Fast respiratory rhythms include medium- (MFO) and high-frequency oscillations (HFO), which are much faster than the fundamental breathing rhythm. According to previous studies, HFO is characterized by high coherence (Coh) in phrenic (Ph) nerve activity, thereby providing a means of distinguishing between these two types of oscillations. Changes in Coh between the Ph and hypoglossal (XII) nerves during the transition from normal eupnic breathing to gasping have not been characterized. Experiments were performed on nine unanesthetized, chemo- and barodenervated, decerebrate adult rats, in which sustained asphyxia elicited hyperpnea and gasping. A gated time-frequency Coh analysis was developed and applied to whole Ph and medial XII nerve recordings. The results showed dynamic Ph-Ph Coh during eupnea, including MFO and HFO. XII-XII Coh during eupnea was broadband and included four distinct peaks, with low-frequency Coh dominating the epochs preceding the onset of Ph activity. During gasping, only MFO-peaks were present in Ph-Ph Coh. Bilateral XII activity showed a significant reduction in Coh and a shift toward lower frequencies during gasping. In contrast, contralateral Ph-XII Coh progressively increased during state changes from eupnea to gasping, a tendency mirrored in the startup part of the Ph activity. These data suggest significant hypoxia/hypercapnia-induced alterations in synchronization between respiratory outputs during the transition from eupnea to gasping, reflecting a reconfiguration of the respiratory network and/or alterations in the circuitry associated with the motor pools, including dynamic coupling between outputs.
- respiratory patterns
- zero-interval subtraction analysis
- motor coupling
- network plasticity
previous studies have used coherence (Coh) analysis to test for the existence of potential common sources of rhythmic drive to multiple outputs in the respiratory control, and other, systems (1, 3, 6, 9, 13, 19, 20, 30, 32). Some have shown that medium frequency oscillations (MFO) in different respiratory nerves is characterized by significantly lower and broadband Coh compared with high frequency oscillations (HFO) (3, 4, 5, 6, 7, 20). Fast respiratory rhythms, particularly HFO, are very sensitive to anesthesia (7, 15, 16). Unfortunately, there are few published observations regarding the coupling of fast rhythms in respiratory outputs during hypoxia in unanesthetized decerebrate cats in vivo (27, 39) or rats in situ (17). These studies concluded that HFO Coh between phrenic (Ph) nerves during gasping became more prominent and was shifted to higher frequencies. However, methodological issues complicate the findings in these studies. For example, the animals in the cat studies did not show the classic “primary apnea” (i.e., the silent phase) that normally precedes gasping (10, 12, 26). In situ artificially perfused precollicularly decerebrated juvenile rats show considerably lower power spectral frequency components compared with adult rats (17, 20, 33, 37; see companion paper Ref. 21). Thus, there remains a need for Coh analysis of adult, in vivo rat motor respiratory outputs in a clearly defined model of gasping.
In the accompanying study (21), we detailed the dynamics of the frequency components for two major respiratory outputs, the Ph and hypoglossal (XII) nerves, under three different conditions (eupnea, hyperpnea, and gasping). We paid particular attention to the major fast respiratory rhythms, MFO and HFO. In this study, we address the dynamic Coh data (i.e., within the breath) during different behavioral states. We take a higher-level systems view of the data presented in the companion paper and ask the questions: “How well are the various power spectral features coordinated among the two Ph or two XII nerves or between the Ph and XII nerves, and how does Coh vary over the course of the breath?” Toward this end, we developed our own nonparametric time-frequency Coh (TFC) analysis based on the zero-interval subtraction (ZIS) method, described in the accompanying paper (21), to analyze dynamic spectral Coh between the various nerve pairs. In addition, we address the question, “How does coherence between the nerve pairs change relative to output behavior in the three states?” Because we are interested in the behavior of the “core” of the central nervous system respiratory rhythm generator, the experiments were performed on unanesthetized, decerebrate, peripherally baro- and chemodenervated adult rats. By performing this comparative analysis and combining this with the results gained from the companion study, we develop a conceptual model of the mechanisms underlying the coordinated respiratory rhythm generation in the rat, in the absence of peripheral feedback effects.
Experiments were performed on the same nine adult, male Sprague-Dawley rats (340–380 g) as in the companion paper (21), in which all details regarding general surgical preparation, decerebration, electrophysiological recording, and gasp induction, are provided. All procedures were approved by the University of Delaware Animal Care and Use Committee. Briefly, under isoflurane anesthesia (5% induction, 1–2% maintenance), intravenous and intra-arterial cannulas were inserted, animals were intubated below the larynx, sinoaortic denervated, and vagotomized bilaterally. The phrenic and (medial branch of the) hypoglossal nerves were dissected free of the surrounding tissue and transected, and electrical activity was recorded from their central ends, along with arterial pressure, expiratory CO2, and lung inflation pressure. Animals were then paralyzed and artificially ventilated, and a bilateral pneumothorax was performed. They were then decerebrated precollicularly, and after 1 h of recovery and stabilization time, monophasic recordings of Ph and XII activity under eupneic, hyperpnea, and gasping conditions commenced, the latter two induced by asphyxia.
As described in the companion paper (21), time tags were generated from the integrated (τ = 50 ms) Ph and XII neurograms to identify the beginning and endpoints of inspiratory bursts, and all raw data were digitally filtered by a low-pass 300- and high-pass 10-Hz filter with stop band 450 Hz (40-dB attenuation at 345 Hz). This allowed us to perform “gated” Coh analysis on the bursts only (or the expiratory periods only, as controls; see below).
Because of the inadequacy of available tools when applied to large data sets with our frequency resolution requirements, we developed our own TFC analysis method (available upon request), running in the Matlab (The Mathworks, v. 7.0.4) environment. This method was derived from the ZIS time-frequency representation algorithm, detailed in the accompanying paper (21). In its present application, the Coh (at all investigated frequencies) at each time point was computed as the difference (Fig. 1B, ΔCoh) between two Coh estimates (1) one computed from a pair of signals (S1 and S2), and the other computed from the same two signals, except that a small segment of data values in one signal are replaced by zeros (S20(t)) centered at a specific time point, t: (2) (3) The difference between these Coh functions, ΔCoh, is the Coh at time = t (see also materials and methods in companion paper). Then the window is slid to the next position (50% overlap with the previous position), and a new ΔCoh calculated (Fig. 1B). This method was used to estimate TFCs of nerve pairs. TFCs may then be time-averaged (i.e., collapsed in the time dimension) to yield Coh as a function of frequency (see Fig. 1D, and results) or frequency-averaged (i.e., collapsed in the frequency dimension) to yield Coh as a function of time for all frequencies (∼10–300 Hz). We used 2.44 Hz/bin resolution in all analyses and a 10-kHz data sampling rate (FFT window was set to 4,096 points, see Ref. 21 for details), and TFCs are presented as interpolated isocontour plots.
To verify our method, we produced two signals (S1 and S2; 10,000 points/s), shown in Fig. 1A. S1 consisted of a sine wave with a frequency of 77 Hz, while S2 was time-varying, with epochs of pure 14 or 77 Hz sine waves (50% of the time spent in each), the latter phase-shifted compared with the S1 signal. Fig. 1C shows the TFC function produced by our ZIS-based algorithm, schematically depicted in Fig. 1B. Time-averaging the TFC produces the Coh estimate shown in Fig. 1D, which demonstrates a sharp peak at 77 Hz, only slightly underestimating (0.45) the theoretical Coh value (0.5) at this frequency. On the other hand, parametric FFT-based methods produce overestimation of Coh at this frequency (∼0.65; Fig. 1D). As with the TFRs in the accompanying paper (21), the time axis of the TCFs is normalized over the burst onset (t = 0) to offset (t = 1) to average breaths of varying lengths appropriately. In addition to analysis of the entire XII burst, the portion of the XII burst concurrent with the Ph burst (Ph-related), and the portion that preceded the Ph burst (pre-Ph) were analyzed separately. Only the former was used for Ph-XII Coh analysis. In showing averaged Coh functions for Ph-XII analyses (see ⇓⇓Fig. 4), “contralateral” data included both pairs (left Ph, right XII, and right Ph, left XII), as did ipsilateral data (right Ph, right XII and left Ph, left XII).
The statistical analyses were described in detail in our companion paper (21) for power spectra analyses. Briefly, TFCs were grouped into either a eupneic, hyperpneic, or gasping category and averaged over all animals (n = 9) to identify consistent trends. Averaged Cohs were reconstructed from the TFCs' results and compared with values estimated from the expiratory phase of nerve activity used as “background”. Depending on the distribution of the data, either parametric or nonparametric tests were applied. Results are presented as means ± SE, and a P value < 0.05 obtained from any statistical tests was interpreted as significant, as was a 95% confidence level. The borders of frequency bands and their maximal peaks were identified using a custom-written optimization algorithm, running in Matlab and described in detail in our companion paper (21). In addition to across-animal analysis, individual animal TFCs were generated to assess interanimal variability.
The results reported are based on analysis of the same 928 eupneic, 386 hyperpneic, and 186 gasping bursts, produced by nine rats that were analyzed in the accompanying paper (21). Coh between left-right pairs are hereafter referred to as Ph-Ph and XII-XII, while heterogeneous pairs are denoted by Ph-XII. In addition to total XII-XII Coh, Coh between phrenic-related (Ph-r) and prephrenic (pre-Ph) parts of the hypoglossal activity was analyzed separately.
Figure 2A1 (thin lines) shows the reconstructed Ph-Ph Coh for individual animals (n = 9) during eupnea. The across-animal ZIS time-averaged Coh between left and right phrenic nerves during eupneic bursts is also shown in Fig. 2A1 (thick trace), with P values given in the lower plot. The major features were, with some exceptions, consistent across animals: one peak each in the MFO and HFO band. There was statistically significant Coh over a wide range of frequencies, including those categorized as MFO and HFO. In addition, local peaks were evident at the same frequency locations as the major peaks in power spectral activity during the same eupneic bursts (75–130 and 137–212 Hz, Table 1; ⇓Fig. 6A1 of Ref. 21), with a maximum Coh of 0.36 at 186 Hz. Unlike the power spectral distribution for the same nerves and condition, the magnitude of the HFO Coh is considerably greater than the MFO Coh, due in part to the diminished Coh in this band in four animals (Fig. 2A1). The TFC for all eupneic breaths is provided by Fig. 2B1. This analysis demonstrates that the highest degree of Coh lies in the 160–220 Hz band (i.e., HFO) and that this is strongest during the first ∼60% of the burst. In latter portions of the breath, Coh is weaker and more widespread in terms of frequencies involved (Fig. 2B1).
Figure 2A2 displays the individual ZIS-derived Ph-Ph Coh during hyperpnea, and all show multiple peaks. The averaged Coh between the phrenic nerves shows considerably sharper peaks (Fig. 2A2) than during eupnea (Fig. 2A1), with statistically significant Coh (compared to expiration) at many frequencies that are characterized as MFO, HFO, and even UHFO (see Ref. 21). The dominant eupneic HFO Coh peak is higher (0.38) than during eupnea, while the MFO eupneic peak is diminished (although another MFO Coh peak emerges at ∼55 Hz). The TFC analysis in Fig. 2B2 reveals that some individual frequency bands (130–150 Hz) are temporally restricted to brief periods of the hyperpneic bursts, others (185–200 Hz) are maintained over most of the burst, and others (50–70 Hz) are present early, wane in the middle, and reappear later. This latter band is a feature that corresponds to that in the typical Ph TFR power spectrum (MFO startup) illustrated in our companion paper (see Fig. 4B2 and C2 of Ref. 21) under the same conditions.
Figure 2A3 demonstrates the consistency between the time-averaged TFC across all seven animals that produced >10 gasps. Specifically, the majority of significant Coh is contained within two bands (MFO1 and MFO2). Both of these bands correspond to peaks in the power spectra of Ph activity in this same state (Fig. 4A3 of Ref. 21). The dynamics of the Coh in these bands (Fig. 2B3) also reflects those in the power spectra (Fig. 4B3 of Ref. 21), with the higher of the two bands (76–93 Hz) appearing early and waning over the breath, and the lower (29–54 Hz) showing significant Coh early in the burst, waning, and recovering by 25% of the breath, waning again, and returning to previous levels from the 60–80% mark of the burst.
Figure 3A1 provides the reconstructed XII-XII Coh for individual animals during the entire eupneic burst and demonstrates remarkable consistency in the location and number of Coh peaks. Figure 3B1 displays the average Coh between the left and right XII activity for the entire burst (black), as well as that for the pre-Ph (blue) and Ph-r (red) portions of the burst (see also Tables 2 and 3). The P value plot identifies a wide band of significantly elevated Coh in frequencies spanning from ∼70 to 240 Hz distributed among several peaks, and reaching maxima similar to maximum Coh levels between Ph nerves (∼0.37). In addition, there is a low-frequency band (20–51 Hz), which demonstrates low, but significant, Coh at the very beginning of the burst.
The TFC shown in Fig. 3C1 characterizes the dynamic nature of the Coh between the XII outputs during eupnea, particularly with respect to the pre-Ph vs. Ph-r portions of the burst (Ph onset indicated by the vertical black line). For example the ∼75–105 Hz band demonstrates very high Coh during most of the pre-Ph period, with more modest and short-lived Coh during the Ph-r activity (Tables 2 and 3). On the other hand, there is significant Coh in the 120–155 and 175–210 Hz bands both during both pre-Ph and Ph-r bursting. Finally, very high frequency (210–247 Hz) Coh is present only during the Ph-related burst.
Figure 3A2 illustrates the relative consistency in ZIS-derived XII-XII Coh between different animals during hyperpnea, with larger variability in the low frequencies (30–60 Hz). Average total Coh (Fig. 3B2, black trace) was significant at almost all frequencies between 70 and 200 Hz (upper P value subplot). The highest Coh was present in the 70–100 Hz range, and this occurred predominantly during the pre-Ph activity (blue trace, and Table 3). Coh in the 25–50 Hz frequency band was evident during pre-Ph activity, reaching higher levels (0.22) than during eupnea (0.16). The TFC (Fig. 3C2) revealed the dynamics of somewhat disjointed bands. The pre-Ph component of XII activity, approximately two-thirds of the total burst duration under hyperpneic behavior, contains numerous scattered bands of Coh, particularly in the 75–100 Hz frequencies, which reappear only modestly during the Ph-r portion (see Table 2). In addition, the TFC shows narrow bands of high Coh (137 Hz peak) that begin in the last sixth of the pre-Ph activity and continue through much of the Ph-r period (Fig. 3C2) and significantly higher level of 175–200 Hz band in Ph-r part (Tables 2 and 3).
XII activity in individual animals was characterized by many lower-Coh peaks, mostly below 150 Hz (Fig. 3A3), with a higher degree of variability compared with the other two states. This caused the across-animal averaged XII-XII Coh to be highly fractured (Fig. 3B3). In particular, peaks at 98 and 117 Hz were evident (Table 2), and both of these lie within a major peak in the XII power spectrum under gasping conditions (Fig. 5A3 of Ref. 21). Overall, XII Coh was considerably lower in this condition than in eupneic or hyperpneic states (see below for interstate quantitative analysis). The TFC (Fig. 3C3) reveals the dynamic nature of the Coh, with the significant bands starting within the first 10% of the burst and continuing some 50–60% through its duration.
To assess the changes in contribution of potential common sources of drive between the Ph and XII motor neuron pools under the three different conditions, Coh between the Ph and XII (Ph-related only) activity was analyzed. Both ipsilateral and contralateral Cohs were estimated, and the TFCs of the contralateral pairs are provided in Table 4.
The Ph-XII Coh was consistently highest between 170 and 200 Hz, as demonstrated by the plots of ipsilateral pairs in Fig. 4A1. In addition, a lower but consistently located high-Coh band was revealed between 130 and 150 Hz. A weak, but statistically significant, Coh band was detected between 80 and 100 Hz, but this was due to contributions from only two animals (Fig. 4, A1 and B1). Figure 4B1 illustrates the nature of the Coh between Ph and XII activity during eupneic breathing across all animals. Statistically significant bands had peaks of ∼81, 139, and 181 Hz in both ipsilateral (red) and contralateral (blue) nerve pairs (Table 4). Figure 4C1 demonstrates the time course of these relationships for the contralateral Ph-XII pairs. The dominant Coh band, ∼160–210 Hz, begins shortly after burst onset and continues unabated until ∼80% through its duration. The middle, weaker-coherent band (∼130–155 Hz) mainly exists in the latter portions (50–80%) of the burst. Finally, the Coh at low frequencies (peak at 81 Hz), when present, appears only in very late stages of the burst (65–85%).
The time-averaged Coh between contralateral Ph and XII activity during hyperpneic bursting is provided in Fig. 4A2, where overall Coh has increased compared with eupneic bursts. Again, ipsilateral (red) and contralateral (blue) pairs show similar Coh patterns (Fig. 4B2 and Table 4) to those during eupnea, if at higher absolute levels, with higher-frequency Coh appearing early in the burst, and lower-frequency Coh contributing later (Fig. 4C2). The overall intensity of the ∼165–210 Hz band is markedly increased and sharpened in hyperpnea compared with the analogous band during eupnea.
During gasping, a dramatic downward shift in frequencies of significant Coh was observed, both in the case of individual animals that produced >10 gasps (n = 7; Fig. 4A3) and across-animal averages (Fig. 4B3). Prominent peaks existed only at 37, 78, and 110 Hz, with no significant Coh above 120 Hz (Table 4). Although not dependent upon power per se, this result mimics the overall shift to lower frequencies in the power spectra of both nerves during gasping (21). The TFC shows the two lower Coh bands spanning the ∼10–70% mark of the burst. The higher Coh band (∼105–120 Hz) is transient, appearing only from ∼20 to 30% into the burst.
Comparative Analysis of States
To quantify the differences between Coh in different states and different nerve pairs, bin-by-bin pairwise comparisons between Coh values within the different states were made. The results of this analysis are shown in Fig. 5. P value subplots of the significance of the differences are color coded <0.05, depending on whether there is a gain (gray) or loss (white) in Coh when going from the first to the second state in the pair.
Figure 5A shows the time-averaged TFCs Ph-Ph Coh during eupnea (black), hyperpnea (blue), and gasping (red). Compared with eupnea, there is a significant gain in Coh in the 40–70 Hz band (gray in E and H) during hyperpnea, whereas several other higher-frequency bands show diminished Coh (white). Likewise, when comparing gasping to eupnea (E and G), there is as significant wideband reduction in Coh between 120 and 270 Hz (white) concomitant with an increase in Coh at low frequencies (30–50 Hz; gray). Finally, when the behavioral state changes from hyperpnea to gasping (H and G), the low-frequency Coh gains made when shifting from eupnea to hyperpnea are mostly lost (white, ∼50–70 Hz), while two bands flanking this diminished band show increases in Coh (gray in H and G plot). In addition, there is widespread loss of Coh at frequencies >120 Hz (white).
Figure 5B shows the same comparative analysis for left-right hypoglossal Coh during the Ph-related portion of the burst. Altering states from eupnea to hyperpnea results in loss of high-frequency Coh, mainly between ∼150 and 235 Hz (white in E-H subplot). Compared with eupnea, gasping produces a Coh loss of an even wider band of higher frequencies (E-G, white), with significant Coh increases in lower frequencies (20–25 and 35–75 Hz; gray). The characteristic changes in Ph-related XII Coh between the hyperpneic and gasping states include significant Coh increases in lower frequencies (∼50–75 Hz, H-G, gray), and selective Coh loss in higher, narrow frequency bands. Pre-Ph XII activity only exists during eupneic and hyperpneic conditions, and the comparison between these states is illustrated in Fig. 5D. Although the pre-Ph Coh bands are similar in the two states, there is a shift toward lower-frequency Coh during hyperpnea, expressed as loss at high frequencies and gain at lower ones (E-H).
The comparison of Coh between contralateral pairs of Ph-XII in the three states is provided by the plots in Fig. 5C. There is a significant Coh increase (gray in E and H P value plot) in medium and high frequencies above 100 Hz when state is altered from eupnea (black) to hyperpnea (blue). When eupnea advances to gasping (E–G), there are large gains in Coh at frequencies between 30 and 100 Hz (gray) and large losses in the 120–270 Hz range (white). When comparing hyperpnea to gasping (H-G), the same general trend exists, although the band of low-frequency gains is slightly more restricted, and the loss in high frequencies is over a wider range.
Direct Comparison Between Nerve Pairs
The Cohs between three different nerve pairs in the three different states are shown on Fig. 6. During eupneic bursts (Fig. 6A) there exists significantly different Coh between the Ph-Ph (black), Ph-XII (blue), and contralateral Ph-XII (red) pairs at different frequencies. Specifically, the XII-XII Coh is higher than the other pairs (Fig. 6A) in the 125–160 Hz band. During hyperpnea (Fig. 6B), the Ph-XII (red) showed consistently higher Coh than the Ph-Ph (black) or XII-XII (blue) Coh, with the notable exception of the large Ph-Ph Coh in the MFO band (50–75 Hz). During gasping (Fig. 6C), Ph-XII (red) and XII-XII (blue) showed consistently higher Coh levels than the Ph-Ph (black) at most frequencies. At frequencies below 100 Hz, the dominant Coh was the Ph-XII (red).
Comparison of Total Coh Among Pairs
To examine the changes in total Coh between the pairs during the three different states, TFCs were averaged over all frequencies (10–300 Hz) and then plotted on the relative time axes for different nerve pairs (Fig. 7). For left-right Ph pairs (solid line), the total (i.e., averaged over all frequencies) Coh changed from a rapid rise/plateau/incrementing pattern during the eupneic burst (Fig. 7A) to a gasping pattern (Fig. 7C), in which the initial, weak initial increase was continuously diminished over the final two-thirds of the burst, with a >50% decrease in peak Coh compared with eupnea. Interestingly, only the Ph-Ph Coh demonstrated a significant late increase in total Coh during eupnea (Fig. 7A), while the XII-XII (Ph-r, dashed) and XII-Ph (dotted) waned after their initial increases, only to “recover” to early levels by the end of the burst. During hyperpnea (Fig. 7B), all total Cohs were saddle-shaped (higher in the beginning and end than the middle of the breath). The weakest total Coh during eupnea was between XII pairs (dashed), while the Ph-XII total Coh was strongest during both hyperpneic and gasping bursts (Fig. 7, B and C, respectively). The left-right XII pairs (dashed line) contain similar total Coh levels as the Ph-Ph pairs (solid line) during hypoxia/hypercapnia only (Fig. 7, B and C).
This study is the first to characterize the intraburst dynamics of Coh between the major respiratory output nerves (Ph and XII) in the unanesthetized, decerebrate adult rat in vivo under normal and hypoxic/hypercapnic conditions. It is also the first to analyze the changes in Coh dymanics between the Ph and XII activity under different systemic states (as defined by output pattern). Coh between output nerves was used as a measure of functional coupling, which may result from either common sources of drive or from similarities in network organization. Using our ZIS technique (see Ref. 21), we developed an analytical technique that produces time-frequency Coh estimates with very short computation time. This allowed us to analyze thousands of inspiratory bursts, thereby allowing for data that could be statistically evaluated.
The important general findings of this study are that 1) Coh between nerve activity is dynamic within frequency bands over the course of the inspiratory period; and 2) substantial changes in Coh between different nerve pairs (Ph-Ph, XII-XII, and Ph-XII) occur during the transition from eupnea to gasping. These changes may reinforce the changes observed in dynamic power under these conditions (21), including a dramatic downward shift in frequencies containing significant Coh during gasping, compared with eupnea. Moreover, the pattern of Coh within certain frequency bands over the course of a burst displays temporal dynamics, which may be indicative of dynamics in the strength of coupling of common inputs to motor pools. Another major finding, in accord with the time-frequency spectral observations (21), is that our functional coupling analysis reveals that hyperpnea appears to be an intermediate state between eupnea and gasping, reflecting characteristics of each. In addition to the frequency dependence of the temporal Coh patterns under different breathing conditions within a given nerve pair, differences in TFC patterns among the different nerve pairs were also evident.
In general, our estimation of Coh during eupnea are well matched to the major power spectral features during the same state (21), with one exception: the MFO peak Coh (averaged among all animals) is smaller than the HFO peak (Fig. 2A1). The stronger HFO Coh indicates that the mechanisms responsible for the generation of HFO have stronger bilateral coupling than those responsible for MFO generation. The dynamics of the Coh differ from those of the power spectrum (Fig. 2B1 of this study vs. Fig. 4B1 of Ref. 21), specifically with regard to the time course of MFO Coh. Significant MFO power begins by the time 30–40% of the eupneic burst has occurred (peaking at 50–70%), whereas MFO Coh, if present, does not increase appreciably until some 70% of the breath duration and is sustained only briefly. This suggests that the functional coupling of Ph-Ph MFO exists only during periods of high activity/recruitment within the Ph motor neuron pool. Meanwhile, significant HFO Coh begins early and plateaus rapidly, like the power in this band (21), and fades by the time two-thirds of the burst has occurred, suggesting that HFO is produced by a consistently bilaterally coupled mechanism. Our results suggest that the bilateral coordination of Ph MFO under eupneic conditions is considerably weaker than HFO, and these relative strengths presumably reflect underlying rhythm generation and distribution mechanisms. Another finding is that of increased Ph-Ph Coh in relatively wide band at the end of inspiration (Fig. 2B1), reflected in the total Ph-Ph Coh (Fig. 7A, solid line), which happens to be concomitant with a significant increase in MFO power (Fig. 4B1 of Ref. 21). Total phrenic Coh during eupnea mimics the incrementing integrated activity shape and may reflect an increase in central synchrony, or an emergent property of increased motor unit recruitment and firing rate.
During hyperpnea, Ph-Ph Coh becomes more fractured in the frequency domain, with the emergence of a new Coh band in low-frequency MFO (Fig. 2, A2 and B2). The same features were found also in Ph power spectra (Fig. 4, A2 and B2 of Ref. 21). These changes may indicate the initiation of a systemic reconfiguration that progresses during prolonged hypoxia/hypercapnia. Despite the high amplitude of the main Coh peaks (MFO and HFO), relative to eupnea, the total Ph-Ph Coh during hyperpnea is lower than during eupnea (Fig. 7, A and B), representing a partial uncoupling (i.e., at some frequencies) of the left and right sides. Ph power, on the other hand, is much greater in almost all bands during hyperpnea compared with eupnea. These seemingly counterintuitive observations may be reconciled by a system designed for efficiency (i.e., of coordinating Ph output) during eupnea, specifically one that is more dependent on strong coupling (see upper P value plot of Fig. 7A). The increase in power during hyperpnea is concomitant with an increase in low-MFO coupling (Figs. 2, A2 and B2) and lower overall coupling (Fig. 7B), but the effect is still increased respiratory effort.
Although Ph activity during eupnea may be “optimized” by near-synchronous oscillations (preferentially in HFO), particularly to overcome muscular inertia present at the start of inspiration, it has been shown to be nonobligatory in forming normal respiratory outputs (see Ref. 11 for a review). Thus it has been suggested that the generation and maintenance of the basic rhythm does not include or require circuitry that produces HFO, but rather, HFO-generating mechanisms may be fused through the activation of additional circuits (2, 29). On the basis of power spectra (Fig. 4 of Ref. 21) and Coh analyses (Fig. 2, A3 and B3), phase-locked Ph MFO becomes the dominant oscillation band during gasping, and this may reflect a drastic network reconfiguration, evidence of which (e.g., emergence of low MFO power and Coh) can be found during hyperpnea. We propose that this reconfiguration may be due to the sensitivity of particular circuit elements to severe or prolonged hypoxia/hypercapnia (i.e., HFO-generating circuits/elements have higher O2 demand and shut down under anoxia). In addition, the decrementing pattern of the Ph total Coh waveform during gasping (Fig. 7C, solid) resembles that of Ph-integrated activity (see Ref. 21, Fig. 2C, bottom) with significant Coh only at the beginning of the burst. Thus it may be surmised that bilateral coupling at low frequencies is mainly useful during Ph startup. Our finding is in accord with the hypothesis that during gasping only, inspiratory initiation (and not termination) mechanisms exist, perhaps driven by pacemaker-like activity of medullary units, particularly as network inhibitory neurons are depressed during anoxia and are not required for the generation of gasps (34, 35, 36).
During eupnea, hyperpnea, and gasping, there is the same general trend in XII-XII Coh as in Ph-Ph Coh: increase in low-frequency bands and decrease in high-frequency bands (Figs. 3 and 5B). However, we observed a clear difference in Coh between the pre-Ph and Ph-r parts of XII discharge (Figs. 3 and Fig. 5D). Specifically, compared with eupnea, there is a marked increase in lower-frequency, and a decrease in higher-frequency, Coh during the pre-Ph portion of the hyperpneic XII burst (Fig. 5D). A difference in central mechanisms (e.g., separate vs. common premotor pools) underlying the distinction between pre-Ph and Ph-related XII discharge has been suggested (14, 28, 38), and our data support this overall concept by demonstrating differential coupling during these two portions of the burst.
Total XII-XII Coh of the Ph-related activity (Fig. 7, dashed line) shows significant Coh only briefly, at approximately the 10% and 65% marks of the eupneic burst (Fig. 8A, dashed line). During hyperpnea (Fig. 7B, dashed line), we observed a modest extension of the second significant Coh epoch. During gasping, there is only significant XII-XII total Coh in the period extending from 20 to 40% through the burst (Fig. 7C, dashed line). Both total XII-XII and Ph-Ph Coh follow a decrementing pattern during gasping, with only brief periods of significant levels in the first third of the burst. Overall XII-XII Coh is not sustained through the breath, but instead is concentrated during specific epochs. This partial bilateral independence implies the presence of independent general oscillators with strong coupling only in certain restricted bands and at certain times (Fig. 3).
We note that total Ph-Ph and XII-XII Coh are very close in shape, amplitude, and time course during hyperpnea and gasping but not during eupnea (Fig. 7, solid vs. dashed lines). It is well known that genioglossal muscles are multifunctional and can be involved in different functional acts such as chewing, drinking, swallowing, and vocalization, implying that common respiratory control of these muscles during eupnea may not be physiologically necessary. During hyperpnea and gasping, when the main goal of the system is to compensate for lower Po2, there is motivation to incorporate genioglossal muscles into the common respiratory output, leading to decreasing upper airway compliance. This cybernetic goal may explain the increase in total XII-XII Coh during hyperpnea and gasping compared with eupnea (Figs. 3A2 and 7B).
Coh between Ph and Ph-related XII did not show any significant difference between ipsilateral and contralateral pairs for any state (Fig. 4). During eupnea, Ph-XII Coh contained three main peaks with related frequency bands corresponding to MFO, HFO, and one at ∼150 Hz (Fig. 4A1 and 4B1), with HFO containing the highest Coh. During hyperpnea, the amplitude of HFO Coh increased, particularly at HFO frequencies, and additional UHFO peaks were detected (Fig. 4, A2 and B2). In contrast, all major Coh peaks were located in the MFO during gasping (Fig. 4, A3 and B3).The total Ph-XII Coh analysis reveals a progressive increase in total coupling strength starting with eupnea and progressing through hyperpnea and gasping (Fig. 7, dotted lines; also, Figs. 5C and 6), with the highest Coh values associated with the gasping state (Fig. 7C, bottom P value plots). Interestingly, there is a dramatic increase in HFO Ph-XII Coh during hyperpnea relative to Ph-Ph and XII-XII and an increase in MFO Coh during gasping (Fig. 6). Taken together, our results imply that functional coupling between phrenic and hypoglossal musculature is increasingly expressed in the kernel or core system as anoxic stress increases.
The increased coupling between Ph and XII outputs as anoxia persists suggests a common oscillator(s) driving both motor neuron pools, consistent with increased incorporation of genioglossal muscles in breathing function (see above). There is anatomical evidence for premotor neurons common to both Ph and XII nuclei located in the ventrolateral portion of the nucleus tractus solitarius and regions dorsomedial to the nucleus ambiguus in the cat (22), ventrally in the medial medullary reticular formation of the ferret (31), and among the bulbospinal neurons in the rat (18). It is not known whether their synaptic efficacy is enhanced during gasping. However, other anatomical (8) and electrophysiological (24, 25) studies suggest separate premotor neurons for Ph and XII motoneurons.
In conclusion, the results presented in this study suggest that coupling, even within the major oscillatory bands, is dynamic, indicating motor output coupling is nerve pair- and condition-dependent. Considering the results from both the accompanying study (21) and this one, we offer a conceptual model, provided in Fig. 8, of the state-dependent dynamics of fast rhythm generation and Coh in the rat respiratory control system. It should be noted that our analysis assumes that the system is stationary within each state, and this may not be true, particularly for intermediate states such as hyperpnea. Nonetheless, each state produced characteristic TFC patterns that were internally consistent. In the eupneic state (Fig. 8, left) the primary rhythm generator (double sinusoid trace) activates medullary startup-associated, low MFO- (light portion of arc, beginning of inspiration, *), and HFO-generating (dark part of arc, first half of inspiration) networks. Outputs of this network are distributed to separate premotor neuronal pools [bulbospinal neurons (Fig. 8, BS) and XII premotor (pMN)] or to common premotor neurons (not shown), either of which may provide the basis for Ph-XII Coh. Given that spinal hemisection mainly affects HFO Ph Coh (11), the local circuitry (Fig. 8, gray ovals) may serve as a major source of MFO generation during the latter portions of inspiration and is partially coupled bilaterally (overlapping of gray ovals), which may account for Ph-Ph and XII-XII Coh. During eupnea, the central generator of the fast respiratory rhythms produces predominantly HFO (Fig. 8, left), but during gasping, predominantly MFO (Fig. 8, right), assuming little or no decrementing ramp in Ph activity (21). Under severe hypoxic/hypercapnic conditions that trigger gasping, MFO may be produced by an entirely different mechanism. Because Ph-XII MFO Coh is present during gasping, a common brain stem mechanism is likely. The very early, brief MFO power and Coh present during eupnea becomes more strongly expressed, as severe hypoxic/hypercapnic conditions persist (Fig. 8, expansion of *). We propose that a quiescent, low-MFO-producing brain stem oscillator is present during eupnea and activated during gasping, which is in agreement with recent descriptions that early inspiratory neurons are active during eupnea and gasping (23). Further study will be required to elucidate which neurons contribute to gasping, and their properties under other functional states. The results in this and the companion study (21) promote the hypothesis that coherent fast oscillations are produced by functionally coupled networks that provide common inputs to local motor neurons or that the local networks themselves (with similar functional architecture across animals) are responsible for the frequency characteristics of the motor pools, or both.
This work was supported by National Institutes of Health Grant R01 HL68143 (to R. F. Rogers)
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2006 the American Physiological Society