Interaction between the ventilatory and cerebrovascular responses to hypo- and hypercapnia at rest and during exercise

Abstract

Cerebrovascular reactivity to changes in the partial pressure of arterial carbon dioxide (P_a,CO2) via limiting changes in brain [H⁺] modulates ventilatory control. It remains unclear, however, how exercise-induced alterations in respiratory chemoreflex might influence cerebral blood flow (CBF), in particular the cerebrovascular reactivity to CO₂. The respiratory chemoreflex system controlling ventilation consists of two subsystems: the central controller (controlling element), and peripheral plant (controlled element). In order to examine the effect of exercise-induced alterations in ventilatory chemoreflex on cerebrovascular CO₂ reactivity, these two subsystems of the respiratory chemoreflex system and cerebral CO₂ reactivity were evaluated (n= 7) by the administration of CO₂ as well as by voluntary hypo- and hyperventilation at rest and during steady-state exercise. During exercise, in the central controller, the regression line for the P_a,CO2–minute ventilation relation shifted to higher and P_a,CO2 with no change in gain (P = 0.84). The functional curve of the peripheral plant also reset rightward and upward during exercise. However, from rest to exercise, gain of the peripheral plant decreased, especially during the hypercapnic condition (−4.1 ± 0.8 to −2.0 ± 0.2 mmHg l⁻¹ min⁻¹, P = 0.01). Therefore, under hypercapnia, total respiratory loop gain was markedly reduced during exercise (−8.0 ± 2.3 to −3.5 ± 1.0 U, P = 0.02). In contrast, cerebrovascular CO₂ reactivity at each condition, especially to hypercapnia, was increased during exercise (2.4 ± 0.2 to 2.8 ± 0.2% mmHg⁻¹, P = 0.03). These findings indicate that, despite an attenuated chemoreflex system controlling ventilation, elevations in cerebrovascular reactivity might help maintain CO₂ homeostasis in the brain during exercise.

Numerous enzymes and ion channels which influence neural activity are modified by changes in pH (Chesler, 2003); therefore, the regulation of pH is a vital homeostatic function. The respiratory chemoreflex is an important feedback control system which keeps the partial pressure of arterial carbon dioxide (P_a,CO2) remarkably constant via ventilatory regulation. For example, the periodic nature of inspiration and expiration is carefully controlled by changes in P_a,CO2 via central and peripheral chemoreflexes so as to maintain pH nearly constant. The resulting hyper- or hypo-ventilation reduces or increases the CO₂ in the blood, respectively, and therefore in the cerebrospinal fluid.

P_a,CO2 serves as an important controlled variable or mediator, especially in the brain. The blood–brain barrier is relatively impermeable to H⁺ and HCO₃⁻ ions; however, molecular CO₂ diffuses across it readily, with the result that the CO₂ in the cerebrospinal fluid parallels the arterial CO₂. Therefore, CO₂ diffuses freely to the cerebrospinal fluid and influences pH which drives ventilation via the central chemoreceptors (Severinghaus et al. 1963; Severinghaus & Carcelen, 1964). Moreover, the middle cerebral artery mean blood velocity (MCA V_mean), as an index of cerebral blood flow (CBF), is highly sensitive to direct changes in P_a,CO2 (Markwalder et al. 1984; Rasmussen et al. 2006). For example, hypocapnia causes cerebral vasoconstriction which reduces MCA V_mean and therefore, because of a reduced ‘washout’, attenuates the fall of brain tissue P_CO2. In contrast, hypercapnia increases MCA V_mean by cerebral vasodilatation, which limits elevations in brain tissue P_CO2.

Cerebrovascular reactivity and ventilatory response to CO₂ seems to be tightly linked (Chapman et al. 1979; Dempsey, 2005; Xie et al. 2005, 2006; Ainslie et al. 2007; Peebles et al. 2007). Changes in CBF might have an important role in stabilizing the breathing pattern during fluctuating levels of chemical stimuli, especially to P_a,CO2 (Xie et al. 2006). In fact, an increase in CBF increases diffusion of CO₂ from the cerebrospinal fluid and the brain extracellular fluid to the cerebral vessels. Therefore, [H⁺] decreases at the level of the central chemoreceptors when CBF increases. Early work by Severinghaus et al. (1963) investigated the regulation of cerebrospinal fluid pH during acclimatization from sea level to high altitude. They proposed three mechanisms for regulating cerebrospinal fluid pH. In addition to active transport across the blood–brain barrier and chemoreflexes, they suggested that cerebral arterioles, which dilate with high P_CO2 and constrict with low P_CO2, also reduce the pH variations of cerebrospinal fluid and may be regarded as a third homeostatic means to regulate cerebrospinal fluid pH and therefore central ventilatory control. In goats, Chapman et al. (1979) reported that severe brain ischaemia blunted ventilatory responses to CO₂. In addition, reports indicate that cerebrovascular responsiveness to CO₂ is an important determinant of eupnoeic and hypercapnic ventilatory responsiveness in otherwise healthy humans (Xie et al. 2006) and those with congestive heart failure and central sleep apnoea (Xie et al. 2005), primarily via its effects at the level of the central chemoreceptors. Such reductions in cerebrovascular CO₂ reactivity affect the stability of the breathing pattern by causing ventilatory overshooting during hypercapnia and undershooting during hypocapnia (Xie et al. 2005). Therefore, changes in cerebrovascular CO₂ reactivity play a critical role in the ventilatory control of P_a,CO2.

High altitude-induced hyperventilation via peripheral chemoreflex activation reduces P_a,CO2 and modifies cerebrospinal fluid pH and central chemoreceptor drive (Severinghaus et al. 1963). Therefore, it is possible that exercise-induced hyperpnoea also modifies the respiratory chemoreflex. In fact, exercise increases ventilation via the respiratory chemoreflex, and also modifies the ventilatory response to CO₂ (Asmussen & Nielsen, 1957; Bhattacharyya et al. 1968; Poon & Greene, 1985); however, it remains unclear how exercise-induced alterations in the respiratory chemoreflex might influence CBF regulation, in particular cerebrovascular CO₂ reactivity. The potential interactions between cerebrovascular reactivity and ventilatory responsiveness to CO₂ during exercise have not been examined. In order to examine the effect of exercise-induced alterations in ventilatory chemoreflex on cerebrovascular CO₂ reactivity, we evaluated two subsystems of the respiratory chemoreflex system using a new equilibrium diagram model (Miyamoto et al. 2004) and cerebral CO₂ reactivity by the administration of CO₂ as well as by voluntary hypo- and hyperventilation at rest and during steady-state exercise. Under a closed-loop condition of the respiratory chemoreflex system, ventilatory output is determined by chemical and metabolic drives (I: central controller), but this ventilatory loading alters these drives in the lung system which feeds back to ventilatory output (II: peripheral plant). We hypothesized that, during exercise, an increase in cerebrovascular CO₂ reactivity will compensate for reductions in the peripheral plant of the respiratory chemoreflex system.

Methods

Seven healthy non-athletic men aged 20 ± 2 years, height 173 ± 8 cm, weight 64 ± 10 kg (mean ±s.d.) were recruited to participate in the study as approved by the Human Subjects Committee of Morinomiya University of Medical Sciences (No. 001). In addition, they were free of any known cardiovascular and pulmonary disorders and were not using prescribed or over the counter medications. Before the experiment, each subject gave informed written consent and visited the laboratory for familiarization with the techniques and procedures. All procedures conformed to the standards set by the Declaration of Helsinki. Subjects were requested to abstain from caffeinated beverages for 12 h and strenuous physical activity and alcohol for at least 24 h before the day of the experiment.

Measurements

All studies were performed at a constant room temperature between 23 and 24°C with external stimuli minimized. Heart rate (HR) was monitored using a lead II electrocardiogram (ECG). A catheter (0.47 mm i.d., 24 gauge) was placed in the brachial artery of the non-dominant arm for arterial blood samples and measurement of the arterial blood pressure (ABP) with a pressure transducer (DX-200, Nihon-Koden, Tokyo, Japan) positioned at the level of the right atrium in the mid-axillary line, fastened to the subject and connected to a pressure-monitoring system (RM-6000, Nihon-Koden). Arterial blood samples were obtained at rest and after reaching steady state in each experimental condition. Samples were immediately analysed for pH, P_a,CO2 and the partial pressure of arterial oxygen (P_a,O2) using a blood gas analyser (IL 1620, Instrumentation Laboratory, USA). The middle cerebral artery blood velocity (MCA V) was obtained by transcranial Doppler ultrasonography (WAKI, Atys Medical, St Genislaval, France). A 2 MHz Doppler probe was placed over the temporal window and fixed with an adjustable headband and adhesive ultrasonic gel (Tensive, Parker Laboratories, Orange, NJ, USA). The MCA V waveform was isonated at the same depth (5 cm from the skin surface of the temple window) in all subjects. Ventilatory responses were measured using an open-circuit apparatus. The subjects breathed through a face mask attached to a low-resistance one-way valve with a built-in hot-wire flow meter. The valve mechanism allowed subjects to inspire room air or a selected gas mixture from a 200 l Douglas bag containing 0.0, 3.5 or 5.0% CO₂ in 40% O₂ with nitrogen (N₂) balance. These concentrations of CO₂ administration were determined by previous studies (Ellingsen et al. 1987a,b). The respiratory CO₂ sensitivity is close to constant within the range 0–5% CO₂ in the inspired gas. We used three progressive CO₂ stimulus points within the range 0–5% CO₂ to identify respiratory chemoreflex in the model. The total instrumental dead space was 200 ml. Respiratory and metabolic data during the experiments were recorded by an automatic breath-by-breath respiratory gas-analysing system consisting of a differential pressure transducer, sampling tube, filter, suction pump and mass spectrometer (ARCO2000-MET, Arcosystem, Chiba, Japan). We digitized expired flow, CO₂ and O₂ concentrations, and derived tidal volume (V_T), minute ventilation , end-tidal O₂ (P_ET,O2) and end-tidal CO₂ (P_ET,CO2). Flow signals were computed to single breath data, and matched to gas concentrations identified as single breaths using the peak P_ET,CO2, after accounting for the time delay in gas concentration measurements. The corresponding O₂ uptake and CO₂ output values for each breath were calculated from inspired–expired gas concentration differences, and by expired ventilation, with inspired ventilation being calculated by N₂ correction. During each protocol, HR, ABP, , P_ET,O2, P_ET,CO2 and MCA V were recorded continuously at 200 Hz.

Experimental protocol

On the first day, each subject performed maximal cycle exercise for the measurement of maximal oxygen uptake . In addition, with the exception of arterial blood gas sampling, each subject underwent the same experiment procedures as those used during the main experimental day to ensure familiarization with the experimental protocols.

On the experimental day, subjects arrived at the laboratory at least 2 h after a light meal. Following instrumentation, the subjects rested in a comfortable chair. Five minutes of baseline data were recorded whilst the subjects breathed room air, wearing the face mask. To characterize the central controller and peripheral plant, subjects underwent two experimental procedures, which consisted of the response to hypercapnia and the P_a,CO2 response to hypo- and hyperventilation, at rest and during exercise .

Exercise capacity

The was assessed with an incremental protocol on a cycle ergometer (Corival1000SS, Lode, Groningen, the Netherlands). The workload was set at 20 W and was increased by 20 W every minute until the subject could no longer maintain the pedalling frequency at 60 r.p.m. despite strong verbal encouragement. The subjects breathed through a facemask attached to a volume transducer while gases were continuously sampled for analysis of fractional concentrations of O₂, CO₂ and N₂. The respiratory gas analysis system was calibrated before each test using known standard gases.

response to hypercapnia (CO₂ administration)

The response to hypercapnia consisted of three trials (fraction of inspired CO₂ (F_I,CO2) 0.00, 0.035, 0.05), which was induced by rapidly changing the F_I,CO2. Each F_I,CO2 trial ran for 12 min at approximately 10–15 min intervals. This duration is long enough to permit CO₂ to reach its new steady-state value at the central chemoreceptors (Honda et al. 1983; Poon & Greene, 1985; Pianosi et al. 1994; Teppema et al. 2000). During the interval periods, the subjects inspired room air. Each subject performed these three trials at rest and during exercise. The order of the trials was randomized for each subject. We performed all trials under the hyperoxic condition to abolish the O₂-sensitive chemoreflex (Ohyabu et al. 1982; Robbins, 1988; Mohan & Duffin, 1997).

P_a,CO2 response to hypo- and hyperventilation (voluntary changes in respiration)

The P_a,CO2 response to ventilation consisted of three trials: two periods of hyperventilation and one period of hypoventilation. To avoid the possible effects of different breathing patterns on the relationship, in the hyperventilation trials, both V_T and breathing frequency were altered deliberately by matching the breathing pattern to that recorded during hypercapnia trials, whilst inhaling 0% CO₂ in 40% O₂ with N₂ balance. In the hypoventilation trial, was set to 80% of during the 0.00 F_I,CO2 trial (i.e. during spontaneous breathing). The breathing pattern was estimated from the relationships between and V_T in each subject. Each trial ran for 12 min with an interval of 10–15 min. Each subject performed these three trials at rest and during exercise, and the order of the trials was randomized.

During hypo- and hyperventilation trials, the inspired and expired volume curves were continuously displayed on a screen monitor. Visual and audio signals were constructed from the breathing pattern of the subjects during the hypercapnia trials. The target V_T level was simultaneously displayed on the same screen monitor in each trial. The subjects were instructed to match their volume curve with the target V_T level and to breathe according to the sound of the metronome. As a result, both the V_T and breathing frequency, and thus , were precisely controlled by the visual feedback.

Since our preliminary measurements indicated that P_a,CO2 responses to , and the response to P_a,CO2 reached steady states within 8–12 min, we represented each response by averaging it in the last 2 min. The arterial blood sample (2.5 ml) was collected at minute 11.5 of each trial period. The measured values of operating points (OPs) in the subjects were defined to be the steady-state values for and P_a,CO2 that were obtained during the 0.00 F_I,CO2 trial without visual feedback (i.e. during spontaneous breathing).

Data analysis

The cerebrovascular and ventilatory equilibrium diagram model is depicted in Fig. 1. The respiratory chemoreflex system consists of two subsystems, the central controller (I) and peripheral plant (II). These subsystems act as a feedback control system, which regulates the systemic CO₂ level. In the brain, CBF (III) is influenced by systemic CO₂ and regulates CO₂ at the brain level, which then feeds back into the central controller.

Figure 1. Equilibrium diagram model.

A, systemically, partial pressure of arterial carbon dioxide (P_a,CO2) is controlled by the respiratory chemoreflex system which consists of two subsystems: the central controller (controlling element; I) and peripheral plant (controlled element; II). In the brain, cerebral CO₂ is strongly regulated by change in cerebral blood flow (CBF; III). In addition, brain tissue CO₂ influences the central chemoreflex, and can be characterized by observing CBF response to changes in CO₂. B, in the central controller the input parameter is P_a,CO2, the output parameter is minute ventilation . To characterize the central controller, fraction of inspired CO₂ (0, 3.5, 5% CO₂ in 40% O₂ with N₂ balance) and the relationship were measured. The central controller can be characterized by observing changes in in response to changes in P_a,CO2. In the peripheral plant, input is , and output is P_a,CO2. To characterize the peripheral plant, was altered by hyper- and hypoventilation using a visual feedback method, which made it possible to control both tidal volume and breathing frequency; the relationship was then quantified. Since both relationships share common variables, the resultant operating point of ventilatory response under the closed-loop condition is determined by intersection of the two relationships.

Central controller (I)

Change in P_a,CO2 (input) stimulates chemoreceptor activity and alters ventilation (output) via the central controller. To characterize the central controller, we used a protocol of CO₂ administration (three levels), a conventional linear equation, , and determined the slope S and intercept B using a least-squares regression method. The slope (S) also identifies the gain of the central controller (G_C); .

Peripheral plant (II)

The central controller-induced changes in ventilation (input) also alters P_a,CO2 (output). To characterize the peripheral plant, we used a protocol of voluntary changes in respiration (four levels), the modified metabolic hyperbola as , and determined the values of A and C by the least-squares regression method. During hyper- and hypocapnia, the gain of the peripheral plant (G_P) to operating point (OP) was calculated from the following equations under hyper- (: −2 l min⁻¹ from OP) and hypocapnia (: +2 l min⁻¹ from OP) conditions, respectively.

Total respiratory loop gain (I + II)

Total respiratory loop gain (G_TR) to OP, hyper- and hypocapnia was calculated by the following equations.

Cerebral CO₂ reactivity (III)

P_a,CO2 (input) alters CBF (output) via cerebral CO₂ reactivity. To characterize cerebrovascular reactivity to CO₂, we used protocols of CO₂ administration and voluntary changes in respiration (six levels), an exponential function, %MCA , and determined the values of K and R. The cerebral CO₂ reactivity (G_B) to OP, hyper- and hypocapnia was calculated from the following equations at the OP, and under hyper- (P_a,CO2: +5 mmHg from OP) and hypocapnic (P_a,CO2: −5 mmHg from OP) conditions, respectively.

Statistical analysis

A paired t test was used to assess the differences in the steady-state haemodynamic variables between rest and exercise conditions. Two-way analysis (CO₂ and exercise) of variance with repeated measures was used to assess the differences in the G_P, G_TR and G_B between all conditions. A Student–Newman–Keul's test was employed post hoc when main effects were significant, i.e. P < 0.05. Data are expressed as mean ±s.e.m. and analyses were conducted using SigmaStat (Jandel Scientific Software, SPSS Inc., Chicago, IL, USA).

Results

Averaged O₂ uptake during the cycling exercise was 32 ± 2%. This mild exercise increased HR and caused slight elevations in MAP (92 to 97 mmHg; P = 0.169) and MCA V_mean (51.4 to 53.3 cm s⁻¹; P = 0.085, Table 1). During exercise, both (P <0.001) and P_ET,CO2 (P = 0.004) were increased whilst pH and P_a,CO2 were unchanged.

Table 1.

Ventilatory and haemodynamic variables at rest and during exercise

	Rest (R)	Exercise (E)	R versus E
PET,CO2 (mmHg)	38.2 ± 1.2	44.1 ± 1.7	P = 0.004
Pa,CO2 (mmHg)	43.2 ± 1.7	44.6 ± 01.2	P = 0.294
Pa,O2 (mmHg)	225 ± 17	244 ± 2	P = 0.578
pH	7.40 ± 0.01	7.39 ± 0.01	P = 0.668
(l min−1)	11.6 ± 0.5	26.0 ± 1.2	P < 0.001
VT (ml)	1059 ± 330	1431 ± 163	P = 0.156
(ml min−1)	330 ± 63	942 ± 40	P < 0.001
(ml min−1)	338 ± 24	849 ± 32	P < 0.001
MAP (mmHg)	92 ± 1	97 ± 2	P = 0.169
HR (beats min−1)	69 ± 5	98 ± 8	P < 0.001
MCA Vmean (cm s−1)	51.4 ± 4.5	53.3 ± 4.2	P = 0.085

Values are means ±s.e.m. P_ET,CO2, end-tidal carbon dioxide (CO₂) tension; P_a,CO2, partial pressure of arterial CO₂; P_a,O2, partial pressure of arterial O₂; , minute ventilation; V_T, tidal volume; , oxygen uptake; , CO₂ uptake; MAP, mean arterial pressure; HR, heart rate; MCA V_mean, middle cerebral artery mean blood velocity.

In the central controller, the regression line of the relation was reset to higher and P_a,CO2 during exercise (Fig. 2) without a change in gain (G_C) (1.7 ± 0.3 to 1.8 ± 0.5 l min⁻¹ mmHg⁻¹, P = 0.837; Fig. 3). The functional curve of the peripheral plant also reset to higher and P_a,CO2 during exercise (Fig. 2); however, the change in gain (G_P) was different from that of G_C, i.e. the G_P at the OP, during both hypercapnia and hypocapnia, was decreased from rest to exercise; the change in G_P was especially marked during hypercapnia (−4.1 ± 0.8 to −2.0 ± 0.2 mmHg l⁻¹ min, P = 0.009; Fig. 3). Therefore, total respiratory loop gain (G_TR) during hypercapnia decreased during exercise (−8.0 ± 2.3 to −3.5 ± 1.0 U, P = 0.019) despite no change in G_TR at OP and under hypocapnia. If the change in P_a,CO2 is 2 mmHg, changes are 3.5 ± 0.7 and 3.6 ± 1.0 l min⁻¹ at rest and during exercise, respectively. However, these similar changes cause a different correction in P_a,CO2 between the hypo- and hypercapnia conditions. Under conditions of hypocapnia, change similarly alters P_a,CO2 at rest (8 ± 2 mmHg) and during exercise (5 ± 2 mmHg); however, under conditions of hypercapnia, large differences in alterations in P_a,CO2 are observed between those at rest (16 ± 5 mmHg) and those during exercise (7 ± 2 mmHg).

Figure 2. Characteristics of central controller (I; A) and peripheral plant (II; B) at rest and during exercise.

A (central controller), linearly increased with P_a,CO2 at rest and during exercise. The averaged regression lines were and at rest and during exercise, respectively. B (peripheral plant), the peripheral plant was characterized by a modified metabolic hyperbola. The averaged fitted hyperbolae were and at rest and during exercise, respectively. Arrows denote operating points.

Figure 3. Group-averaged central controller gain (G_C, A), peripheral plant gain (G_P, B) and respiratory total loop gain (G_TR, C) at rest and during exercise.

The central controller was characterized by a conventional linear equation, thus the gain under hypocapnic conditions was the same as that at the operating point (OP) and hypercapnia. The peripheral plant was characterized by a hyperbola therefore gains at OP, hypo- and hypercapnia were analysed. Values are means ±s.e.m.*P <0.05, different from rest; #P <0.05, different from operating point; $P <0.05, different from hypocapnia.

Hypercapnia resulted in an exponential elevation in MCA V_mean during exercise as well as at rest (Fig. 4). However, the functional curve of cerebral CO₂ reactivity was not reset during exercise because of small changes in P_a,CO2 and MCA V_mean. In contrast to the ventilatory chemoreflex, all cerebrovascular reactivities (G_B) to OP, hyper- and hypocapnia were increased during exercise despite unremarkable changes in both K (P = 0.662) and R (P = 0.286) of these curves (Table 2 and Fig. 5). The increases in G_B were more marked in the hypercapnic condition (2.4 ± 0.2 to 2.8 ± 0.2 % mmHg⁻¹, P = 0.025) compared to other conditions (OP, P = 0.049; hypocapnia, P = 0.086).

Figure 4. Characteristics of cerebrovascular CO₂ reactivity (III) at rest and during exercise.

The cerebrovascular CO₂ reactivity was characterized by an exponential function. The averaged fitted exponential equations were MCA and MCA at rest and during exercise, respectively. Arrows denote operating points.

Table 2.

Characteristics of central controller, peripheral plant and cerebrovascular reactivity at rest and during exercise

	Rest (R)	Exercise (E)	R versus E
Central controller (I)
S	1.74 ± 0.32	1.81 ± 0.49	P = 0.837
B	29.2 ± 8.51	21.28 ± 6.62	P = 0.073
Peripheral plant (II)
A	344 ± 54	1124 ± 118	P < 0.001
C	13.2 ± 3.0	3.1 ± 2.8	P = 0.004
Cerebrovascular CO2 reactivity (III)
K	40.0 ± 2.5	38.9 ± 3.1	P = 0.662
R	0.0216 ± 0.0013	0.0232 ± 0.0016	P = 0.286

Values are means ±s.e.m. Central controller (I), ; peripheral plant (II), ; cerebrovascular reactivity to CO₂ (III), MCA .

Figure 5. Group-averaged cerebrovascular CO₂ reactivity (G_B) at rest and during exercise.

The cerebrovascular CO₂ reactivity was characterized by an exponential function therefore G_B values at the operating point and under hypo- and hypercapnia were analysed. Values are means ±s.e.m.*P <0.05, different from rest; #P <0.05, different from operating point; $P <0.05, different from hypocapnia.

Discussion

The main finding of the present investigation was that, under conditions of hypercapnia and exercise, the total respiratory loop gain was markedly reduced. These changes in total loop gain occurred independently of the change in central controller gain because of a marked decrease in peripheral plant gain. Furthermore, cerebrovascular CO₂ reactivity during each condition, especially during hypercapnia, was increased during exercise. These findings indicate that, despite an attenuated chemoreflex system controlling ventilation, elevations in cerebrovascular reactivity might help maintain CO₂ homeostasis in the brain during exercise.

The respiratory chemoreflex

The respiratory chemoreflex is a powerful feedback control system which acts to maintain P_a,CO2 or pH remarkably constant; the tight regulation of pH is critical to maintain homeostatic function for all tissues (Chesler, 2003), especially neural activity. Exercise, which activates muscle metabolism and produces CO₂, causes hyperpnoea via the ventilatory chemoreflex. The exercise-induced hyperpnoea was reflected in elevations in from 12 to 26 l min⁻¹ (P <0.001). The mechanism(s) subserving ventilatory control during exercise remain controversial (Ward, 2007); however, traditionally these mechanisms are proposed to include elements of proportional feedback, central and carotid chemosensory, and feedforward systems, central command and muscle reflex (Dempsey et al. 2006; Waldrop & Iwamoto, 2006; Ward, 2007). Acute hypoxia causes hyperventilation and alkalosis at the medullary chemoreceptors, which reduce their drive (Crawford & Severinghaus, 1978). In addition, the change of this cerebrospinal fluid alkalosis modifies the respiratory control (Severinghaus et al. 1963). Although the mechanism of exercise-induced hyperpnoea is different from that associated with hyperventilation at high altitude, it seems that exercise-induced hyperpnoea alters central chemoreflex. The role of the ventilatory chemoreflex in the regulation of exercise hyperpnoea has been extensively investigated (Cunningham, 1987). The ventilatory sensitivity to hypoxia is increased from rest during exercise (Bhattacharyya et al. 1968); however, the effect of exercise on the respiratory chemoreflex remains controversial. For example, Asmussen & Nielsen (1957) demonstrated that the ventilatory–P_CO2 relationship line shifted to the left without a change in its sensitivity. In contrast, Poon & Greene (1985) showed that the slope of the ventilatory–P_CO2 relationship was increased by exercise. In addition, the ‘chemoreflex response’ is not dictated by the level of chemical drive. Such an integrative response involves a dynamic interaction between the respiratory controller and the chemical drive, and is influenced by respiratory mechanical constraints (Poon et al., 2007). Under a closed-loop condition (Fig. 1), ventilatory output is determined by chemical and metabolic drives, although this ventilatory loading alters these drives in the lung system which feeds back to ventilatory output. However, previous studies have failed to consider the importance of metabolic changes due to the work of breathing (Miyamoto et al. 2004). We have used a new equilibrium diagram model (Miyamoto et al. 2004) to resolve this question. The previous studies by Severinghaus et al. (Severinghaus et al. 1963; Severinghaus & Carcelen, 1964; Crawford & Severinghaus, 1978) demonstrated the regulation of cerebrospinal fluid pH during the hyperventilation associated with high altitude by using a similar model for ventilatory control. The respiratory model of the present study has a limitation in identifying the regulation of cerebrospinal fluid pH or the interaction between central and peripheral chemoreflexes. In the present study, the equilibrium diagram model demonstrates the effect of CO₂ change on respiratory control or the effect of respiratory change on cerebral CO₂ haemodynamics during exercise. Compared with the model used in previous work (Severinghaus et al. 1963; Severinghaus & Carcelen, 1964; Crawford & Severinghaus, 1978), our model gave similar information about respiratory control during conditions of exercise rather than at high altitude.

The regression line of the central controller was shifted rightward and upward around the higher operating point of and P_a,CO2 during exercise (Fig. 2). Both neural and humoral mechanisms may be involved in the ventilatory chemoreflex responses of the central controller associated with exercise. During exercise, increases in peripheral chemoreflex hypoxic sensitivity can be related to lactic acidosis (Asmussen & Nielsen, 1958; Wasserman et al. 1975), circulating catecholamines (Cunningham et al. 1963) and potassium (Linton et al. 1984; Qayyum et al. 1994). However, the sensitivity of the central controller (G_C) was unchanged during exercise (P = 0.837, Fig. 3). This finding may be related to the small changes in lactate, catecholamines and potassium concentrations during such a light exercise workload (32%), or a differential influence of CO₂ on peripheral chemoreflex activity as opposed to hypoxia. Moreover, because oscillations in pH increase during exercise (Band et al. 1980), these changes might also modify the results of respiratory control during exercise as identified in the present study.

The lung system (peripheral plant) is an important subsystem of respiratory chemoreflex, because it is an effector to change CO₂ systemically via an alteration in ventilation (Miyamoto et al. 2004). The sensitivity of peripheral plant is non-linear and is changed by ventilation (Fig. 2). At rest, G_P was increased due to a decrease in ventilation, suggesting that the peripheral plant is more effective in controlling CO₂ at low levels. This effect of ventilatory loading is particularly acute during hyperventilation as the respiratory apparatus is subject to increasing mechanical limitations, i.e. dead space (Poon et al. 2007). Therefore, considering the multiple effects of these subsystems, the central controller is generally more pronounced at low than high levels (Clark et al. 1980; Poon, 1989a,b). The respiratory total loop gain (G_TR) was much higher under the hypercapnic condition caused by hypoventilation compared with normal (P <0.001) and hypocapnic (P <0.001) conditions at rest (Fig. 3).

During exercise the functional curve of the peripheral plant also reset rightward and upward around the higher and P_a,CO2 (Fig. 2). However, the change in G_P was different from that of G_C. From rest to exercise, during hypercapnia and hypocapnia, there was a decrease in G_P at the operating point (OP). Importantly, the change in G_P at hypercapnia was larger (−51%) compared with that at other conditions (OP, −39% and hypocapnia −27%; Fig. 3). The sensitivity of the peripheral plant is non-linear and was decreased exponentially during elevations in ventilation. Therefore, these exercise-induced G_P reductions were related such that the OP moved rightward on the functional curve of the peripheral plant during exercise compared with rest (Fig. 2). The rightward shift of the exercise OP was determined by the resetting of central controller (rightward and upward shift), indicating that the mechanism of the change in G_P during exercise depends on the interaction with alteration in the central controller. As a consequence, total respiratory loop gain (G_TR) at hypercapnia decreased during exercise despite no changes in G_TR at the OP and under the hypocapnic condition. These findings suggest that the respiratory chemoreflex was attenuated during exercise under the hypercapnic condition despite no change in the sensitivity of the central controller. The interaction between the central controller and the plant was non-linear. Moreover, these results were not consistent with the traditional chemoreflex feedback model, which ignores the mechanical plant. The ventilatory response to chemical or exercise inputs is also potentiated by increases in physiological dead space or shunt (Poon et al. 2007). In addition, congestive heart failure patients with increased physiological dead space are reported to have an augmented sensitivity (Wasserman et al. 1997). Therefore, an interaction with the attenuation in peripheral plant gain may be another mechanism underlining the lack of change in the controller gain during exercise.

Cerebrovascular CO₂ reactivity

At rest, hypercapnic cerebral CO₂ reactivity was greater than the hypocapnic reactivity (Fig. 5) because of the increase in CO₂ exponentially elevated MCA V_mean when a wider range of CO₂ challenge was applied (Rasmussen et al. 2006). Animal studies indicate that the mechanisms underlying the normal greater reactivity to hypercapnia compared with hypocapnia may be related to a greater influence of vasodilator mediators on intracranial vascular tone compared with vasoconstrictive mediators (Toda & Okamura, 1998). During exercise, cerebral CO₂ reactivity (G_B) to the OP, in both the hyper- and hypocapnia conditions, was increased. Enhanced cerebral CO₂ reactivity at OP with exercise has been reported (Rasmussen et al. 2006). Our new finding is that the increase in G_B during the hypercapnic condition was much larger compared with other conditions at rest and during exercise. Moreover, cerebral CO₂ reactivity (G_B) to OP (P = 0.049) and hypercapnia (P = 0.025) was increased during exercise while G_B to hypocapnia was unchanged (P = 0.086, Fig. 5). This enhanced cerebral CO₂ reactivity during exercise may relate to interactions with the central controller; however, the mechanism(s) underpinning such changes remain unclear.

The role of autonomic neural control of the cerebral circulation is controversial and, despite rich sympathetic nerve innervation of the cerebral arteries (Nielsen & Owman, 1967; Nelson & Rennels, 1970; Edvinsson, 1975), the traditional thinking is that changes in sympathetic tone appear to have a limited effect on CBF. In contrast, Meadows et al. (2003) found that sleep decreased cerebral CO₂ reactivity, suggesting that the level of cerebral activation influences the cerebrovascular reactivity to CO₂. In addition, sympathetic nervous activation attenuates the CO₂-induced increase in CBF at rest (Jordan et al. 2000). Therefore, exercise-induced physiological changes (e.g. autonomic neural control) may also modify the cerebral CO₂ reactivity. However, these findings contrast with a study which reported that sympatho-excitation induced with lower body negative pressure did not alter the cerebral CO₂ reactivity (LeMarbre et al. 2003). Collectively, the mechanisms underlying heightened sympathetic nerve activity during exercise on the regulation of CBF remain unclear.

Cerebral autoregulation is well maintained during mild and moderate dynamic exercise (Brys et al. 2003; Ogoh et al. 2005a,b, 2007), suggesting that CBF regulation is not influenced by ABP during exercise. However, at rest in the supine position, Aaslid et al. (1989) have reported that cerebral autoregulation is also affected by the basal vascular tone and it is attenuated by hypercapnia. Via sympathoexcitation, arterial blood pressure increases with CO₂ administration (Ainslie et al. 2005). Thus, because of an attenuation in normal cerebral autoregulation under hypercapnic conditions, CBF may be influenced by an increased ABP with CO₂ administration and this phenomenon may be further altered by exercise. During exercise the additional CO₂-induced elevations in blood pressure and a lowered cerebral autoregulation might explain the exponential change in cerebral CO₂ reactivity during hypercapnia.

The interaction between total respiratory chemoreflex and cerebrovascular reactivity

An increase in cerebrovascular CO₂ reactivity compensated an attenuated respiratory chemoreflex system during steady-state exercise, especially under the hypercapnic condition. Although the interaction between systemic and cerebral CO₂ controlling mechanisms during exercise remains unknown, previous investigations (Chapman et al. 1979; Dempsey, 2005; Xie et al. 2005, 2006; Ainslie et al. 2007; Peebles et al. 2007) indicate that cerebral CO₂ reactivity is linked with the ventilatory response to CO₂. Changes in cerebrovascular CO₂ reactivity affect the stability of the ventilatory responsiveness to CO₂ via alterations in the degree of washout in central chemoreceptor hydrogen [H⁺]; these changes have been documented in a range of physiological (Xie et al. 2006; Ainslie et al. 2007) and pathophysiological disorders (Xie et al. 2005). Peebles et al. (2007) reported that hypercapnic cerebral CO₂ reactivity was inversely related to the increase in ventilatory change and suggested that a reduced cerebral CO₂ reactivity resulted in less central CO₂ washout and greater ventilatory stimulus. However, our findings indicate that the relationship between the two systems during exercise cannot be explained only by these mechanisms, because the central controller gain was unchanged during exercise despite an enhanced cerebral CO₂ reactivity. This dissociation may depend on a peripheral chemoreflex distribution to the central controller or CBF distributions to central chemoreflex that are different between rest and exercise. These findings highlight the interdependence of total respiratory chemoreflex to many other variables through a complex and probably non-linear relationship.

Technological considerations

P_ET,CO2 measurement has been used as an estimate of P_a,CO2. The difference between P_a,CO2 and P_ET,CO2 is influenced by metabolic CO₂ production and tidal volume and the relationship between these two variables is not altered by breathing frequency and exercise (Jones et al. 1979). In addition, the estimated different cerebral CO₂ reactivity between rest and exercise from P_a,CO2 was the same as that from P_ET,CO2 (Rasmussen et al. 2006). However, P_ET,CO2 is higher than P_a,CO2 when metabolic CO₂ production and are increased (Jones et al. 1979). Peebles et al. (2007) demonstrated that cerebrovascular CO₂ reactivity is underestimated by P_ET,CO2 when compared with P_a,CO2; therefore, we have made the calculations of central controller, peripheral plant and cerebral CO₂ reactivity using P_a,CO2. Another important consideration is that P_CO2 from the internal jugular vein is likely to be a closer index of brain tissue P_CO2 than P_a,CO2 (Xie et al. 2006) and the medullary central chemoreceptors are not stimulated directly by P_a,CO2; rather, they are stimulated by [H⁺] via alterations in brain tissue CO₂ tension (Peebles et al. 2007). Whilst studies have ‘corrected’ ventilatory reactivity against brain tissue P_CO2 (Xie et al. 2006), these experiments were conducted at rest (Fencl, 1986; Peebles et al. 2007). Because exercise would modify the relationship between P_a,CO2 and internal jugular vein P_CO2 as an index of brain tissue P_CO2, we decided not to ‘correct’ ventilatory reactivity against brain tissue P_CO2. Another potential limitation of estimating MCA V using transcranial Doppler ultrasonography is that changes in the diameter of the isonated vessels could modulate MCA V independently of flow. However, the MCA diameter appears to remain relatively constant in humans under several conditions (Giller et al. 1993; Schreiber et al. 2000; Serrador et al. 2000). In addition, the changes in MCA V_mean during submaximal dynamic exercise appear to be similar to the changes in CBF determined by other techniques, i.e. internal carotid artery blood flow (Hellström et al. 1996) and the ¹³³Xe clearance technique (Jørgensen et al. 1992a,b). It should be noted, however, that the functional anatomy of the arteries supplying the brain varies between individuals. In addition, the proportion of total flow in any single vessel may not be constant, and flow redistribution between major cerebral vessels may occur. Therefore, accurate measurement of global CBF cannot be assured unless simultaneous flow in all major vessels is measured. Third, the change in MCA V_mean in relation to a ‘central controller’versus a ‘peripheral plant’ could not be evaluated in the model of the present study. Each mechanism was identified separately at rest and during exercise. Thus, its relationship remains unclear at rest and during exercise and further studies incorporating linear and non-linear models are needed to provide further insight.