Tracking pulmonary gas exchange by breathing control during exercise: role of muscle blood flow

Populations of group III and IV muscle afferent fibres located in the adventitia of the small vessels appear to respond to the level of venular distension and to recruitment of the vascular bed within the skeletal muscles. The CNS could thus be informed on the level of muscle hyperaemia when the metabolic rate varies. As a result, the magnitude and kinetics of the change in peripheral gas exchange – which translates into pulmonary gas exchange – can be sensed. We present the view that the respiratory control system uses these sources of information of vascular origin, among the numerous inputs produced by exercise, as a marker of the metabolic strain imposed on the circulatory and the ventilatory systems, resulting in an apparent matching between pulmonary gas exchange and alveolar ventilation.

The inability of the cardiovascular system to appropriately increase an already high resting level of O 2 supply to the tissues and CO 2 transport to the lungs requires blood flow to be redistributed toward the exercising muscle wherein muscle metabolic rate increases (Astrand & Rodahl, 1977;Guyton, 1987;Laughlin et al. 1999). The theory presented in this paper is that the changes in vascular resistance within the exercising muscles produce a specific signal that may prove to be essential for the 'respiratory neurons' involved in the coupling between ventilation and peripheral/pulmonary gas exchange. The elements supporting the 'vascular distension hypothesis' are briefly described. For more details the reader can refer to the following reviews (Haouzi et al. 2004b;Haouzi, 2006;Forster et al. 2012). 454 P. Haouzi J Physiol 592.3 Finally, a model will be presented on how this information may be used by the CNS to produce a matching between alveolar ventilation (but not minute ventilation) and the pulmonary gas exchange rate.

Setting the scene: rate of O 2 and CO 2 transport by the blood and muscle blood flow redistribution
The transfer of molecules of oxygen from the atmosphere to the mitochondria of millions of cells to eventually 'feed' the electron chain, along with the elimination of the molecules of CO 2 produced in the process, relies on a profound interaction between the circulatory and the ventilatory systems (Dejours, 1964;Astrand & Rodahl, 1977).
In resting mammals, including humans, the rate of O 2 delivery by the arterial blood (D aO 2 ) -the product of the cardiac output and the arterial O 2 concentrations (C aO 2 ) -is several fold the rate of body O 2 consumption (Astrand & Rodahl, 1977;Dejours, 1981Dejours, , 1988. For instance, in an average-sized adult human, resting D aO 2 is about 1000 ml min −1 , assuming an arterial O 2 content of 200 ml l −1 and a cardiac output of 5 l min −1 (Guyton et al. 1973), while restingV O 2 ranges between 250 and 300 ml min −1 (Dejours, 1990). The volume of O 2 delivered per minute in the arterial tree is therefore 3-4 times the volume of O 2 consumed over the same time. As a consequence, D VO 2 (the rate of O 2 delivery back to the lungs) is only 25% less than D aO 2 (Dejours, 1990). Incidentally, in air-breathing animals, the rate at which CO 2 is transported by the venous system (D VCO 2 ) is also 3-4 times its rate of production: There is, in addition to the CO 2 produced by all the tissues, about 1 litre min −1 of CO 2 coming for the arterial side which must be transported by the venous system towards the lungs (Dejours, 1981). The latter only eliminates, through alveolar ventilation, the equivalent of the endogenous production of CO 2 , 200-250 ml min −1 at rest.
During a dynamic exercise, muscleV O 2 andV CO 2 can increase up 20-fold (Astrand & Rodahl, 1977;Dejours, 1990); D aO 2 to the muscles must therefore rise to prevent or limit a reduction in muscle P O 2 . Similarly, asV CO 2 must rise by about the same amount asV O 2 and at a similar rate -or even at a higher rate during heavy exercise -the only way to limit the rise in muscle CO 2 is for the cardiovascular system to allow for an increase in blood flow to (and therefore from) the metabolically active tissues. As O 2 extraction and cardiac output can, at the very best, increase by 3 times each at the peak of a maximal exercise in a trained athlete (Astrand & Rodahl, 1977;Dejours, 1990), a redistribution of blood flow towards the exercising muscles (Saltin et al. 1968;Rowell, 1974;Rowell & O'Leary, 1990;Laughlin, 1999) is the only mechanism through which O 2 delivery rate can match the increased rate of O 2 demand in the muscles. This redistribution of muscle blood flow involves a decrease in vascular resistance in the metabolically active territories, while peripheral conductance decreases in the non-exercising muscles and tissues, for example the skin or the gastrointestinal system (Mitchell, 1990b).

The challenge of exercise-induced hyperpnoea: 'matching' peripheral/pulmonary gas exchange
The involvement of the ventilatory system during exercise should be viewed as one of the components of the adjustments described in the previous paragraph, as: (1) the level of breathing must cope with the consequences of an increase in pulmonary gas exchange, (Whipp, 1981;Forster et al. 2012), a direct effect of the increase in muscle gas exchange; and (2) the magnitude of the ventilatory response dictates the level of alveolar -and thus arterial -O 2 and CO 2 partial pressures, and contributes to maintaining an adequate rate of O 2 delivery and CO 2 removal (Whipp & Ward, 1998;Haouzi, 2006). Keeping P aO 2 and P aCO 2 constant will certainly limit the 'circulatory cost' required to raise D aO 2 to the exercising muscles. As the circulatory system is a closed circuit (if, as a first approach, one ignores venous capacitance), the changes in pulmonary blood flow and pulmonary gas exchange mirror the averaged gas exchange occurring in all of the peripheral tissues -unless disrupted by experimental means (see below for the effects of venous vs. arterial Note that ventilation increases with no increase in P aCO 2 during a moderate level of exercise, while hypocapnia develops at a higher level of work rate and breathing. occlusion). Consequently, the convection of a gas in the regions of the lungs wherein gas exchange takes place, i.e. the alveolar regions, must increase to prevent alveolar P CO 2 from rising and alveolar P O 2 from dropping in proportion to the rate at which peripheral gas exchange increases. To get a quantitative idea of the importance of the adjustment of the convection of gas to the pulmonary gas exchange, consider that a walking human increases O 2 consumption (and CO 2 production) and thus lung gas exchange by about 3 times. As P aO 2 = P iO 2 -kV O 2 /V A and P aCO 2 = P iCO 2 + kV CO 2 /V A , P ACO 2 (and P aCO 2 ) will increase by 3 times (up to 120 Torr!) and P AO 2 (and P aO 2 ) would be close to zero, ifV A did not increase during this very moderate form of exercise.
As shown in Figs 1 and 2, not only does minute ventilation increase in proportion to the pulmonary gas exchange, but this ventilatory adjustment has kinetics which seem to 'follow' that ofV O 2 orV CO 2 during both the onset of and the recovery from any form of dynamic exercise. This remains true whether a step, an impulse, a ramp or a fluctuating change in work load is applied (Fujihara et al. 1973a, b;Casaburi et al. 1977;Whipp, 1978Whipp, , 1981Haouzi et al. 2004a). An important point needs to be clarified about P aCO 2 homeostatis during exercise: the ventilation tracks in all types of exercise factors related or proportional to the pulmonary gas exchange, but: (1) it is not the pulmonary gas exchange by itself which seems to contain the signal driving breathing; and (2) because the ventilatory kinetics are slower than that of gas exchange, a transient, albeit small, change in P aCO 2 does occur in the unsteady state conditions of a step or sinusoidal change in work load (Whipp, 1978). The fundamental question As the period of change in walking speed decreases, the pulmonary gas exchange response becomes dissociated from the motor activity due to a slower time constant response than the locomotor activity, which follows very precisely the change in walking speed. Minute ventilation follows the change in pulmonary gas exchange, with a reduction in amplitude and an increase in phase lag despite unchanged locomotor and motor control. Similar results have long been reported in humans during cycling exercise (Casaburi et al. 1977). of exercise-induced hyperpnoea is therefore to address the mechanisms accounting for the ventilatory-gas exchange matching. In steady-state and unsteady-state conditions, the result of this matching is to prevent (or limit) the rise in P aCO 2 (and decrease in P aO 2 ), which would result from an increase inV CO 2 (orV O 2 ).
Finally, the evolution of the lung anatomy in air-breathing vertebrates has led to the development of a significant pulmonary dead space (Dejours, 1981), wherein gas circulates between the alveolar regions and the atmosphere without being exchanged. It is alveolar, and not minute, ventilation that controls the levels of CO 2 and O 2 to be exchanged with the blood. The puzzling question is how could alveolar ventilation be regulated while it is minute ventilation (a tidal volume at a given frequency) that is generated by the respiratory neurons. This certainly adds some complexity to the conundrum represented by the mechanism of exercise-induced hyperpnoea.

The 'peripheral vascular distension hypothesis'
The challenge is not to propose a theory that can account for any rise in ventilation (indeed multiple structures evoked during an exercise can stimulate breathing, for example muscle receptors, chemoreception and supra-medullary structures), but rather to propose a mechanism that can account for the magnitude and time course of the ventilatory response and its link to the pulmonary gas exchange (Whipp & Ward, 1992).
Over the last few years, our group, following the original work of Huszczuk et al. (1993), has defended the view that one of the pathways through which the control of breathing may follow metabolic changes in the lungs is to follow metabolic changes in the muscle but based on the neural monitoring of the peripheral vascular events (Haouzi et al. 2004b).
Briefly, skeletal muscle afferent fibres are traditionally divided into four groups according to their conduction velocity. The group III and IV muscle afferent fibres (small myelinated or unmyelinated fibres) do increase breathing (McCloskey & Mitchell, 1972;Amann et al. 2010). The natural stimuli of these endings, besides 'painful' stimuli, include the mechanical distortion of their receptive field (mechanoreception), the accumulation of 'metabolic by-products' of the contractions and an increase in local temperature (Hertel et al. 1976;Kaufman et al. 1993Kaufman et al. , 2002Mense, 1996). Stacey (1969) had already noted that although there is a large range of termination sites within the muscle structures, many group III and IV fibres are originating from the blood vessel adventitia, including the arterioles and venules (Fig. 3). Also, Von Düring & Andres (1990) found striking anatomical relationships between muscle group IV endings and the vessels in the cat. In keeping with these anatomical findings, we found that a population of these fibres did respond to the distension of the vessels, predominantly at venular level, and could monitor muscle blood flow by encoding the degree of recruitment of the post-capillary network (Haouzi et al. 1999). For instance, in 60 slowly conducting afferent fibres present in the dorsal roots coming from the cat triceps surae, we found that 31% of group IV and 15% of group III nerves were stimulated by the vascular smooth muscle relaxant papaverine (2-2.5 mg kg −1 ; Fig. 4). Sixty-two per cent of them were also stimulated by isoproterenol, and more than half of the fibres that were stimulated by papaverine were also stimulated during an occlusion of the vena cava, suggesting that these fibres are located within or close to the venous or venular structures. Finally, a large number of group IV fibres respond to both dynamic contractions and venous distension or vasodilation (Fig. 4).
In accordance with these neurophysiological findings, reflexes triggering ventilation changes could be elicited when altering muscle circulation. For instance, injection of a vasodilatory agent into the isolated hindlimb circulation of a sheep stimulates breathing (Haouzi et al. 1996). This occurs well before the drug could have reached the central circulation and, thus, could have affected the arterial chemo-or baroreceptors. Conversely, total obstruction of the blood flow to and from the limbs at rest or at the cessation of dynamic exercise (Fig. 5) does not stimulate, but actually decreases, ventilation much faster than in control conditions, which decline towards resting levels in humans and in animals (Dejours et al. 1957b;Haouzi et al. 1993;Fukuba et al. 2007). However, impeding, by intravascular occlusion, the circulation from (venous side) or to (arterial side) the hindlimbs during electrically induced    leads to opposite ventilatory outcomes, regardless of the circulatory changes occurring in the central circulation. Indeed, despite a similar reduction in O 2 uptake and blood pressure resulting from obstruction of the caudal vena cava or the distal abdominal aorta, ventilation typically rises when the venous side is occluded and decreases when the arterial side is impeded during exercise Haouzi et al. 2004b). Every condition associated with an impediment of the arterial supply prevents the normalV E response produced both at the onset and during the steady-state response to dynamic contractions in various animal models. A similar reduction inV E response can be observed in patients with peripheral vascular disease of the lower extremities when walking on a treadmill, below their pain threshold (Haouzi et al. 1997). Such reduction in breathing during exercise is observed despite all of the other stimuli being present (e.g. control of movements and intensity of contractions). These results also fit with a puzzling observation that whenever the motor act is dissociated from metabolic or gas exchange changes, the strategy adopted by the ventilatory control system is not to follow factors related to the motor activity but to follow, in a systematic and predictive way (Haouzi et al. 2004a), factors proportional or related to some of the changes associated with the rate at which CO 2 is exchanged in the lungs (Fig. 2). Any concept that neglects this crucial observation cannot account for the fundamental mechanism ofV E control in exercise, as this approach preserves all other inputs to the CNS (Whipp & Ward, 1991).

How to understand the apparent matching between alveolar ventilation and pulmonary gas exchange in exercise
The question is how much the system described in the above section can contribute to the ventilatory response to exercise, wherein minute ventilation can increase by more than 100 l min −1 (Dejours, 1964)? Answering this question requires an understanding of how the structures in the CNS involved in breathing control process the multitude of available sources of information reaching

Figure 5 Effects of vascular occlusion during recovery from exercise in humans
Left panel: example of the breath-by-breath minute ventilation (V E ), and alveolar P CO 2 during recovery from a light level of exercise with intact circulation and while impeding the circulation to the post-exercising legs (Dejours et al. 1957). Right panel: Minute ventilation (V E ), O 2 uptake (V O 2 ), and end-tidal P CO 2 (P ETCO 2 ) during recovery (filled symbols) from a constant work rate cyclo-ergometer exercise performed above the lactate threshold and during recovery with cuffs inflated for 2 min around the upper thigh (open symbols) (adapted from Haouzi et al. 1993). The first arrow indicates the cessation of exercise and cuff inflation; the second arrow indicates the moment of occlusion release. Note that in both studies, the normal ventilatory decline was depressed during cuff occlusion, resulting a large ventilatory deficit, despite expected accumulation of metabolites in the muscle circulation. the medullary and supra-medullary structures. Although the neurophysiological basis to understand this neuronal processing is still lacking, precious information can still be obtained from what we already know on theV E response to exercise, as presented in the following section.
Exercise is a complex physiological condition, which combines the consequences of a motor activity with those related to an increase in metabolism. Motor activity involves the control and effects of elementary muscle contractions and movements (gate, cycling); the structures implicated in this process include specific cortical (Thornton et al. 1998;Williamson et al. 2003) or subcorticical regions (Eldridge & Waldrop, 1991;Eldridge, 1994) as well as peripheral inputs related to contractions or movement (Kniffki et al. 1978;Mense, 1992;Kaufman et al. 2002). The metabolic changes can trigger many different inputs: (1) an increase in CO 2 venous content and venous blood flow (Greco et al. 1978;Bennett et al. 1984) and a decrease in the mixed venous O 2 content which could in turn affect the arterial chemoreceptors (Phillipson et al. 1981); (2) a change in the chemical composition in the muscles (Mense & Stahnke, 1983;Kaufman & Rybicki, 1987) affecting muscle afferents; (3) an increase in local and systemic temperature (Dejours et al. 1958;Budzinska, 1975;Hertel et al. 1976); and (4) an increase in systemic and muscle blood flow potentially stimulating receptors located in the central (Jones et al. 1982;Huszczuk et al. 1983) as well as peripheral circulation (Haouzi, 2009), while blood flow decreases in many non-exercising tissues. Despite the fact that all of these inputs, with their different magnitude and time constants, can 'reach' the CNS in an almost

Figure 6. Schematic representation of the 'vascular hypothesis' presented in this review
Any event distending the venular system (venous occlusion, hyperemia or mechanical deformation of the receptive field in keeping with volume of blood present) can in turn stimulate group III and IV endings.
infinite number of combinations and can all increase breathing separately, the ventilatory response to exercise seems to follow the metabolism in a rather simple and perfectly predictive manner (Whipp, 1978;Whipp et al. 1982). The strategy used by the 'respiratory neurons' seems to rely on properties that have more to do with the selection of information and pattern recognition than to the integration of individual inputs. In other words, all of the signals produced by exercise are not qualitatively important in translating into an increase in breathing. According to this view, the CNS does not respond in proportion to (or as a function of) a given stimulus, which could be predicted by, for instance, the elementarẏ V E response to CO 2 , hypoxia, a muscle contraction or a change in blood flow. The respiratory control system seems to use all of the various sources of information to determine that an exercise is actually being performed and to select the most relevant sources of information for adjusting breathing (see Fig. 2). Tracking the change in peripheral gas exchange, via the change in circulation, may well be part of this strategy. This is, after all, what the response to 'sinusoidal' or 'impulse' exercise reveals (Whipp, 1981;Haouzi, 2006). Similarly, when there are antagonist sources of information, such as dissociating venous return to the lungs from the arterial supply, the strategy adopted by the CNS always seems to follow factors associated with an increase in the local vascular response, as if it were the only relevant information for controlling breathing related to exercise (for discussion see Haouzi, 2006). By contrast, the respiratory control system may not be able to recognize that an exercise is being performed when different conditions are created, such as in patients with peripheral vascular disease (Haouzi et al. 1997) or, even more dramatically, during a cardiac arrest, resulting in a completely novel and unpredictableV E response ). With such a view, the matching betweenV A and pulmonary gas exchange cannot be predicted by any signal considered individually  Finally, the debate over the fact that it isV A and noṫ V E that appears to be regulated while it isV E which is generated could be understood by the constraint imposed on the respiratory system by the 'non-proportional' relationship between the amplitude and the duration of any breath (Haouzi et al. 2007;Haouzi & Bell, 2009;Haouzi, 2011). Indeed, it is the change in dead space ventilation (V D ) and not in dead space (V D ) that can disrupt blood gas during exercise; we have recently proposed that there is an obligatory relationship between the amplitude and duration of any breaththis relationship is not proportional but has a positive intercept of the magnitude of V D , which results in an apparent regulation ofV A (Haouzi et al. 2007;Haouzi & Bell, 2009) whatever the level of breathing frequency is adopted. Following such relationship, while it is minute J Physiol 592.3 ventilation (the product of a tidal volume by the breathing frequency) that is 'generated' by the medullary and spinal respiratory moto-neurons (Mitchell, 1990a), it isV A that is being regulated.

Concluding remarks
The view presented in this paper is that the volume of blood at the venular level in the muscles could constitute a crucial stimulus informing the CNS that metabolism is changing in peripheral tissue. The CNS produces a ventilatory output that follows the levels of gas exchange but which, intriguingly, appears to neglect other signals, at least in terms of their kinetics. The matching between alveolar ventilation and pulmonary gas exchange observed during exercise may result from a complex process in which the ventilatory strategy chosen by the CNS is determined by (1) the magnitude and kinetics of the vascular response in the muscles; (2) all other inputs produced by exercise, related or not to gas exchange or to one of its circulatory surrogates, which inform the respiratory neurons that an exercise is being performed (see Fig. 2); and (3) the fundamental relationship between the amplitude and the duration of any breath, which results in a regulation of alveolar rather than minute ventilation.