage map for this group (Fig. 3E).
In addition to the inspiratory neurons linked by offset-peak correlations, we identified overlapping sets of mutually correlated inspiratory neuron pairs with central peak features. These associations are represented by colored circles within the green spheres in the linkage map (Fig. 3E). For example, all of the correlograms calculated for every pair composed of the seven neurons marked with a blue circle contain a central peak; one example is shown in correlogram 21 from the 829–862 pair (Fig. 3C). A central peak (Fig. 3C, 22) was also identified in the cross-correlogram for tonic expiratory neurons 847 and 879.
All of the inspiratory neurons shown in this sample (within the gray background) had transient reductions in their discharge probability following trigger spikes of tonic expiratory neuron 847; the offset troughs in correlograms 23–27 (Fig. 3C) document 5 of these 12 inferred inhibitory relationships. Correlograms triggered by tonic expiratory neuron 879 for three of these inspiratory target cells also had offset troughs, providing evidence for a functional convergence of coordinated tonic expiratory neurons upon the same pre-BötC target neurons (cells 810, 862, and 890; Fig. 3E).
The average of the phrenic nerve signal triggered by inspiratory neuron 821 included a narrow offset peak superimposed upon a broader peak that spanned the trigger origin (Fig. 3Dc). Together, these features suggest the influence of other correlated premotor inspiratory neurons. The offset trough (Fig. 3Dd) in the average triggered by tonic expiratory neuron 847 provides further support for functional inhibition of the inspiratory neuron chain.
Evidence for Functional Connectivity of RTN-pF Neurons with the BötC-to-VRG Domain
The spike trains of 4,028 neuron pairs, each composed of an RTN-pF and a BötC-VRG neuron, were evaluated for short-time scale correlations. Table 5 shows the numbers of offset and central peaks and troughs detected in correlograms calculated using an RTN-pF trigger neuron and a BötC-VRG target. As in Table 3, pairs are arranged according to the respective respiratory discharge patterns of the RTN-pF reference (left) and BötC-VRG target neuron (top).
A total of 37 RTN-pF neurons (26%) were correlated with at least one BötC-VRG neuron. Of these, 24 were triggers of 68 correlograms with positive time lag offset peaks (n = 47) or troughs (n = 21), results consistent with paucisynaptic functional neuronal connectivity from the RTN-pF to BötC-VRG neurons. Eleven RTN-pF reference neurons were correlated with more than one BötC-VRG target neuron, evidence for divergent connectivity. Functional input from at least 1 RTN-pF cell could be inferred for 45 BötC-VRG neurons (19%); in 12 cases, there was evidence for the convergent influences of 2 or more RTN-pF cells upon a single BötC-VRG neuron.
Twenty-eight (19%) RTN-pF neurons were associated with 1 or more BötC-VRG neurons via shared influences as indicated by 85 correlograms with a central peak or trough. In 13 instances, correlograms triggered by the same RTN-pF neuron but with RTN-pF and BötC-VRG target neurons had offset features, while the correlogram for those two target neurons had a central feature. Each such result from a trio of neurons is consistent with a particular RTN-pF neuron serving as a putative input source shared by the target cells.
Twenty-four chemoresponsive RTN-pF neurons were correlated with at least one chemoresponsive BötC-VRG neuron as indicated by an offset or central feature in the correlogram. Table 6 shows the numbers of correlograms with offset peaks and troughs for chemoresponsive RTN-pF-BötC-VRG cell pairs arranged by respiratory modulation category and response to chemoreceptor stimulation. The recording site coordinates and responses of three neurons from this data subset are represented in Fig. 4A. The rostral tonic decrementing expiratory RTN-pF neuron 415 had a transient decline in average activity late in the inspiratory phase (Fig. 4B, top CTH). This neuron responded to chemoreceptor stimulation with a reduced firing rate, whereas the rates of phasic VRG inspiratory neuron 818 and augmenting expiratory cell 826 (Fig. 4B, middle and bottom CTHs) increased. The correlograms triggered by the RTN-pF neuron and computed for both VRG target spike trains had offset troughs (Fig. 4C, 28 and 29). These features and the neurons' responses represented in the linkage map (Fig. 4D) are consistent with inhibition of both VRG target neurons by cell 415 and suggest the hypothesis that evoked disinhibition contributes to the increased VRG neuron activity with chemoreceptor stimulation; see discussion.
Multiple correlations were identified in a group of 14 RTN-pF and VRG neurons from a fourth animal; recording site coordinates and responses are represented in Fig. 5A. Firing rate histograms and CTHs from a subset of the neurons (Fig. 5, B and C, respectively) show that tonic decrementing expiratory RTN-pF neurons 418 and 402 both had transient rate reductions late in the inspiratory phase similar to that of neuron 415 in Fig. 4B. The VRG sample included phasic augmenting (817) and decrementing (829) inspiratory neurons and tonic decrementing (820) and phasic augmenting (831) expiratory neurons (Fig. 5C, right).
Cross-correlograms from RTN-pF neuron pairs in this data set have been previously reported (Ott et al. 2011). Here we document specific correlogram features (Fig. 5D) and a corresponding correlation linkage map that incorporates identified functional associations with VRG neurons (Fig. 5F). Central peaks were the primary features in correlograms for pairs composed of RTN-pF neurons 402, 408, and 418. This feature set is represented by the small blue circles (Fig. 5F; e.g., Fig. 5D, 30) and is indicative of shared coordinating influences.
This trio had other associations with RTN-pF neurons (Fig. 5F). Notable features in correlograms triggered by neuron 402 included an offset “peak-trough” with target neuron 427 (Fig. 5D, 31), an offset peak with target 416 (Fig. 5D, 32), and an indirect link with neuron 427 via the offset trough in the correlogram triggered by cell 416 (Fig. 5D, 33). Neuron 418 had similar correlational relationships with cells 416 and 427 (Fig. 5D, 35, 37). These results are consistent with the operation of parallel excitatory and feedforward inhibitory RTN-pF circuits and are considered further in discussion. In addition to triggering offset trough-feature correlograms with RTN-pF targets (Fig. 5F; e.g., Fig. 5D, 36), cells 402, 408, and 418 also triggered correlograms with offset peaks for VRG augmenting expiratory target neuron 831 (e.g., Fig. 5D, 34, 39).
Evidence for Multiple Sources of Pre-BötC-VRG Tonic Expiratory Neuron Modulation
Other correlation features identified in this set of neurons included offset peaks in correlograms triggered by neurons 408 and 418 with pre-BötC-VRG tonic expiratory target neuron 820. The 418–820 correlogram (Fig. 5D, 38) had an offset peak “superimposed” upon a broader central peak, the latter feature indicative of a shared influence. The correlogram triggered by decrementing inspiratory neuron 829 with target cell 820 had an offset trough (Fig. 5D, 45). These and other results shown in Tables 5 and 6 suggest that multiple influences shape the discharge pattern of pre-BötC-VRG tonic expiratory neurons; see discussion.
Correlograms triggered by tonic expiratory neuron 820 with each of four VRG inspiratory target neurons (809, 812, 815, and 817, grouped within the gray background in Fig. 5F) all had positive time lag offset troughs (Fig. 5D, 40–43). Each pair of neurons in this group of four tended to discharge synchronously as illustrated by the central correlogram peak for pair 815–817 (Fig. 5D, 44). The STAs of phrenic nerve signals triggered by three of the VRG inspiratory neurons (Fig. 5E, e–g) had short-latency narrow offset peaks (3.9 ± 0.4 ms lag to peak; 1.7 ± 0.3 ms half-width) superimposed upon broader peaks. These results are consistent with the influence of tightly synchronized premotoneurons correlated by functionally antecedent shared inputs. The offset trough in the average triggered by neuron 820 (Fig. 5Eh) shows a functional inhibition of inspiratory drive correlated with the spikes of this expiratory neuron.
The combinations of responses and correlogram features identified in this study suggest that eight distinct operations among VRC neurons contribute to central chemoreceptor modulation of breathing. As shown in Fig. 6A the increases (↑) and reductions (↓) in the reference and target neuron firing rates in response to chemoreceptor challenge together with the associated offset correlogram feature (peak or trough) are simply interpreted as distinct functional actions (excitation, disfacilitation, inhibition, and disinhibition) that either promote or limit changes in target neuron activity following chemoreceptor stimulation.
The mosaic of identified correlational subassemblies incorporating these operations (Fig. 6B) offers a new perspective on respiratory network architecture and includes multiple sites for regulating the motor pattern for breathing. Selective stimulation of central CO2/pH chemoreceptors evokes increased firing rates in the rostral-to-caudal VRC excitatory inspiratory neuron chain (Rekling and Feldman 1998; Segers et al. 1987) that extends from the pre-BötC region to more caudal VRC populations (Fig. 6B, square 1). The chain includes putative premotor bulbospinal neurons identified by offset peaks in STAs of phrenic motoneuron signals. Neurons in the chain excite inhibitory inspiratory neurons (Fig. 6B, square 2) with recurrent and feedforward connections to both the excitatory and other inhibitory inspiratory circuit elements, shaping their discharge pattern. Inhibitory inspiratory neurons modulate the inspiratory-phase firing rates of tonic expiratory neurons (Fig. 6B, square 3) that inhibit pre-BötC and more caudal inspiratory neurons, thereby also influencing phrenic motor output. Divergent actions of inhibitory inspiratory neurons also suppress spiking in phasic augmenting expiratory (E-Aug-P) populations.
The RTN-pF modulation of the pre-BötC-VRG in this scheme includes six distinct categories of “indirect” actions through tonic expiratory neurons (Fig. 6B, square 4). Several chemoreceptor-evoked operations converge upon phasic decrementing expiratory (E-Dec) neurons to reduce and limit their activity (Fig. 6B, operations 3, 4, and 5). Neurons with this discharge pattern are widely considered to play a role in controlling expiratory phase duration (e.g., Hayashi et al. 1996; Rybak et al. 2004).
Recent studies have proposed that RTN-pF E-Aug-P neurons, distinct from inhibitory BötC neurons (Fedorko et al. 1989; Jiang and Lipski 1990; Lindsey et al. 1989; Merrill and Fedorko 1984), are involved in generating active expiration (Molkov et al. 2010; Pagliardini et al. 2011), complementing prior related observations on expiratory neuron discharge patterns during the expiration reflex and their connectivity (Baekey et al. 2004). Although this study did not specifically address hypotheses on the control of expiratory drive, we did identify correlogram features consistent with excitatory and inhibitory actions of tonic RTN-pF neurons (Fig. 6B, square 5) upon phasic augmenting expiratory neurons in the caudal VRG, a region containing bulbospinal neurons that drive active expiration (Baekey et al. 2004; de Almeida et al. 2010; De Troyer et al. 2005; Iscoe 1998). We note that while there is strong evidence for excitatory glutamatergic projections of chemoresponsive and putative chemoreceptor RTN-pF cells to the BötC-VRG (Abbott et al. 2009; Mulkey et al. 2004), the inhibitory interactions in the present results suggest other parallel projections that are modulated by chemoresponsive circuitry within the RTN-pF (Ott et al. 2011) or sign-changing multisynaptic connectivity.
Consideration of Methods
The decerebrate cat model used in this study avoids the confounding effects of anesthetics and has ventilatory responses to hypercapnia similar to those in the awake intact cat, although suprapontine influences are absent (Tenney and Ou 1977). We identified chemoresponsive neurons with selective stimulation of central chemoreceptors (Arita et al. 1988a, 1988b). We evaluated responses and correlations only in animals with an enhanced peak amplitude of integrated phrenic nerve activity following stimulus onset, an established metric for identifying a change in inspiratory drive with chemoreceptor stimulation in animals vagotomized to eliminate the effects of pulmonary afferent feedback upon the medullary respiratory network (see, e.g., Clark and von Euler 1972; Hwang et al. 1983; Scott 1908).
Our multiarray approach with submicron depth adjustment for each electrode is an effective method for monitoring neurons in widely distributed sites deep within the brain stem with high temporal resolution. By recording many neurons simultaneously, we identified extended correlational linkages of neurons assessed during the same pre- and poststimulus conditions with the goal of defining simple circuit models that reproduce experimentally observed features (Aertsen et al. 1989). Advantages and limitations of the approach have been considered elsewhere (Nuding et al. 2009a; Ott et al. 2011; Shannon et al. 2000).
Diversity of Correlogram Features
The present data set included offset correlogram peaks and troughs with half-widths ranging from 0.7 ms (Fig. 3C, 20) to >100 ms and is consistent with actions of diverse neurotransmitters and receptors associated with different categories of neurons in the VRC (Alheid and McCrimmon 2008; Haji et al. 2000; Stornetta 2008). Glutamate is a predominant excitatory transmitter found in inspiratory interneuron and premotoneuron populations. GABA and glycine both act as inhibitory neurotransmitters in other types of inspiratory and expiratory populations (Bongianni et al. 2010; Shao and Feldman 1997). Phasic BötC E-Aug and E-Dec or “postinspiratory” neuron populations use glycine as an inhibitory transmitter (Büsselberg et al. 2003; Ezure et al. 2003; Schreihofer et al. 1999). GABA acts as a gain modulator limiting both control and reflex-evoked activity of inspiratory and expiratory neurons and, via distinct receptors, contributes to the silent phase of phasic neurons (Tonkovic-Capin et al. 2003; Zuperku and McCrimmon 2002). The present observation of functional inhibition of inhibitory inspiratory neurons by other inspiratory neurons (e.g., neuron 816 in Fig. 2) is consistent with the concept that glycinergic inspiratory neurons act on both excitatory and other glycinergic inspiratory neurons (Winter et al. 2009).
Differential Modulation of Nodes Within the Inspiratory Neuron Chain
When considered together with prior work (reviewed in Segers et al. 2008), the results support a circuit architecture with an excitatory inspiratory chain (Fig. 7A, green populations) embedded within a partly hierarchical structure that incorporates both feedforward and recurrent inhibition by other inspiratory neurons (Fig. 7, B and C, pink populations). In this model, the inhibitory inspiratory neurons participate in several temporally overlapping tuning functions, including adjustment of burst onset time and the slope of augmenting burst ramps in other inspiratory neurons, as well as directly moderating chemoreceptor-evoked increases in firing rates and inspiratory drive. Notably, chemoreceptor-evoked changes in firing rate included reduced activity in some pre-BötC-VRG inspiratory neurons and increased activity in others (Fig. 7, arrows). In this regard, the present results are reminiscent of the differential modulation of rostral I-Driver and caudal inspiratory neurons in response to selective peripheral chemoreceptor stimulation (Morris et al. 1996a, 1996b, 2000; Morris and Gozal 2004).
Central peaks detected in correlations of some inspiratory neuron pairs are consistent with the effects of shared synaptic drive from antecedent sources. These sets of mutually correlated inspiratory neuron pairs suggest the operation of several distinct excitatory (or inhibitory) influences acting upon the cluster. Functional heterogeneity of our sample of VRG inspiratory neurons is likely. For example, neurons without short-latency offset peaks in the phrenic STAs may include somatic motoneurons innervating laryngeal muscles or their premotor drivers (Baekey et al. 2001), as well as preganglionic vagal motoneurons (McAllen and Spyer 1978).
Tonic expiratory neurons provide a “reservoir” for inspiratory drive modulation. The enhanced respiratory modulation of tonic E neurons during central chemoreceptor stimulation—presumably attributable in part to increased inhibition by inspiratory neurons relative to control conditions—diminishes the inspiratory-phase inhibition of inspiratory neurons, thereby contributing to the observed augmented inspiratory drive (Fig. 7, C and D). We identified several instances of short-latency positive-lag offset troughs in both phrenic signal averages and multiple cross-correlograms with inspiratory target neurons, all triggered by spikes of the same tonic expiratory neuron. These results suggest divergent actions of individual tonic expiratory neurons upon clusters of inspiratory neurons, perhaps amplified by spike synchrony among expiratory neurons with common targets (e.g., neurons 847 and 879 in Fig. 3).
When considered together with prior evidence of a role for tonic expiratory neurons in baroreceptor reflex inhibition of phrenic motoneurons and inspiratory drive (Lindsey et al. 1998), the present results support the concept that tonic VRC expiratory neurons constitute a node for converging afferent systems that regulate inspiratory drive intensity and thus tidal volume (Fig. 7E). These tonic expiratory neurons may also contribute to the well-known tonic inhibitory bias on inspiratory activity under conditions of hypocapnia (Nuding et al. 2009b; Sears et al. 1982).
The correlation linkage map in Fig. 5F complements previously reported pairwise relationships between chemoresponsive RTN-pF neurons (Ott et al. 2011). Two of the correlograms represented in that map include offset “peak-trough” sequences (31 and 37 in Fig. 5D). Simple interpretations of this feature include 1) dual actions of a particular transmitter or cotransmitter (Stornetta 2008) and 2) connections composed of excitatory and multisynaptic inhibitory actions evoked by the same trigger neuron. The multiple correlations among a trio of neurons in the circuit (correlograms 31 through 33, Fig. 5D) are simply interpreted as evidence for the second conjecture: Neuron 402 excites neurons 427 and 416, with neuron 416 also inhibiting cell 427. The offset peak superimposed upon a broader central peak in correlogram 38 (Fig. 5D) is consistent with shared synaptic influences amplifying an excitatory RTN-pF-to-BötC-VRG interaction.
Collectively, the correlogram features in Fig. 5 suggest distributed circuit mechanisms that include interactions between RTN-pF cells and more caudal tonic VRC expiratory neurons (Fig. 7F). As considered previously, parallel paths in the respiratory network (e.g., Lindsey et al. 1987; Nuding et al. 2009a; Segers et al. 2008), including circuits mediating chemoreceptor modulation of breathing (Abbott et al. 2009; Nuding et al. 2009b; Song and Poon 2009; Spyer and Gourine 2009), provide a substrate for multiple regulatory mechanisms and enhanced system robustness through redundancy—the duplication of critical components—and degeneracy, where different operations yield similar changes in output (Tononi et al. 1999).
This study was supported by National Institute of Neurological Disorders and Stroke Grant R37-NS-19814.
No conflicts of interest, financial or otherwise, are declared by the author(s).
Author contributions: M.M.O., S.C.N., K.F.M., and B.G.L. conception and design of research; M.M.O., S.C.N., K.F.M., and B.G.L. performed experiments; M.M.O., S.C.N., L.S.S., R.O., and B.G.L. analyzed data; M.M.O., S.C.N., L.S.S., K.F.M., and B.G.L. interpreted results of experiments; M.M.O., S.C.N., L.S.S., and B.G.L. prepared figures; M.M.O., S.C.N., L.S.S., and B.G.L. drafted manuscript; M.M.O., S.C.N., L.S.S., K.F.M., and B.G.L. edited and revised manuscript; M.M.O., S.C.N., L.S.S., R.O., K.F.M., and B.G.L. approved final version of manuscript.
We thank Peter Barnhill, Kimberly Ruff, Kathryn Ross, and Andrew Ross for excellent technical assistance.
- Copyright © 2012 the American Physiological Society