Through oxygen-dependent release of the vasodilator ATP, the mobile erythrocyte plays a fundamental role in matching microvascular oxygen supply with local tissue oxygen demand. Signal transduction within the erythrocyte and microvessels as well as feedback mechanisms controlling ATP release have been described. Our understanding of the impact of this novel control mechanism will rely on the integration of in vivo experiments and computational models.
Erythrocytes play a fundamental role in tissue oxygen supply via the controlled release of ATP in areas of increased oxygen need.
The matching of oxygen supply with oxygen demand in metabolically active tissue is a fundamental physiological process. Although a number of theories to explain this critical function have been proposed, none have been either universally accepted or fully tested in the intact microcirculation. Any attempt to comprehend the mechanism(s) by which oxygen delivery and metabolic need are matched must, by necessity, be based on the combination of experimental observations and theoretical models. In 1919, August Krogh, in conjunction with a mathematician colleague, K. Erlang, proposed the first oxygen transport model based on the assumption that each capillary is the sole supplier to a cylindrical region of tissue surrounding it (the Krogh cylinder) (55). Under this simple model, oxygen is assumed to diffuse in the radial direction with a uniform diffusivity and to be consumed in the tissue at a uniform constant rate (FIGURE 1A⇓). This model permitted Krogh to propose that capillary density must be actively regulated in response to changing metabolic activity to ensure adequate tissue oxygenation. Specifically, this model predicts that, in response to a fall in oxygen delivery, capillary density would increase to maintain tissue oxygenation. However, no mechanism for this alteration was proposed. Krogh’s model and the interpretations derived from it have continued to influence the way most researchers and clinicians view tissue oxygenation. However, recent studies of microvascular oxygen transport have demonstrated the inadequacies of Krogh’s idealized single capillary model as a comprehensive descriptor of tissue oxygenation.
Over 30 years ago, Duling and Berne (19) reported that oxygen levels in the blood diminished along the arteriolar tree with up to two-thirds lost before entering the capillary bed. Numerous researchers, using a variety of techniques in different organs and species, have confirmed this finding (23, 90, 92). Thus it is clear that there is exchange of O2 that occurs among microvessels that is not accounted for by the Krogh model. Although it is not fully understood why there is such a large precapillary decrease in O2, Ellsworth and Pittman (26) provided evidence that some of the O2 leaving the arterioles diffuses to erythrocytes flowing through nearby capillaries, resulting in an increase in their O2 saturation. Since O2 is transported by diffusion from arterioles to capillaries, it is also likely that O2 exchange occurs between capillaries with different O2 levels (23), as proposed in theoretical models (27, 37). This exchange would be consistent with quantitative studies of microvascular blood flow, which have demonstrated considerable spatial heterogeneity of capillary perfusion (93) and a corresponding heterogeneity in O2 delivery. Thus groups of capillaries, rather than the single capillary on which Krogh’s model was based, need to be considered in evaluating the mechanisms responsible for maintaining tissue O2 requirements and uniform O2 delivery. As depicted in FIGURE 1B⇑, the current view is that the microcirculation supplies oxygen to tissue using multiple vessel types that are functionally connected through both convective oxygen flow (blood flow) and diffusive oxygen transfer (23, 37). Inherent in this complexity is the need for a mechanism to direct the convective oxygen flow to regions of the tissue where oxygen is required to satisfy metabolic needs. This directed distribution of peripheral blood flow is analogous to the mechanisms inherent in the pulmonary circulation that direct blood flow to those regions of the lung that are well ventilated (96).
One might expect that mechanisms associated with blood flow regulation such as wall shear stress and pressure (myogenic tone) would be sufficient to properly distribute blood flow and, consequently, O2 supply within the arteriolar tree. However, at the microvascular level, total blood flow may not accurately reflect local O2 supply. The convective distribution of oxygen within the arteriolar tree is impacted by the diffusional loss of oxygen to nearby erythrocytes in microvessels. Most of the diffusional loss of oxygen in the arteriole is from erythrocytes near the wall of the vessel, that is, at the edge of the cell free plasma layer. Thus these cells have a lower O2 saturation than those nearer the center of the vessel (9, 25). At vessel bifurcations, this uneven distribution affects convective oxygen supply to a greater or lesser extent depending on the relative flow in the downstream branches. For example, if blood flow into a downstream branch is very low, only plasma and the erythrocytes with lower O2 will be skimmed into this branch (FIGURE 2⇓). As flow into the side branch increases as a result of downstream vasodilation, flow into that branch will be derived from a larger cross-sectional region of the lumen of the upstream vessel, allowing blood with higher hematocrit and erythrocytes with higher O2 saturation to enter that branch. Thus an increase in O2 demand in one region of tissue would induce a redistribution of erythrocytes (O2 delivery) not solely described by the simple redistribution of blood flow. In light of these complexities, it becomes abundantly clear that the simplistic Krogh approach to oxygen delivery will not adequately explain the complex interactions apparent in the intact microvasculature. Therefore, a local control mechanism must exist that is capable of sensing O2 need and adjusting flow within the small arterioles feeding the capillary network.
Numerous studies have focused on the blood vessels themselves or on discrete regions within the tissue as the sensor of O2 requirements in a tissue (18, 23, 26, 27, 43, 49, 64, 72). However, the exquisite accuracy in matching O2 supply with tissue O2 demand in skeletal muscle in vivo mandates a control system that is both more sensitive and responsive than those previously described. The system required must be able to sense localized need and to initiate an integrated response that results in appropriate increases in local oxygen supply. Could this controller be as simple as the erythrocyte itself?
A Case for the Erythrocyte as a Vascular Controller
Fundamental to any system that regulates the delivery of appropriate amounts of oxygen to meet changing tissue metabolic needs is the requirement that the need be detected, quantified, and subsequently coupled to a mechanism that will appropriately alter blood flow (O2 delivery). Such a mechanism requires interplay among tissue gas exchange, tissue metabolism, and vascular smooth muscle function. Moreover, the process must be regulated within a narrow range (45). The vascular endothelium participates in controlling vascular caliber (7, 32, 54, 94) and coordinating the response to local, diverse stimuli initiated within the tissue (13, 17, 32, 44, 76, 77, 78, 94). It would be reasonable to suggest that one or more components of the oxygen transport pathway communicate directly or indirectly with the endothelium to appropriately alter microvascular perfusion.
In 1993, Stein and Ellsworth suggested (88) that, in severe hypoxia, the oxygen content of the blood supplying the tissue was more important than its oxygen tension for the maintenance of oxygen supply in hamster skeletal muscle. Oxygen content (oxygen saturation), a reflection of the extent of binding of oxygen to hemoglobin within the erythrocyte, is related to oxygen tension by the characteristic oxyhemoglobin dissociation curve. Oxygen tension determines the diffusive transfer of oxygen from the erythrocyte to the tissue. Thus, if oxygen content rather than oxygen tension were the important factor in regulating oxygen delivery, then the erythrocyte itself would assume a central role in the process since it contains the only component of the oxygen transport pathway that is directly influenced by oxygen content, hemoglobin. The oxygen content of the erythrocyte as it traverses a tissue is directly linked to the level of oxygen utilization of that tissue (FIGURE 3⇓). Therefore, if the erythrocyte itself were able to sense oxygen need and affect an alteration in vascular caliber leading to appropriate changes in blood flow, this property of the erythrocyte would provide an efficient means of matching oxygen delivery (blood flow) with metabolic need, eliminating the requirement for a diverse network of sensing sites throughout the vasculature. It is intriguing to think that the mobile erythrocyte, whose level of oxygen content at a particular point in a tissue is directly linked to the level of oxygen utilization by that tissue, could itself augment blood flow and oxygen delivery wherever and whenever the need might arise.
The establishment that the erythrocyte, the major supplier of oxygen, also functions as a sensor of oxygen requirements and affector of changes in oxygen supply would provide an important level of precision to local vascular control. How could such a small, anucleated bag of hemoglobin accomplish this task? Erythrocytes contain millimolar quantities of adenosine 5′ triphosphate (ATP) (65), which is produced primarily by membrane-bound glycolytic pathways. In 1992, Bergfeld and Forrester reported that human erythrocytes release ATP in response to the combined effects of hypoxia and hypercapnia (5). More recent studies have shown that exposure to reduced oxygen tension (~35 Torr) alone is sufficient to stimulate ATP release from erythrocytes of hamsters (24), rabbits (85), rats (50), and humans (84). Although these studies have examined ATP release in response to reductions in oxygen tension, Jagger et al. (50), demonstrated that ATP efflux was linearly related to hemoglobin O2 saturation, suggesting that the conformational change of hemoglobin, as it desaturates during a fall in oxygen levels, elicits the release of ATP. This conclusion was further supported by an observed inhibition of ATP release on exposure of the erythrocytes to carbon monoxide, which would prevent the conformational change of hemoglobin in response to a drop in oxygen tension (50). The importance of hemoglobin oxygen saturation in ATP release was later confirmed in human studies by González-Alonso and his collaborators (38, 39, 75). If release of ATP from erythrocytes is directly linked to a physiological stimulus, such as a decrease in oxygen saturation, then there must be a signal transduction pathway in the erythrocyte connecting the stimulus to the release.
A Proposed Signal Transduction Pathway for ATP Release from Erythrocytes
Strong evidence exists supporting the controlled release of ATP from erythrocytes in response to both physiological and pharmacological stimuli. Physiologically, erythrocytes release ATP in response to mechanical deformation (82, 83, 86), as would be encountered when these cells traverse the microcirculation, as well as in response to exposure to reduced oxygen tension (5, 15, 22, 50, 70). In both cases, the amount of ATP released is influenced by the magnitude of the stimulus. In addition to physiological stimuli, erythrocytes release ATP in response to receptor-mediated activation of erythrocyte membrane-bound β-adrenergic receptors or prostacyclin receptors in a concentration-dependent manner (68, 80).
The finding that ATP is released in a controlled fashion suggests that erythrocytes possess a mechanism that directly links the stimuli to release of ATP. Over the past several years, Sprague’s group has delineated a signaling pathway for ATP release from erythrocytes (FIGURE 3⇑). This pathway includes the heterotrimeric G proteins Gs and Gi (68, 69, 70, 81), adenylyl cyclase (AC) (86), protein kinase A (86), and the cystic fibrosis transmembrane conductance regulator (CFTR) (83). Studies have clearly shown that activation of Gs-coupled β-adrenergic receptors and prostacyclin (IP) receptors in erythrocytes results in concentration-dependent increases in cAMP and ATP release from erythrocytes (6, 68, 80). However, Olearczyk et al. demonstrated that the G protein that is activated when erythrocytes are exposed to deformation or reduced oxygen tension is not Gs but rather Gi (69, 70). The direct activation of Gi with the wasp venom extract mastoparan 7 stimulates increases in cAMP and ATP release from erythrocytes (69, 70). Although the α subunit of Gi is well known to inhibit the activity of some AC isoforms, its associated β subunits, specifically subunits 1 through 4 (4, 28, 33, 60, 89, 91), have been shown to activate AC isoforms II, IV, and VII. Importantly, Giα, β subunits 1 through 4, and AC II are all components of human erythrocyte membranes (81). A central role for Gi in the pathway for ATP release from erythrocytes in response to reduced oxygen tension was demonstrated by the finding that incubation of erythrocytes with pertussis toxin, which binds to Gi preventing its dissociation into the component subunits, prevented ATP release in response to this physiological stimulus (70). Although Gi activation is clearly involved in ATP release from erythrocytes exposed to reduced oxygen tension, the mechanism that directly couples the decrease in oxygen saturation of the hemoglobin molecule to the activation of Gi remains under investigation. One possibility is that the conformational change in the hemoglobin molecules bound to the erythrocyte membrane directly activate Gi or some other aspect of the release pathway (50).
Vascular Control by ATP
If we accept that ATP is released from erythrocytes in a controlled manner as they perfuse a region of tissue with a low SO2, then this ATP must initiate a conducted vasodilation that extends beyond the site of initiation for there to be an effective increase in vascular perfusion (O2 delivery) (56) (FIGURE 3⇑). Using arterioles in the intact hamster cheek pouch retractor muscle, McCullough et al. (59) demonstrated a dose-dependent conducted vasodilator response to the intraluminal application of ATP with the maximum dilation occurring at 10−6 M. Importantly, similar amounts of adenosine were ineffective in producing a conducted response in these vessels. The vasodilator response initiated by ATP was conducted as far as 1,200 μm upstream at a rate of approximately 50 μm/s. Since one would anticipate that the oxygen tension at the downstream end of the capillaries and in the venules would be most reflective of local tissue oxygen utilization, Collins et al. (10) investigated the impact of application of similar amounts of ATP into collecting venules. They observed a similar conducted vasodilation, the speed of which was influenced by the architecture of the intervening vasculature. It is important to note that the intra-arteriolar application of 10−6 M ATP, the concentration that produced the maximum conducted vasodilation, is of the same order of magnitude as would be predicted to be released from erythrocytes perfusing a microvessel within a hypoxic tissue region (24, 59). Dietrich et al. (15) later established that the time course for sensing of a low-oxygen environment, the release of ATP from erythrocytes, and a vasodilatory response is on the order of 500 ms, supporting the potential physiological importance of this control mechanism. In recent studies in which the oxygen tension on the surface of an intact muscle was lowered in a stepwise fashion using a computer-controlled gas flow chamber, the increases in flow that occurred within the capillary bed were consistent with such a time course (21).
When ATP is released into the vascular lumen, it can interact with receptors present on the endothelium that can elicit both endothelium-dependent and smooth muscle cell-dependent vasoactive responses, which would be conducted along the vasculature (FIGURE 4⇓). Endothelial cells possess purinergic receptors that, when activated, stimulate the synthesis and release of several vasodilators. These vasodilators include nitric oxide (NO) as well as products of arachidonic acid metabolism. Although the receptor activated and the mediator released may vary in different tissues, the receptors present in the cerebral circulation have been particularly well characterized.
In the cerebral circulation, ATP and its related breakdown products, including adenosine, are vasoactive mediators. Forrester et al. found that intraluminal ATP was a potent vasodilator in the baboon cerebral circulation (29) and that topically applied ATP dilated cat pial vessels at a much lower concentration than adenosine (30). These reports led to the hypothesis that ATP, or a closely related breakdown product rather then adenosine, is the vasoactive purine in the cerebral circulation (30). Although in the cerebral microcirculation ATP, ADP, and adenosine are all potent vasodilators (14, 53), only ATP and ADP consistently caused conducted vasomotor responses (14, 53). Since at equimolar concentrations ADP was slightly less potent in causing dilatory local and conducted responses (53) than ATP, it is likely that ATP may be the primary agonist. This corresponds with observations by Ikeuchi and Nishizaki where, in brain artery endothelial cells, ATP caused stronger potassium currents than ADP, whereas AMP had no effect (48).
Numerous studies have attempted to link neuronal activation, local oxygen tension, and cerebral blood flow. Neuronal activation can lead to a small, but significant, drop in local oxygen tension, which is followed by an increase in local blood flow (2). This neuronal activation-induced decrease in oxygen tension could elicit ATP release from erythrocytes, resulting in an increase in blood flow. Indeed, ATP is released from erythrocytes perfusing isolated cerebral arterioles within fractions of a second after entering the hypoxic vessel (15). Although not all studies have detected the drop in oxygen tension associated with neuronal activation (79), there is increasing evidence suggesting that, in the brain, ATP released from erythrocytes could contribute to a rapid initial increase in blood flow in response to neuronal activity.
ATP released from erythrocytes is known to activate specific P2 purinergic receptors on the vascular endothelium. Although it is simpler to consider two distinct classes of these P2 receptors, P2X and P2Y, more recent evidence indicates that the mammalian P2X and P2Y receptors are actually families of receptors consisting of seven P2X receptors (P2X1–7) and eight P2Y receptors (P2Y1, P2Y2, P2Y4, P2Y6, P2Y11–14) (1, 34). Studies in cerebral microvessels have delineated the specific receptors present and the vasoactive response to their stimulation. In isolated nonpressurized pial arterioles, Lewis et al. concluded that P2Y2- and P2Y6-like receptors are responsible for the observed vessel constriction (57). In isolated and pressurized rat penetrating arterioles, Horiuchi et al. reported that ATP constricts the vessels transiently via smooth muscle P2X1 receptors and dilates the vessels via endothelial P2Y2 (47). Similarly, You et al. found that, in larger size cerebral arterioles, dilation to intraluminal ATP results from P2Y2 stimulation (98). These receptors were confirmed to exist on hamster skeletal muscle arterioles (unpublished observations).
Regardless of the receptor that is activated, ATP-induced vasodilation is the result of the synthesis and activity of endothelium-derived relaxing factors. In skeletal muscle, McCullough (59) and Collins (10) each observed that the conducted vasodilation to intraluminal ATP was eliminated following administration of a NO synthase inhibitor implicating NO as an important vascular mediator in these vessels. In addition, in large cerebral arterioles, endothelial P2Y2-specific stimulation was shown to release NO as well as a non-NO, non-cyclooxygenase-dependent factor (97, 98). Similarly, it was reported that in cerebral arterioles, endothelial P2Y1-specific stimulation releases both NO and a non-NO, non-cyclooxygenase-dependent factor, whereas P2Y2-specific stimulation releases only the non-NO, non-cyclooxygenase-dependent factor (46), possibly a cytochrome P450 monooxygenase product such as epoxyeicosanoic acids (EETs) (16). EETs can activate calcium-sensitive potassium channels, resulting in hyperpolarization (8, 31). In brain artery endothelial cells, purines induced strong potassium currents with ATP > ADP > AMP (48). In cerebral arterioles, the dilation to ATP was preceded by hyperpolarization, and the dilation to ATP depended on large conductance BKCa and intermediate conductance IKCa calcium-sensitive potassium channels but not small conductance SKCa channels (16). Thus the mediator released in response to activation of endothelial purinergic receptors by ATP will depend on the identity of the specific receptor activated and the signaling pathway present in a particular blood vessel (FIGURE 4⇑).
Although there is significant evidence to suggest that NO is a regulator of vascular perfusion, controversy remains as to its source and mechanism of action. In 1996, Stamler and colleagues (51) proposed that the erythrocyte was responsible for regulating O2 delivery through the transport of NO, produced in the lungs, to the periphery in the form of the bioactive compound, S-nitrosothiol (SNO). SNO, reported to be a potent vasodilator, is carried by hemoglobin and released as the hemoglobin O2 saturation falls in response to local O2 demand. Although there is support for this hypothesis (58, 87), questions remain as to its role under physiological conditions (35, 36, 71, 74). Recent work from the laboratories of Gladwin and Patel has provided evidence that deoxyhemoglobin acts as a nitrite reductase, converting nitrite to NO, and hence making it possible for the erythrocyte to vasodilate arterioles in response to low SO2 (11, 36, 71). However, it is not clear that this mechanism has sufficient temporal resolution to account for the observed rapid changes in the distribution of perfusion in response to local decreases in O2 tension. Importantly, unlike erythrocyte-derived ATP, neither SNO nor nitrite have been shown to be associated with conducted vasodilation of upstream resistance vessels in an intact vascular bed. Therefore, although each of these mechanisms could play a role in the regulation of certain aspects of tissue perfusion, neither would appear to fulfill the role we propose for the release of ATP from erythrocytes in response to local O2 need in the normal microvasculature.
Control of ATP Release In Vivo: Feedback Mechanisms
Although ATP can stimulate the synthesis and release of multiple endothelium-derived vasodilators, ATP-induced increases in NO are important in both the cerebral and skeletal muscle microcirculations. It is important to recognize that NO synthesized in endothelial cells diffuses not only to vascular smooth muscle, where it stimulates vasodilation, but it is also released into the vascular lumen. When erythrocytes enter a microcirculation in which large amounts of NO are already present, additional ATP release would be unnecessary. Indeed, NO was shown to attenuate agonist-induced ATP release from erythrocytes in a negative-feedback fashion. In these studies, incubation of erythrocytes with the NO donor spermine NONOate results in inhibition of ATP release from rabbit and human erythrocytes exposed to decreased oxygen tension (67). It has also been proposed that ADP, the first degradation product of ATP, can inhibit ATP release from erythrocytes (95). Under this hypothesis, ATP released from the erythrocytes in response to physiological stimuli is metabolized by ecto-enzymes, resulting in the generation of ADP. ADP can then activate P2Y13 receptors present on erythrocytes resulting in decreases in intracellular cAMP and reduced ATP release (95). The finding that ATP release from erythrocytes is inhibited by the endothelium-derived vasodilator released from the endothelium in response to ATP (NO) as well as the first degradation product of that nucleotide (ADP) demonstrates the potential for negative feedback regulation of this physiologically important signaling pathway.
Pathological Consequences of a Defect in this Control Mechanism
Recently, it was reported that erythrocytes of humans with Type 2 diabetes (DM2), a condition in which cardiovascular disease accounts for nearly one-half of associated deaths, express decreased amounts of the α2 subunit of the heterotrimeric G protein Gi compared with erythrocytes of healthy humans (86). Interestingly, decreased Gi expression is present in animal models of diabetes as well (40–42, 63, 66, 99). In humans, Giα2 expression was decreased selectivly; that is, expression of other Gi αsubtypes, Gsα, and AC II was unaltered (86). This selective decrease in Giα2 expression in human erythrocytes was associated with impaired cAMP synthesis as well as with decreased ATP release in response to an agent that directly activates Gi, mastoparan 7 (86). It is of interest that the degree of impairment of ATP release correlated inversely with gylcemic control. That is, the higher the average blood sugar, as measured by glycated hemoglobin level (HbA1C), the greater the defect in ATP release (86). Since vascular complications in humans with DM2 also correlate inversely with HbA1c level, this finding raises the possibility of a connection between the failure of erythrocytes to release ATP and the vascular complications of DM2 (12, 61). Moreover, such reports suggest that the erythrocyte could be a novel target for the development of drugs for the treatment of vascular insufficiency.
How the Controlled Release of ATP from Erythrocytes in the Microcirculation Changed the Conceptual Approach to our Understanding of Oxygen Supply
Our understanding of the mechanisms that control matching of O2 supply with demand has evolved from the concept of a single, idealized capillary supplying oxygen to a simple Krogh cylinder into a vastly more complex system of oxygen delivery and regulation. This evolution resulted from a more complete understanding of the diverse rheological and transport properties of the microvasculature as well as the concept of the erythrocyte as a sensor of O2 need and regulator of vascular tone via its release of ATP. As a result, a significantly more complex modeling approach than that provided by the pioneering work of Krogh is required to describe this system. It is clear that, to make the model more comprehensive, additional variables must be included such as 1) 3D network geometry of capillaries, arterioles, and venules; 2) realistic hemodynamics of erythrocyte flow and O2 distribution within these networks; 3) 3D oxygen transport between blood and tissue; 4) changes in erythrocyte-derived ATP release as a function of oxygen content; 5) local and conducted changes in microvessel diameter; and 6) other microvascular control mechanisms based on local wall shear rate (shear-dependent vasodilation) and blood pressure (myogenic factors). Krogh’s original model of oxygen supply assumed that increased oxygen delivery to tissue resulted from an increase in the number of perfused capillaries (“capillary recruitment”). However, in vivo experiments in skeletal muscle (20) indicate that most capillaries are perfused under resting conditions. As such, increased flow would result in a more uniform distribution of erythrocytes to already perfused capillaries rather than perfusion of additional capillaries.
Most of the above features have been included in theoretical models of diameter adaptation in microvascular networks (73) and more recently in a theoretical model of blood flow regulation (3). This latter model is consistent with the hypothesis that regulation of microvascular O2 delivery based on O2 saturation-dependent release of ATP by erythrocytes leading to conducted vasodilation can account for experimentally observed increases in perfusion in response to increased oxygen demand. However, this idealized model used a simulated network consisting of only seven representative segments and did not consider diffusive exchange of O2 between capillaries and arterioles. Therefore, it remains to be shown whether the features described above, including erythrocyte-derived ATP, when included in a more comprehensive model, can fully explain the O2-dependent microvascular flow regulation that occurs in vivo. In addition, other key issues that need to be addressed include 1) whether there is a primary location within the network where the need for changes in O2 supply are sensed or control is exercised throughout the arteriolar tree to ensure proper distribution of O2 supply; 2) the inherent time scale of the regulatory system (i.e., determining how fast can the system respond and what the limiting factors are); 3) the vessels in the network that are involved in the response to a given stimulus; and 4) the limitations of the system’s ability to control local O2 delivery. Future progress toward understanding the role of erythrocyte-derived ATP in the regulation of O2 delivery within the microcirculation will depend on a combination of new in vitro experiments, detailed computational modeling, and model testing and validation using in vivo experiments under a range of physiological and pathophysiological conditions.
Original work in the authors’ laboratories is supported by the National Institutes of Health, American Diabetes Association, Heart and Stroke Foundation of Ontario, and Canadian Institutes of Health Research.
- Copyright © 2009 the American Physiological Society