A negative interaction between brainstem and peripheral respiratory chemoreceptors modulates peripheral chemoreflex magnitude

Abstract

Interaction between central (brainstem) and peripheral (carotid body) respiratory chemosensitivity is vital to protect blood gases against potentially deleterious fluctuations, especially during sleep. Previously, using an in situ arterially perfused, vagotomized, decerebrate preparation in which brainstem and peripheral chemoreceptors are perfused separately (i.e. dual perfused preparation; DPP), we observed that the phrenic response to specific carotid body hypoxia was larger when the brainstem was held at 25 Torr P_CO2 compared to 50 Torr P_CO2. This suggests a negative (i.e. hypo-additive) interaction between chemoreceptors. The current study was designed to (a) determine whether this observation could be generalized to all carotid body stimuli, and (b) exclude the possibility that the hypo-additive response was the simple consequence of ventilatory saturation at high brainstem P_CO2. Specifically, we tested how steady-state brainstem P_CO2 modulates peripheral chemoreflex magnitude in response to carotid body P_CO2 and P_O2 perturbations, both above and below eupnoeic levels. We found that the peripheral chemoreflex was more responsive the lower the brainstem P_CO2 regardless of whether the peripheral chemoreceptors received stimuli which increased or decreased activation. These findings demonstrate a negative interaction between brainstem and peripheral chemosensitivity in the rat in the absence of ventilatory saturation. We suggest that a negative interaction in humans may contribute to increased controller gain associated with sleep-related breathing disorders and propose that the assumption of simple addition between chemoreceptor inputs used in current models of the respiratory control system be reconsidered.

Since the initial work of Gesell et al. (1940), many contradictory reports have been published on how inputs from central and peripheral respiratory chemoreceptors are coordinated to control breathing. In non-rodent preparations, the mathematics of this interaction has been reported to be: (a) additive, where the two receptor compartments do not interact in any significant way (e.g. van Beek et al. 1983; Daristotle & Bisgard, 1989; Clement et al. 1992, 1995; St. Croix et al. 1996); (b) hyper-additive, where activation of one receptor increases the gain of the other (i.e. positive; e.g. Adams et al. 1978; Robbins, 1988); and (c) hypo-additive, where decreased stimulation of one receptor increases the gain of the other (i.e. negative; e.g. Gesell et al. 1940; Tenney & Brooks, 1966; Ou et al. 1976; Berger et al. 1978; Giese et al. 1978; Eldridge et al. 1981; Adams & Severns, 1982; Smith et al. 1984). Furthermore, the contribution of the carotid bodies to the overall P_CO2 ventilatory response has been reported to range from insignificant (Fencl et al. 1966) up to 60% (e.g. Bellville et al. 1979; Berkenbosch et al. 1979; Heeringa et al. 1979; Daristotle & Bisgard, 1989; Honda, 1992; Pan et al. 1998; Rodman et al. 2001; Fatemian et al. 2003). If the contribution of central and peripheral chemoreceptors is other than additive, then it follows that the contribution of one chemoreceptor to the overall chemoreflex will depend in part on the activity of the other chemoreceptor (Adams & Severns, 1982). Thus, the large discrepancy in previous reports of the relative contribution of the carotid bodies may be reconciled in part by the demonstration of a negative interaction between central and peripheral chemoreceptors: the level of stimulation of one compartment affects the contribution of the other to chemoresponses.

Currently, all models of the human respiratory control system assume simple addition between chemoreceptor inputs (Khoo et al. 1991; Painter et al. 1993; Ursino et al. 2001; Longobardo et al. 2002; Duffin & Mahamed, 2003; Topor et al. 2004). However, due to the multiplicative relationship between frequency and tidal volume in generating minute ventilation, simple addition between central and peripheral chemoreceptors in generating , as demonstrated by some investigators, necessitates that a negative interaction must exist in the contribution of chemoreceptors to frequency and/or tidal volume (Mitchell et al. 1990; Day & Wilson, 2007).

Using a novel in situ dual perfused preparation (DPP) for rodents, we previously demonstrated that a single bout of specific carotid body hypoxia (60 Torr P_O2) elicited larger phrenic responses when the brainstem was held at 25 Torr P_CO2 compared to 50 Torr P_CO2. Thus, phrenic responses to one level of carotid body hypoxia while at two levels of brainstem P_CO2 suggest there is a negative chemoreceptor interaction (Day & Wilson, 2007). Here, we test whether this negative interaction is generalizable to a broad range of carotid body chemostimuli, including those above and below eupnoeic levels. Specifically, we investigated phrenic responses to random perturbations of specific carotid body oxygen (isocapnic) and carbon dioxide (isooxic) under three different levels of steady-state brainstem P_CO2 (25, 35 and 50 Torr).

Our findings suggest the integration of central and peripheral respiratory chemoreceptors in the control of breathing involves a negative interaction regardless of whether the carotid body perturbation results in an increase or decrease in activation from that at eupnoeic levels. We conclude that the ventilatory response to carotid body stimulation from both P_CO2 and P_O2 perturbations (i.e. peripheral chemoreflex magnitude) depends largely on the level of brainstem P_CO2.

Methods

In situ dual perfused preparation (DPP)

We modified the working heart–brainstem preparation (WHBP; Paton, 1996; Wilson et al. 2001), making it appropriate for studies of respiratory chemoreceptor interaction. We will refer to our preparation as the in situ dual perfused preparation (DPP; Day & Wilson, 2005). Our modifications include: (a) separate perfusion of brainstem and peripheral chemoreceptor compartments, (b) clamping of descending aorta perfusion pressures at ∼90 mmHg via computer feedback control, (c) vagotomy, and (d) normokalaemic perfusate.

Dissection and perfusion system

Experiments were conducted using juvenile male Sprague–Dawley albino rats (Charles River, Quebec, Canada; 120–150 g, ∼4–6 weeks old) and 48 animals were included in the study. Experimental procedures were approved by the University of Calgary Animal Care Committee and were in accordance with national guidelines. Procedures were identical to a previous study (Day & Wilson, 2005). Briefly, rats were heparinized (1500 units i.p.) ∼15–30 min prior to being anaesthetized for dissection. Deep anaesthesia was induced using an overdose of halothane via inhalation. Adequate anaesthesia was assessed by testing for absence of response to noxious tail pinch. In rapid succession, rats were transected subdiaphragmatically and decerebrated at the midcollicular level. The decerebration was performed with a vertical cut using a new scalpel, removing all tissue rostral and dorsal to the colliculi. Transection and decerebration were performed in cold perfusate (∼5–8°C, volume: 500 ml) containing (in mm) 115 NaCl, 24 NaHCO₃, 4 KCl, 2 CaCl₂, 1 MgSO₄, 1.25 NaH₂PO₄ and 10 dextrose, equilibrated with 95% O₂–5% CO₂. The skin was removed and the preparation was placed into an experimental chamber and secured with ear bars. The descending aorta was canulated with a double-lumen catheter. One lumen of the catheter was connected to a peristaltic pump (Minipuls 3, Gilson Middleton, WI, USA) and used to perfuse the descending aorta retrogradely with room temperature perfusate containing 1.33% Ficoll Type-70 (Sigma-Aldrich F-2878) equilibrated with 40 Torr P_CO2 in O₂. The other lumen was attached to a pressure transducer to monitor perfusion pressure. Once canulated, the speed of the peristaltic pump was increased so as to increase perfusion pressure to ∼90 mmHg using a custom-built computer-controlled feedback system. A bilateral vagotomy was performed at the mid-cervical level. The common carotid arteries were tied off above the clavicles and canulated proximal to the carotid bodies. A separate peristaltic pump with two channels was used to perfuse the common carotid arteries at ∼15–20 ml min⁻¹ per carotid. Up to this stage in the dissection, central and peripheral perfusion was from the same tonometer (∼300 ml).

Independent perfusion of central (descending aorta) and peripheral (carotid arteries) circuits was then initiated by pulling fresh Ficoll-containing medium from two different reservoirs of a custom-built tonometer. This custom-built system was designed to accommodate a common return while preventing mixing of perfusate once equilibrated in the two reservoirs. Between the reservoir and the preparation, the central perfusate passed through a bubble trap, heat exchanger and a 25 μm filter. The carotid perfusate passed through a bubble trap and a heat exchanger. Perfusate leaked from the many cut vessels in the preparation and returned to the experimental tonometer where it was recycled and re-equilibrated (total volume of perfusate: 750 ml).

Following the initiation of independent perfusion, the central perfusate was equilibrated with 40 Torr P_CO2 in O₂ and the peripheral perfusate was equilibrated with 40 Torr P_CO2 and 100 Torr P_O2 in N₂. Over the next 30 min, the temperature of the preparation and the central perfusion pressure were ramped to 33–34°C and ∼90 mmHg, respectively. The central P_CO2 was then adjusted slowly to the level utilized for the subsequent protocol (25, 35 or 50 Torr P_CO2; estimated pH for 34°C: 7.58, 7.43 and 7.27, respectively). Protocols were started at 70 ± 5 min after central canulation. Gases used to equilibrate perfusate were mixed by computer-controlled mass flow controllers (MKS Instruments) and sampled to ensure accuracy (SensorMedics (Cardinal Health), Medical Gas Analyser LB-2 (for CO₂) and Sable Systems (Las Vegas, NV, USA) PA-1B O₂ analyser, calibrated daily). Gas concentrations reported here were accurate to within 0.1% (∼1.3 Torr in Calgary). No neuromuscular blocker was used.

Electrophysiology

The left phrenic nerve was dissected out with a small piece of diaphragm attached to its distal end and placed in a custom made Plexiglas recording chamber. The nerve was bathed in perfusate and protected with petroleum jelly. Silver bipolar extracellular electrodes were used to record the phrenic neurogram, which was amplified (Differential AC Amplifier Model 1700, A-M Systems Inc., Carlsborg, WA, USA), filtered (low cut-off, 300 Hz, high cut-off, 5 kHz), rectified and integrated (Amplitude Demodulator, Saga Tech, Calgary, AB, Canada), computer archived (Digidata 1322A and Axoscope 9.0, Axon Instruments/Molecular Devices, Union City, CA, USA) at a sampling rate of 50 Hz, and analysed off-line.

Phrenic neurogram analysis

Only preparations that exhibited a ramping phrenic discharge (the peak of phrenic activity occurs in the last 50% of the duty cycle) and an inspiratory time (T_I) of = 1 s under control conditions were included in the study. Preparations with a T_I > 1 s were considered apneustic. Analysis of phrenic neurogram characteristics was performed using in house software written by R.J.A.W. The following respiratory variables were quantified from the integrated phrenic neurogram: period (T_TOT), respiratory frequency (f_R, 60 times the inverse of the period), time to peak (T_p), inspiratory duration (T_I), expiratory duration (T_E), neural tidal volume (nV_T, the peak phrenic amplitude), neural minute ventilation (, the product of f_R and nV_T) and eupnoeic index (E_I, T_p divided by T_I). A schematic of this quantification has been previously published (Day & Wilson, 2005).

Protocols and baseline conditions

Forty-eight preparations were used to quantify the phrenic responses to specific carotid body (1) isooxic carbon dioxide perturbations (25 preparations) or (2) isocapnic oxygen perturbations under various levels of steady-state brainstem P_CO2 (23 preparations). During the 70 min recovery period following dissection, baseline conditions were established: brainstem 35 Torr P_CO2 in O₂ and carotid bodies, 35 Torr P_CO2 and 100 Torr P_O2 in N₂. In both protocols, a 5 min baseline was obtained at 35 Torr P_CO2 brainstem. This baseline was intended as a common point for all preparations to which data could be normalized (within animal) for graphical presentation. Randomly, the brainstem of each preparation were then changed over 5 min to either 25 Torr P_CO2, 35 Torr P_CO2 (no change from baseline) or 50 Torr P_CO2. A second baseline was established under these new conditions. One of two experimental protocols was then superimposed upon these new baseline conditions (see Fig. 1). Protocol 1, carotid body P_CO2: randomly, we applied five carotid body carbon dioxide perturbations lasting 5 min each: 15, 25, 35, 45 and 55 P_CO2 in 100 Torr P_O2. Protocol 2, carotid body P_O2: randomly, we applied five carotid body oxygen perturbations lasting 5 min each: 400, 200, 100, 60 and 40 Torr P_O2 in 35 Torr P_CO2. A 5 min washout period was then applied to all preparations (100 Torr P_O2 and 35 Torr P_CO2) to ensure the preparations still fitted the inclusion criteria for the study (T_I < 1 s, E_I greater than 0.5).

Figure 1. Experimental protocol.

All experiments began with a common baseline (BL 1): brainstem (BS) 35 Torr P_CO2 in O₂ and carotid bodies 100 Torr P_O2 and 35 Torr P_CO2 in N₂. The brainstem was then ramped randomly to one of three experimental P_CO2 levels: 25, 35 or 50 Torr. Following a second baseline at this new level (BL 2), randomly chosen carotid body (CB) treatments were applied (CB TX 1–5). The carotid body treatments were either P_CO2 or P_O2 perturbations. Experiment 1, CB P_CO2: 15, 25, 35, 45 and 55 Torr (in 100 Torr P_O2). Experiment 2, CB P_CO2: 400, 200, 100, 60 and 40 Torr P_O2 (in 35 Torr P_CO2). Each perturbation was 5 min and the last minute of each was analysed and normalized to the values at BL 1 for graphical presentation in Fig. 2 (CB P_CO2) and Fig. 5 (CB P_O2).

Data analysis

From the phrenic neurogram, f_R, nV_T and were quantified and averaged in 1 min bins. Data from the last minute of each perturbation were used for comparison. For each animal (1) all data were normalized to the first baseline (35 Torr P_CO2 brainstem) for graphical comparison between animals and (2) a mathematical function was fitted to normalized data resulting from the carotid body perturbations. For the carotid body P_CO2 perturbations, a linear function best described the relationship:

Here, y is the respiratory variable (f_R, nV_T or ), y₀ is the y intercept, x is the level of carotid body P_CO2 and a is the slope (representing responsiveness). The slope of each line (a) was calculated, tested for significance and compared between brainstem conditions. For the carotid body P_O2 perturbations, an inverse first order polynomial function was used:

In this case, y is the respiratory variable (f_R, nV_T or ), y₀ is the y asymptote, x is the level of carotid body P_O2 and c is the curvature (representing responsiveness). This function allowed us to compare a term quantifying the responsiveness (curvature, c) of each plot between brainstem conditions, similar to the slope term in the linear function for carotid body P_CO2.

The inverse first order polynomial function utilized to quantify the phrenic responsiveness to changes in carotid body P_O2 was similar to the function used by Weil et al. (1970) to characterize human responses to alveolar isocapnic P_O2 perturbations. However, in Weil et al. (1970), there was an extra parameter representing the x asymptote. In our experiments, the curvature of some responses was such that there was no x asymptote above 0 Torr P_O2; hence our decision to use a simplified function without this parameter.

Statistical tests

In choosing our statistical analysis, we aimed to test two key questions: (1) whether there was a significant respiratory response (f_R, nV_T or ) to carotid body P_CO2 or P_O2 perturbations and (2) whether there was a difference in the sensitivity of ventilatory responses to carotid body P_CO2 and P_O2 perturbations (slope and curvature, respectively) between brainstem P_CO2 conditions.

To address the first question, we tested for the presence of a slope (carotid body P_CO2) or curvature (carotid body P_O2) by testing if the calculated a or c values, respectively, were statistically different from zero. The a or c coefficients were calculated (as above) and each was tested for significance using the Fisher (F) statistic using SigmaPlot 10.0 software (assumed at P < 0.05). To determine whether each group of coefficients was significant we used chi squared (χ²) analysis to combine the P-values from each preparation and generate an overall P-value for each data set using χ²= 2Σu_i, where u_i is the negative natural log (–ln) of each individual probability (i.e. P-value; Tables 1 and 2; Winer, 1970). With the degrees of freedom equal to twice the number of preparations, the overall P-value was obtained from a standard χ² table. This comparison generated a group P-value, indicating whether or not the a or c values were significantly different from zero (assumed at P < 0.05), and thus whether the coefficient contributed to the fit of the function to the data.

Table 1.

Carotid body carbon dioxide data

	1	2	3	4	5	6	7	8	9	All
Frequency
BS 25
a	0.008	-0.0005	0.01	0.005	0.007	0.008	0.03	0.02	0.0002	0.009
R2	0.96	0.84	0.95	0.95	0.87	0.96	0.95	0.64	0.16	0.95
P	0.0037	0.03	0.0047	0.0049	0.02	0.0034	0.0046	0.1	0.5	0.0051
BS 35
a	0.0091	0.0017	0.0053	0.0031	0.0028	0.0008	0.0033	0.0022	—	0.0035
R2	0.91	0.87	0.98	0.83	0.97	0.16	0.75	0.9	—	0.93
P	0.012	0.021	0.0013	0.031	0.0021	0.51	0.057	0.015	—	0.0079
BS 50
a	0.0022	0.0021	–0.0006	–0.0012	0.0042	0.0005	0.0016	0.0012	—	0.0013
R2	0.59	0.67	0.006	0.22	0.88	0.051	0.55	0.12	—	0.50
P	0.13	0.09	0.9	0.42	0.018	0.72	0.15	0.57	—	0.19
Neural tidal volume
BS 25
a	0.0017	0.0018	0.008	–0.0046	-0.0011	-0.0004	–0.0004	0.008	0.0005	0.0015
R2	0.84	0.26	0.81	0.53	0.2	0.55	0.067	0.46	0.049	0.62
P	0.029	0.38	0.037	0.16	0.45	0.15	0.67	0.21	0.72	0.11
BS 35
a	–0.0002	0.0049	0.0009	0.001	0.0006	−2.08 × 10−5	2.23 × 10−5	0.0011	—	0.001
R2	0.021	0.86	0.52	0.64	0.017	0.0011	7.75 × 10−5	0.055	—	0.81
P	0.82	0.024	0.17	0.11	0.84	0.96	0.99	0.7	—	0.037
BS 50
a	–0.001	–0.0009	0.0003	0.0007	0.0005	0.0002	-7.26 × 10−5	-0.0001	—	−4.86 × 10−5
R2	0.097	0.24	0.062	0.45	0.02	0.0082	0.0055	0.0037	—	0.0032
P	0.61	0.4	0.69	0.22	0.82	0.89	0.91	0.92	—	0.93
Neural minute ventilation
BS 25
a	0.0076	0.0017	0.016	1.81 × 10−5	0.0042	0.0079	0.023	0.019	0.0007	0.0089
R2	0.99	0.18	0.95	1.30 × 10−5	0.63	0.95	0.94	0.57	0.11	0.97
P	0.0002	0.47	0.0044	0.1	0.11	0.0045	0.0062	0.14	0.58	0.0018
BS 35
a	0.0092	0.0059	0.0059	0.0037	0.0031	0.0007	0.0033	0.0028	—	0.0043
R2	0.94	0.87	0.1	0.82	0.41	0.2	0.48	0.29	—	0.95
P	0.0066	0.022	<0.0001	0.034	0.24	0.46	0.2	0.35	—	0.0047
BS 50
a	0.0005	0.0009	–0.0003	–0.0004	0.0046	0.0009	0.0016	0.0008	—	0.0011
R2	0.041	0.11	0.0012	0.069	0.69	0.38	0.85	0.3	—	0.4
P	0.75	0.6	0.96	0.67	0.08	0.27	0.025	0.34	—	0.25

Frequency, neural tidal volume and neural minute ventilation calculations of individual responses to specific carotid body P_CO2. The a value represents the slope (i.e. responsiveness) in y=y₀+ax. The R² value represents the goodness of fit of this function to the data. The P-value represents whether or not the slope of each response is statistically different from zero (assumed at P < 0.05). BS, brainstem. Mean data (All; shaded) are calculated from the mean curves in Fig. 2. Note that the value for a is the same if calculated as a mean of all the individuals, or if calculated from fitting the function to the mean response, as shown here.

Table 2.

Carotid body oxygen data

	1	2	3	4	5	6	7	8	9	All
Frequency
BS 25
c	0.27	33.12	51.65	63.33	22.66	49.32	21.2	18.57	23.3	31.49
R2	0.0044	0.94	0.94	0.93	0.85	0.95	0.73	0.94	0.97	0.96
P	0.92	0.0068	0.007	0.0078	0.026	0.0051	0.067	0.0059	0.0026	0.0034
BS 35
c	13.59	2.87	54.87	27.26	–3.26	14.44	11.89	—	—	17.38
R2	0.88	0.4	0.93	0.91	0.47	0.97	0.86	—	—	0.1
P	0.019	0.25	0.0077	0.012	0.2	0.0018	0.024	—	—	0.0001
BS 50
c	5.89	4.11	–1.87	0.4	17.68	9.26	0.28	—	—	5.11
R2	1.0	1.0	0.5	0.025	0.98	0.95	0.0073	—	—	0.98
P	<0.0001	<0.0001	0.18	0.80	0.0013	0.0056	0.89	—	—	0.0014
Neural tidal volume
BS 25
c	12.6	4.68	25.24	15.73	7.5	0.87	5.3	0.97	3.15	8.45
R2	0.98	0.52	0.79	0.7	0.76	0.11	0.23	0.11	0.45	0.99
P	0.0008	0.17	0.044	0.081	0.053	0.59	0.41	0.58	0.22	0.0005
BS 35
c	5.58	2.03	0.85	2.79	5.35	0.43	4.13	—	—	3.022
R2	0.52	0.26	0.034	0.29	0.52	0.0028	0.34	—	—	0.69
P	0.17	0.38	0.77	0.35	0.17	0.93	0.31	—	—	0.08
BS 50
c	1.22	1.35	9.98	3.81	0.64	–4.19	−0.89	—	—	1.7
R2	0.11	0.16	0.78	0.95	0.051	0.39	0.23	—	—	0.65
P	0.58	0.51	0.047	0.0047	0.72	0.26	0.41	—	—	0.1
Neural minute ventilation
BS 25
c	0.27	33.12	51.65	63.33	22.66	49.32	21.21	17.32	16.77	34.22
R2	0.0044	0.94	0.94	0.93	0.85	0.95	0.73	0.86	0.85	0.99
P	0.92	0.0068	0.007	0.0078	0.026	0.0051	0.067	0.025	0.025	0.0009
BS 35
c	16.54	4.47	55.53	27.35	1.74	11.6	13.41	—	—	18.66
R2	0.77	0.69	0.91	0.88	0.063	0.63	0.77	—	—	0.98
P	0.049	0.084	0.012	0.019	0.69	0.11	0.052	—	—	0.0011
BS 50
c	6.24	4.88	8.19	4.25	22.1	6.22	−0.61	—	—	7.32
R2	0.75	0.8	0.74	0.76	0.98	0.53	0.023	—	—	0.99
P	0.057	0.039	0.062	0.055	0.0013	0.16	0.81	—	—	0.0004

Frequency, neural tidal volume and neural minute ventilation calculations of individual responses to specific carotid body P_O2. The c value represents the curvature (i.e. responsiveness) in y=y₀+c/x. The R² value represents the goodness of fit of this function to the data. The P-value represents whether or not the curvature of each response is statistically different from zero (assumed at P < 0.05). BS, brainstem. E, exponent. Mean data (All; shaded) is calculated from the mean curves in Fig. 5. Note that the value for c is the same if calculated as a mean of all the individuals, or if calculated from fitting the function to the mean response, as shown here.

To address the second question, for carotid body P_CO2 perturbations, we calculated the a value (slope) of each of the three respiratory variables (f_R, nV_T and ) for each preparation. These values were then compared between brainstem P_CO2 conditions using 1-factor (1F) ANOVAs. For carotid body P_O2 perturbations, we performed a comparable analysis using c values (curvature).

Curve fitting and statistical testing of the significance of the a and c values were performed using SigmaPlot 10.0 (Systat Software Inc., San Jose, CA, USA). Statistical tests comparing the magnitude of the a or c values between brainstem P_CO2 levels (i.e. 1F ANOVAs) were performed using SigmaStat 2.03 (Systat) on values calculated from normalized data. In all cases, once significant differences had been determined (assumed at P < 0.05), Student–Newman–Keuls (SNK) post hoc tests were used for pair-wise comparisons.

Results

Phrenic responses to specific carotid body P_CO2

Did changes in carotid body P_CO2 produce respiratory responses?

Specific carotid body P_CO2 perturbations (15–55 Torr; P_O2 100 Torr) resulted in linear responses in f_R, nV_T and (Fig. 2). The R² values in Table 1 represent the goodness of fit of the linear function to data from each preparation. Table 1 also contains P-values demonstrating whether the slope of the responses (a values) of individual preparations were significant.

Figure 2. Mean normalized phrenic response slopes to specific carotid body (CB) iso-oxic (100 Torr)P_CO2perturbations.

All points were normalized within animal to a common point where the brainstem (BS) was 35 Torr P_CO2 (Baseline 1; see Fig. 1). A, mean normalized phrenic f_R responses to CB P_CO2; B, mean normalized phrenic nV_T responses to CB P_CO2; C, mean normalized phrenic responses to CB P_CO2. Brainstem 25, 35 and 50 Torr P_CO2 as indicated by circle, triangle and square symbols, respectively. The trend lines are a linear function (y=y₀+ax) that best described the responses. The R² values indicate the goodness of fit of the function to the data can be found in Table 1.

Individual P-values were combined to give an overall P-value using the χ² test (see Methods). This test yielded the following results. Overall, for both f_R and , the response slopes were significantly different from zero when the brainstem was maintained at either 25 or 35 Torr P_CO2 (P < 0.001). Frequency effects were mediated predominantly through changes in T_E (Fig. 4). For nV_T, the response slope was significantly different only when the brainstem was maintained at 25 Torr P_CO2 (P < 0.05).

Figure 4. Phrenic T_I and T_E in response to specific carotid body iso-oxic (100 Torr)P_CO2perturbations at various brainstemP_CO2levels.

A, brainstem 25 Torr P_CO2 (n= 9); B, brainstem 35 Torr P_CO2 (n= 8); C, brainstem 50 Torr P_CO2 (n= 8). Filled squares are T_E. Open diamonds are T_I. Error bars are s.e.m.

Did sensitivity of responses to carotid body P_CO2 depend on brainstem P_CO2?

For f_R, response slopes to carotid body P_CO2 were statistically different between different brainstem P_CO2 conditions (1F ANOVA; P= 0.018, F= 4.831; Fig. 3A). Specifically, the slope of the response to carotid body P_CO2 when the brainstem was maintained at 25 Torr P_CO2 was greater (increased responsiveness) than when the brainstem was maintained at either 35 or 50 Torr P_CO2 (SNK post hoc test; P < 0.042, q > 3.048; Fig. 3A). These differences mainly reflect changes in T_E (Fig. 4). For nV_T, there were no statistically significant differences between slopes under any brainstem P_CO2 condition (1F ANOVA; P= 0.493, F= 0.731; Fig. 3B). For , there was a statistically significant difference between response slopes under different brainstem P_CO2 conditions (1F ANOVA; P= 0.023, F= 4.528). Specifically, when the brainstem was maintained at 25 Torr P_CO2 the slope of the response to carotid body P_CO2 was greater (increased responsiveness) than when the brainstem was maintained at 50 Torr P_CO2 (SNK post hoc test; P= 0.018, q= 4.22; Fig. 3C).

Figure 3. Comparison of phrenic response slopes to specific carotid body (CB) iso-oxic (100 Torr)P_CO2perturbations under differing brainstemP_CO2.

Slopes (i.e. responsiveness) were calculated from the normalized data using linear function (a value in y=y₀+ax) fitted through all 5 CB P_CO2 points (see Fig. 2 for mean curves). A, comparison of phrenic f_R slopes at different brainstem P_CO2; B, comparison of phrenic nV_T slopes at different brainstem P_CO2; C, comparison of phrenic slopes at different brainstem P_CO2. Calculations and statistics were performed on normalized data. Filled bars represent that the slope was statistically significant from zero (P < 0.05, see results section). *Statistically significant difference between brainstem 25 and both 35 and 50 Torr P_CO2 in A (P= 0.018), and between brainstem 25 and 50 Torr in C (P= 0.023). Significance assumed at P < 0.05. NSD, no significant differences. Error bars represent s.e.m. a.u., arbitrary units. See Table 1 for analysis of individual data.

Phrenic responses to specific carotid body P_O2

Did changes in carotid body P_O2 produce respiratory responses?

Specific carotid body P_O2 perturbations (400–40 Torr, P_CO2 35 Torr) resulted in curvilinear (hyperbolic) responses in f_R, nV_T and (Fig. 5). The R² values in Table 2 represent the goodness of fit of the inverse first order function to data from each preparation. Table 2 also contains P-values demonstrating whether the curvatures of the responses (c values) of individual preparations were significant.

Figure 5. Mean normalized phrenic response slopes to specific carotid body (CB) isocapnic (35 Torr)P_O2perturbations.

All points were normalized within animal to a common point where the brainstem (BS) was 35 Torr P_CO2 (Baseline 1; see Fig. 1). A, mean normalized phrenic f_R responses to CB P_O2; B, mean normalized phrenic nV_T responses to CB P_O2; C, mean normalized phrenic responses to CB P_O2. Brainstem 25, 35 and 50 Torr P_CO2 as indicated by circle, triangle and square symbols, respectively. The trend lines are an inverse first order function (y=y₀+c/x) that best described the mean responses. The R² values indicate the goodness of fit of the function to the data and can be found in Table 2.

Individual P-values were combined to give an overall P-value using the χ² test (see Methods). This test yielded the following results. For f_R, the response curvatures were significantly different from zero under conditions of brainstem 25, 35 and 50 Torr P_CO2 (P < 0.001) mainly through an affect on T_E (Fig. 7). For nV_T, the response curvatures were significantly different from zero under conditions of brainstem 25 (P < 0.01) and 50 Torr P_CO2 (P < 0.05). For , the response curvatures were significantly different from zero under conditions of brainstem 25, 35 and 50 Torr P_CO2 (P < 0.001).

Figure 7. Phrenic T_I and T_E in response to specific carotid body isocapnic (35 Torr)P_O2perturbations at various brainstemP_CO2levels.

A, brainstem 25 Torr P_CO2 (n= 9); B, brainstem 35 Torr P_CO2 (n= 7); C, brainstem 50 Torr P_CO2 (n= 7). Filled squares are T_E. Open diamonds are T_I. Error bars are s.e.m.

Did sensitivity of responses to carotid body P_O2 depend on brainstem P_CO2?

For f_R, response curvatures to carotid body P_O2 were statistically different between different brainstem P_CO2 conditions (1F ANOVA; P= 0.018, F= 4.956). Specifically, curvatures when the brainstem was 25 Torr P_CO2 were greater (increased responsiveness) than when held at 50 Torr P_CO2 (SNK post hoc test; P= 0.014, q= 4.427; Fig. 6A). These differences mainly reflect changes in T_E (Fig. 7). For nV_T, there were no statistically significant differences between curvatures under any brainstem P_CO2 condition (1F ANOVA; P= 0.064, F= 3.166; Fig. 6B). For , there was a statistically significant difference between response curvatures under different brainstem P_CO2 conditions (1F ANOVA; P= 0.039, F= 3.836). Specifically, response curvatures when the brainstem was 25 Torr P_CO2 were greater (increased responsiveness) than when held at 50 Torr P_CO2 (SNK post hoc test; P= 0.031, q= 3.903; Fig. 6C).

Figure 6. Comparison of a values (curvature) from phrenic response curves to specific carotid body (CB) isocapnic (35 Torr)P_O2perturbations under differing brainstemP_CO2.

A measure of curvature (i.e. responsiveness) was calculated from normalized data using an inverse first order function (c value in y=y₀+c/x) fitted through all 5 CB P_O2 points (see Fig. 5 for mean curves). A, comparison of phrenic f_R responsiveness values at different brainstem P_CO2. B, comparison of phrenic nV_T responsiveness at different brainstem P_CO2. C, comparison of phrenic responsiveness at different brainstem P_CO2. Calculations and statistics were performed on normalized data. Filled bars represent that the curvature was statistically significant from zero (P < 0.05, see results section). *Statistically significant difference between brainstem 25 and 50 Torr P_CO2 in A (P= 0.018) and in C (P= 0.039). Significance assumed at P < 0.05. NSD, no significant differences. Error bars represent s.e.m. a.u., arbitrary units. See Table 2 for analysis of individual data.

To ensure that the results presented here were not an artifact of testing normalized data, slopes (carotid body P_CO2) and curvatures (carotid body P_O2) were also calculated from raw data and compared statistically: the conclusions from these tests were identical to those presented here (data not shown).

Discussion

Using a dual perfused rodent preparation (DPP), we report that inputs from specific carotid body CO₂ and O₂ perturbations are integrated with inputs from central chemoreceptors and translated to respiratory variables in a hypo-additive manner. This negative interaction between chemoreceptors holds for carotid body perturbations under conditions of both decreased activation (hypocapnia and hyperoxia) and increased activation (hypercapnia and hyperoxia), in a brainstem P_CO2 dose-dependent manner. This finding suggests our previous demonstration of a negative interaction is generalizable and not caused by saturation of phrenic response (previously we stimulated the carotid bodies with 60 Torr P_O2 and compared between 2 levels of brainstem P_CO2: 25 and 50 Torr; Day & Wilson, 2007).

Similar to our previous study, the respiratory responses to specific carotid body perturbations and the negative interaction we report here were most apparent in frequency (f_R) (owing mainly to a pronounced effect on T_E; see Figs 4 and 7 for trends), which translated to neural minute ventilation (). There was a general trend toward a reduction in peripheral chemoreflex responsiveness in all respiratory variables as the brainstem P_CO2 was increased (i.e. a dose response; see Figs 3 and 6). Further, the fit of functions used to quantify carotid body P_O2 and P_CO2 responses (as assessed by the coefficient of predictive ability of the function, R²) decreased with increasing brainstem P_CO2 (see Tables 1 and 2). We suggest that the reduction in R² is also consistent with the conclusion that the efficaciousness of the peripheral chemoreflex in contributing to a ventilatory response is reduced as the brainstem becomes progressively more hypercapnic. It should be noted that this study does not address how the responsiveness of the peripheral chemoreflex to P_CO2 changes with varying levels of carotid body P_O2 (e.g. Nielsen & Smith, 1952; Lloyd et al. 1958; Lahiri & DeLaney, 1975; Daristotle et al. 1987). Nor does this study address (1) how the P_O2–P_CO2 relationship is affected by brainstem P_CO2 and (2) whether time-dependent responses such as short-term potentiation, short-term depression and posthypoxic frequency decline are individually affected by brainstem P_CO2 (Day & Wilson, 2007; Song & Poon, 2009).

Critique of the DPP

The DPP provides a novel method for studying respiratory chemoreceptor interaction in the rat, acting as an important bridge between recent studies in rodents at the cellular level and whole animal responses (Mulkey et al. 2004; Takakura et al. 2006). An important strength of our model is that both brainstem and peripheral chemoreceptor compartments are intact and independently perfused. Additionally, unlike in vivo preparations used to study the interaction of central and peripheral chemoreceptors that breathe room air or inspired gases, we are able to hold the brainstem at a P_CO2 below eupnoeic levels. Thus, the chemoreflex resulting from one chemoreceptor can be explored across the entire physiological range while the other chemoreceptor is maintained at a constant level of chemostimulation.

Uniquely, the DPP allows the study of brainstem and carotid body respiratory chemosensitivity in non-anaesthetized, decerebrate rodents, independent of heart rate, systemic blood pressure, hormonal factors and vagal inputs. Consequently, one of the key advantages of our model is that the question of brainstem and peripheral chemoreceptor interaction can be examined without the confounding influence of other factors. Indeed, one might argue that only by first understanding the fundamental interaction between the two types of chemoreceptors in non-anaesthetized animals will we be able to understand how the superimposition of vagal afferents, suprapontine structures and other influences modulates chemosensitivity.

While on the one hand the above factors are advantageous for studying brainstem–carotid body chemoresponse interactions, on the other they inevitably decrease the facsimile to chemosensitivity in an intact animal (discussed below). Therefore, our data should be interpreted in the light of the possibility that the very inputs we have gone to lengths to remove, may prove to play key roles in chemosensitivity in intact animals.

Comparison of chemosensitivity of the DPP to intact animals

The main differences between the chemosensitivity of the DPP with that of an intact animal is that the apnoeic threshold and CO₂ responsiveness are shifted in the hypocapnic direction (compare Boden et al. 1998 with Day & Wilson, 2005) and the response to changes in brainstem P_CO2 above normocapnic levels are reduced (see supplementary data). Other studies have demonstrated that both decerebration (Nielsen et al. 1986; Mitchell, 1990; Hayashi & Sinclair, 1991; St.-John & Paton, 2000) and vagotomy (Takakura et al. 2006) have qualitatively similar effects.

In the DPP, the maximum sensitivity of the peripheral chemoreflex to P_CO2 in normoxia is about 1% Torr⁻¹ (when the brainstem was maintained at 25 Torr P_CO2; Figs 2 and 3). In humans, reports of normoxic P_CO2 responsiveness have been in the order of 1.5–2 l min⁻¹ Torr⁻¹P_CO2 (Xie et al. 2001). This amounts to approximately 30% Torr⁻¹P_CO2 (calculation: 6 l min⁻¹ baseline, 2 l increase per Torr P_CO2=∼30% increase per Torr). If one estimates that about 30% of the total CO₂ chemoresponse is from the carotid bodies in normoxia as others do (e.g. Smith et al. 2006; Tananka et al. 2000), then this amounts to a sensitivity of ∼10% increase per Torr P_CO2 resulting from carotid body stimulation. This is an order of magnitude larger than our preparation.

Some of the blunted chemosensitivity of the DPP compared to humans might partly be explained by species differences. The CO₂ chemosensitivity of the DPP is similar in magnitude to that measured in awake freely behaving rats of the same species by some investigators (compare Czapla & Zadina, 2005 with Day & Wilson, 2005), but not by others (Nattie & Li, 2002).

Another possible explanation for the blunted chemoresponses of the DPP is that decerebation removes suprapontine structures that contribute to and/or modulate system chemosensitivity. In this respect, we note recent work in which activation of the dorsomedial hypothalamic region in anaesthetized rats excites putative respiratory chemosensitive cells in the medulla and up-regulates phrenic output (Guyenet, 2008). The hypothalamus is likely to be removed during decerebration in the DPP. Anaesthesia also eliminates cortical inputs that influence respiration and blunts chemoresponses. However, unlike decerebration, anaesthetics are likely to have widespread desensitizing effects throughout the respiratory control circuit including the brainstem and peripheral chemoreceptors themselves (e.g. Faber et al. 1982; Wyon et al. 1998; Czapla & Zadina, 2005).

Negative interaction between central and peripheral chemoreceptors

Some studies that address the question of interaction between central and peripheral chemoreceptors have demonstrated simple addition in generating minute ventilation (see references in Introduction). In these studies, the multiplicative relationship between frequency and tidal volume in generating minute ventilation necessitates that a negative interaction exists between central and peripheral chemoreceptors in at least one respiratory variable (f_R or V_T; Mitchell, 1990; Day & Wilson, 2007). However, due to the shift in the chemosensitive range of the DPP, it is possible that the demonstration of negative interactions in frequency and reported here is the consequence of saturation occurring within a common pathway in the CNS (Eldridge et al. 1981; Smith et al. 1984). Data from a previous study showed that the maximum frequency reached when the carotid body was stimulated with hypoxia was higher when the brainstem was maintained at 25 Torr P_CO2 than when it was maintained at 50 Torr P_CO2, in the same animals (Day & Wilson, 2007). Thus, it seemed unlikely that the negative interaction can be explained by saturation in ventilatory frequency of the preparation.

The current study aimed to extend the previous finding of a negative interaction and to address the possibility of saturation more directly. Here, we aimed to test the possibility of saturation in two ways, namely (1) using a brainstem P_CO2 dose–response (3 levels) and (2) testing the peripheral chemoreflex both above and below eupnoeic levels (i.e. both increased and decreased carotid body activation). The fact that (1) a portion of the peripheral chemoreflexes measured here contained responses under conditions of decreased activation and (2) the magnitude of the peripheral chemoreflexes varied between three brainstem P_CO2 conditions suggests that there is a negative interaction between chemoreceptors that is not explained by saturation of a common neural pathway in the CNS.

Using an artificially perfused awake goat model, whereby the cisterna magna was perfused with mock CSF containing various levels of bicarbonate, Smith et al. (1984) also demonstrated a negative interaction between chemoreceptors. By injecting various concentrations of sodium cyanide intravenously to stimulate the carotid bodies, they found that the efficacy of carotid body stimulation at eliciting increases in frequency and tidal volume increased as the cisterna magna perfusion was made progressively more alkaline (Smith et al. 1984). Although our study differs in the species and carotid body stimuli used (e.g. sodium cyanide versusP_CO2 and P_O2), our conclusions are in agreement with Smith and colleagues’ study (1984). In the current study, combining conclusions from both the statistical tests of (1) whether the slopes (carotid body P_CO2) or curvatures (carotid body P_O2) were significantly different from zero and (2) whether the magnitude of slopes and curvatures differs between brainstem P_CO2 conditions supports the conclusion that f_R and responses to changes in both carotid body P_CO2 (Fig. 3) and carotid body P_O2 (Fig. 6) are modulated by brainstem P_CO2 in a dose-dependent manner.

Site of interaction

Potential sites of convergence between brainstem and carotid body chemoreceptors include the nucleus tractus solitarii (NTS), raphe, retrotrapazoid nucleus (RTN) and the lateral parabrachial nucleus (LPBN) of the pons. While any of these sites may influence or dictate the mathematical nature of the interaction, recent studies suggest that the RTN and LPBN may be particularly important.

Takakura et al. (2006) showed that there are direct anatomical connections between the NTS (where afferent input from the carotid bodies impinge) and the RTN, a putative central chemoreceptor site close to the ventrolateral surface of the brainstem. Carotid body stimulation by inspired hypoxia increased the responsiveness of RTN neurons in anaesthetized rats to increases in inspired CO₂. These data suggest that the RTN is a good candidate for a site of carotid body–brainstem chemoreceptor pathway interaction. Interestingly, with direct relevance to our main finding, analysis of firing frequencies of RTN cells with and without hypoxic stimulation of the carotid bodies demonstrated an additive interaction. However, in these experiments the carotid bodies were also likely to have been stimulated by the inspired CO₂: a negative interaction between central and peripheral chemoreceptors may have been masked by the known hyper-additive interaction between P_CO2 and P_O2 within the carotid bodies (e.g. Lahiri & DeLaney, 1975). Thus, while the RTN may be involved in mediating the interaction between chemoreceptor pathways, the nature of this interaction at the cellular level requires further investigation.

The LPBN receives inputs from the RTN and the NTS and plays a powerful role in regulating acute hypoxic and hypercapnic responses in unanaesthetized rats (Mizusawa et al. 1995). Recently, Song & Poon (2009) demonstrated that the LPBN plays a critical role in modulating T_E during acute hypoxic or hyperoxic hypercapnic chemostimulation. In this respect, we note that the negative interaction we observed in frequency was also largely dependent on changes in T_E (see Figs 4 and 7). Specifically, reductions in T_E during carotid body stimulation were progressively eliminated as brainstem P_CO2 increased. Thus, the convergence of brainstem and carotid body chemoreceptor information onto the LPBN and the importance of the LPBN in regulating T_E both raise the possibility that the negative interaction between chemoreceptors we observe may in part be mediated through this pontine nucleus.

Implications for respiratory stability in humans

In sleeping humans, breathing stability depends critically upon central and peripheral chemoreflex gain (i.e. controller gain; Chapman et al. 1988; Xie et al. 1995; Solin et al. 2000; Topor et al. 2001, 2004; Dempsey et al. 2004). The higher the chemoreceptor gain, the more unstable the system and the more likely a hyperpnoea will reduce the P_CO2 below the apnoeic threshold, causing apnoea (Xie et al. 2001, 2002; Dempsey et al. 2004). Our data demonstrate that the mathematics of chemoreceptor interaction can have a sizable effect on the magnitude of peripheral chemoreflex gain. Currently, the mathematics of chemoreceptor interaction in humans is uncertain (see Introduction) and whether the nature of the interaction is specific to (a) different phases of the hypoxic response and/or (b) modulated by sleep state is unknown. Nonetheless, to the extent that a negative interaction between central and peripheral chemoreceptors exists in humans, this interaction has the potential to play an important role in breathing stability in sleeping humans by influencing the controller gain.

During some circumstances, a negative interaction between chemoreflexes could destabilize breathing. The central chemoreflex has a delayed onset and a longer time constant in response to blood gas perturbations compared to the peripheral chemoreflex (Bellville et al. 1979; Pedersen et al. 1999; Fatemian et al. 2003; Smith et al. 2006). Consequently, transient swings in blood gases resulting from an apnoeic event may be sufficient to stimulate the carotid bodies without significantly affecting the activity of the central chemoreceptors. In this case, if the central compartment was relatively hypocapnic compared to the peripheral compartment, our results would predict an increase in peripheral chemoreflex gain which, in turn, may have a destabilizing effect on breathing. Thus, transient swings in blood gases may be particularly detrimental in populations that are chronically hypocapnic (e.g. congestive heart failure, sleeping at high altitude; see Day & Wilson, 2008).

During other circumstances, a negative interaction may be stabilizing. For example, the sensitivity of the carotid body to hypoxia is greatly magnified by hypercapnia owing to the strongly multiplicative interaction between CO₂ and O₂ sensing at the level of the carotid body (e.g. Lahiri & DeLaney, 1975). While this gain–enhancing multiplicative interaction is essential for the function of the carotid bodies in the breath-by-breath monitoring of blood gases (Lahiri et al. 2006), it may lead to instability in the event of a brief bout of hypoxia in the presence of concomitant hypercapnia. Our data predict that if the brainstem chemoreceptors are also stimulated (i.e. hypercapnic), the relative influence of the peripheral chemoreceptor on the subsequent chemoreflex is mitigated.

Given the possible importance of chemoreflex interaction to breathing in sleeping humans, we urge new experimental and modelling studies to clarify the nature of the chemoreceptor interaction in humans and reexamine its consequences on the chemoreflex control of breathing.

Conclusions

Our data demonstrate a negative interaction between central (brainstem) and peripheral (carotid body) chemoreceptors, such that when both central and peripheral chemoreceptors are strongly stimulated, the central chemoreceptors largely dictate the level of ventilation. We also demonstrate that regardless of whether activation of the peripheral chemoreceptor is increased or decreased, the gain of the resulting peripheral chemoreflex is inversely proportional to brainstem P_CO2 (e.g. decreasing or increasing carotid body P_CO2 and P_O2 had a larger effect on neural ventilation when the brainstem was relatively hypocapnic). These results suggest that (1) the common assumption of simple addition of central and peripheral chemoreceptor inputs used in mathematical models of the respiratory controller should be reconsidered and (2) the effects of a negative interaction between central and peripheral chemoreceptors on the stability of breathing be evaluated.

Supplemental material

Online supplemental material for this paper can be accessed at:

http://jp.physoc.org/cgi/content/full/jphysiol.2008.160689/DC1