Contribution of non-endothelium-dependent substances to exercise hyperaemia: are they O2 dependent?


  • Janice M. Marshall,

    1. School of Clinical & Experimental Medicine, College of Medical & Dental Sciences, The Medical School, Vincent Drive, University of Birmingham, Birmingham B15 2TT, UK
    Search for more papers by this author
  • Clare J. Ray

    1. School of Clinical & Experimental Medicine, College of Medical & Dental Sciences, The Medical School, Vincent Drive, University of Birmingham, Birmingham B15 2TT, UK
    Search for more papers by this author

  • This report was presented at The Journal of Physiology Symposium on Blood flow regulation: from rest to maximal exercise, which took place at the Main Meeting of The Physiological Society, Edinburgh, UK on 3 July 2012. It was commissioned by the Editorial Board and reflects the views of the authors.

J. M. Marshall: School of Clinical & Experimental Medicine, College of Medical & Dental Sciences, The Medical School, Vincent Drive, University of Birmingham, Birmingham B15 2TT, UK. Email:


Abstract  This review considers the contributions to exercise hyperaemia of substances released into the interstitial fluid, with emphasis on whether they are endothelium dependent or O2 dependent. The early phase of exercise hyperaemia is attributable to K+ released from contracting muscle fibres and acting extraluminally on arterioles. Hyperpolarization of vascular smooth muscle and endothelial cells induced by K+ may also facilitate the maintained phase, for example by facilitating conduction of dilator signals upstream. ATP is released into the interstitium from muscle fibres, at least in part through cystic fibrosis transmembrane conductance regulator-associated channels, following the fall in intracellular H+. ATP is metabolized by ectonucleotidases to adenosine, which dilates arterioles via A2A receptors, in a nitric oxide-independent manner. Evidence is presented that the rise in arterial inline image achieved by breathing 40% O2 attenuates efflux of H+ and lactate, thereby decreasing the contribution that adenosine makes to exercise hyperaemia; efflux of inorganic phosphate and its contribution may likewise be attenuated. Prostaglandins (PGs), PGE2 and PGI2, also accumulate in the interstitium during exercise, and breathing 40% O2 abolished the contribution of PGs to exercise hyperaemia. This suggests that PGE2 released from muscle fibres and PGI2 released from capillaries and venular endothelium by a fall in their local inline image act extraluminally to dilate arterioles. Although modest hyperoxia attenuates exercise hyperaemia by improving O2 supply, limiting the release of O2-dependent adenosine and PGs, higher O2 concentrations may have adverse effects. Evidence is presented that breathing 100% O2 limits exercise hyperaemia by generating O2, which inactivates nitric oxide and decreases PG synthesis.

  • image

[ Janice Marshall is Bowman Professor of Physiology in the College of Medical and Dental Sciences, University of Birmingham. She has wide reaching interests in the ways in which neural and local influences interact to regulate the cardiovascular system in health and disease, with a particular focus on the mechanisms that regulate O2 delivery in systemic hypoxia and exercise. Clare J. Ray is a lecturer in Cardiovascular and Respiratory Sciences in the College of Medical and Dental Sciences at the University of Birmingham. Her research interests are in cardio-respiratory integration and control, with particular focus on the local control of skeletal muscle blood flow.]




cytochrome oxidase


maximal voluntary contraction


nitric oxide


nitric oxide synthase








inorganic phosphate


partial pressure of O2

inline imageO2

oxygen consumption


Over the last 10 years or so, much attention has been given to the role played by endothelium-derived substances, such as nitric oxide (NO), prostacyclin (PGI2) and endothelium-derived hyperpolarizing factor, in exercise hyperaemia. The recognition that endothelial dysfunction is associated with cardiovascular disease and even accompanies healthy ageing has fuelled interest in how the contribution of the endothelium-dependent substances changes with disease and ageing. The aim of this review is to redress the balance a little, to consider the contribution made by substances that accumulate in the interstitial space of muscle during contraction and which may act independently of the endothelium. These substances include the adenine nucleotides, adenosine, prostaglandins (PGI2 and PGE2), plus several ions – inorganic phosphate (Pi), K+, H+ and lactate – which are generally considered to be released from the contracting muscle fibres rather than the endothelium. This is a reasonable assumption, because the endothelium limits, to a greater or lesser extent, the movement of these substances between interstitial fluid and plasma. A particular argument made in this review is that the accumulation and, consequently, the action of some of these substances is O2 dependent and can therefore be attenuated when additional O2 is provided during exercise. This proposal recognizes the well-known evidence that the magnitude of exercise hyperaemia is directly associated with work done by the contracting muscle and therefore with oxygen consumption (inline image; Clifford & Hellsten, 2004).

Given the focus of this review, we have devoted little attention to the role that red blood cells may play in local vascular regulation during exercise; this topic was reviewed in recent symposium reports published in The Journal of Physiology and elsewhere. To start with, there is extensive evidence that red blood cells release ATP when haemoglobin deoxygenates and that ATP acts on P2 receptors on the endothelium to induce dilatation (see Ellsworth & Sprague, 2012). Furthermore, deoxygenated haemoglobin reduces nitrite to NO, leading to maximal generation of NO when oxyhaemoglobin saturation is ∼50% (see Owusu et al. 2012 for review). On the contrary, S-nitrosohaemoglobin, which is formed in the lung when haemoglobin undergoes a conformational change to bind O2, is considered to regenerate NO via transnitrosation reactions with thiols when haemoglobin deoxygenates in tissue microcirculation (Singel & Stamler, 2005). These mechanisms are proposed to contribute to exercise hyperaemia in a manner that is dependent on the level of oxyhaemoglobin, ATP and NO being released in proportion to O2 unloading and therefore to muscle inline image and exercise intensity. Although the indirect evidence for their roles is accumulating, direct evidence has so far proved elusive because there are no selective antagonists of P2 receptors that can be used in vivo and it is difficult to differentiate the NO that escapes red blood cells from that released from endothelial cells (see, for example, Dufour et al. 2010; González-Alonso, 2012; Hellsten et al. 2012 for reviews).

The early phase of exercise hyperaemia

There seems little doubt that the early phase of exercise hyperaemia, recorded as an increase in arteriolar diameter and blood flow occurring within 1–2 s of the onset of contraction (Marshall & Tandon, 1984; Hamann et al. 2004), is dependent on K+. Recordings with K+-sensitive electrodes clearly showed that interstitial K+ concentration rises very steeply from contraction onset, reaching a peak within 5–10 s and achieving concentrations of up to 10 mm, depending on duration and force of contraction (Hník et al. 1976). More recent measurements with microdialysis confirmed that interstitial K+ concentration increases in a graded manner with exercise intensity (Juel et al. 2000; Lott et al. 2001) and to much higher levels than measured in venous efflux from muscle or than are required to induce similar dilatation when infused intra-arterially (Hník et al. 1976; Wilson et al. 1994; Juel et al. 2000). This disparity highlights one of the dangers inherent in attempting to estimate the contribution made by a substance to exercise hyperaemia from the degree of dilatation induced when it is infused at rest at levels chosen to mimic concentrations measured in venous efflux during exercise.

There is compelling evidence that the arteriolar dilatation recorded at the fourth second of contraction over a range of frequencies was attenuated when release of K+ through voltage-sensitive K+ channels on the muscle fibres was inhibited or when the dilator actions of K+ on arteriolar smooth muscle were inhibited by blocking the Na+–K+ pump or by inhibiting inwardly rectifying K+ channels; there was less consistent effect on the dilatation at the 20th second of contraction (Armstrong et al. 2007). Moreover, when muscle vasculature was perfused with high-K+ solution to depolarize the vascular smooth muscle, the rapid dilatation in response to 1 s contraction was abolished. This indicates that smooth muscle hyperpolarization induced by physiological levels of K+ is more important than the mechanical effect of the ‘muscle pump’ in the early, rapid phase of hyperaemia (Hamann et al. 2004). The factors that contribute to the early phase of exercise hyperaemia were reviewed by Clifford (2007).

The correlation between interstitial and venous concentrations of K+ and the magnitude of exercise hyperaemia leaves the possibility that K+ also contributes to the maintained phase (Hilton et al. 1978; Kiens et al. 1989; Lott et al. 2001). It may be the case that the hyperpolarization induced by K+ in vascular smooth muscle and endothelial cells facilitates the action of dilators that are released later. For example, adenosine and prostaglandins (PGs) may cause dilatation partly by opening one or more types of K+ channel, and dilatation is conducted from distal to upstream vessels by cell-to-cell conduction of hyperpolarization (see Duza & Sarelius, 2003; Jackson, 2005; Domeier & Segal, 2007; Ray & Marshall, 2006).

The maintained phase of exercise hyperaemia

The maintained phase lasts throughout and beyond exercise, continuing into what is often called postexercise or postcontraction hyperaemia. The magnitude of the maintained phase correlates well with various measures of the level of exercise undertaken, namely the work done, the force exerted or inline image. This relationship has led to the traditional view that exercise hyperaemia is induced by a substance or, more probably, substances that are released in proportion to exercise intensity. They are presumed to act in a feedback manner such that the muscle vasodilatation induced by these substances and the consequent increase in muscle blood flow provides for the metabolic requirements of contracting muscle. The implication is that the dilating substance(s) acts as an error signal, released when nutrient supply is inadequate in some way, such that the feedback mechanism is balanced when the magnitude of hyperaemia matches the metabolic requirements of the muscle. Clearly, factors that have no obvious relationship to muscle metabolism may also contribute to exercise hyperaemia. For example, endothelium-dependent dilator substances generated by increased shear stress (Koller & Kaley, 1990) could be released by the increase in blood flow per se and bring shear stress back into balance. Their involvement may well limit the contribution required by the byproducts of muscle metabolism to keep supply and demand matched.

Involvement of different factors triggered by the different types of stimuli arising in exercising muscle could explain the phenomenon of redundancy, which has received much attention in recent years (e.g. Clifford & Hellsten, 2004; Joyner & Wilkins et al. 2007). This is the idea that the reason it has proved difficult to identify the substances responsible for exercise hyperaemia and to establish their relative contributions is that when the release or action of one substance is prevented by use of a pharmacological antagonist, the magnitude of hyperaemia changes little because other factors simply make a larger contribution. It is worth noting, in relation to the discussion below on O2-dependent substances, that if, rather than pharmacologically blocking the contribution of a substance released as an ionic, metabolic or mechanical error signal, the conditions that created the error are ameliorated, then the same exercise intensity may be accomplished with a smaller exercise hyperaemia.

As indicated above, the maintained phase continues imperceptibly into postcontraction or postexercise hyperaemia. The factors that contribute to this phase were comprehensively reviewed by Bangsbo & Hellsten (1998). Not surprisingly, these factors include substances generated during exercise that are still acting after contraction has ceased. Their long-lasting effect probably reflects the fact that some substances are slow to disappear by metabolism, uptake or diffusion away from the site and the fact that, once dilated, vascular smooth muscle is rather slow to regain initial tone. Interestingly, the magnitude of postexercise hyperaemia is greater than is required to support the increase in muscle inline image at this time (Bangsbo & Hellsten, 1998). This may be because, especially after static contraction, cessation of the mechanical occluding effect of muscle contraction leads to a sudden increase in blood velocity and a shear stress-related release of dilator substances from the endothelium.

Involvement of interstitial ATP and its metabolites It is now widely accepted, from measurements made by microdialysis, that ATP, ADP, AMP and adenosine accumulate in muscle interstitium during exercise in humans and other mammals (Hellsten et al. 1998; Mo & Ballard, 2001; Lo et al. 2001). Each accumulates in proportion to the exercise intensity, but it is the increase in ATP that correlates best with the increase in muscle blood flow (Hellsten et al. 1998). When applied extraluminally, ATP predominantly causes constriction of muscle arterioles, but with an underlying vasodilator component attributable to the action of adenosine formed extracellularly by ectonuleotidases (McCullough et al. 1997; Duza & Sarelius, 2003). Thus, the general consensus is that it is interstitial adenosine that contributes to the maintained phase of exercise hyperaemia, rather than interstitial nucleotides.

Origin and contribution of adenosine Studies on cultured skeletal muscle cells showed that muscle contraction increased the concentrations of ADP and AMP in the culture medium; release of ATP was not detected. It was demonstrated that adenosine was generated extracellularly by ectophosphatases on the sarcolemma; the final step from AMP to adenosine is catalysed by 5′-nucleotidase, the mobilization of which to the cell membrane is increased by contraction (Hellsten & Frandsen, 1997; Hellsten, 1999). The possibility has been considered that the ATP that appears in the interstitial space in vivo is released from the sympathetic nerve fibres as a cotransmitter with noradrenaline, rather than from the skeletal muscle fibres. However, the quantities released from sympathetic terminals are small, localized to the neurovascular junction and unlikely to explain the high ATP concentrations measured in the interstitium. Furthermore, ATP released into plasma, for example, from red blood cells when haemoglobin unloads O2, does not easily cross the endothelial barrier (Mo & Ballard, 2001). An alternative explanation has been provided by Ballard and co-workers, who have built up substantial evidence, in vitro and in vivo, that AMP and adenosine are generated by contracting skeletal muscle fibres in proportion to the increase in muscle H+ concentration (Achike & Ballard, 1993; Cheng et al. 2000). They recently showed that ATP itself was released from skeletal muscle fibres when intracellular H+ was increased, via a channel associated with the cystic fibrosis transmembrane conductance regulator (Tu et al. 2010). Contracting skeletal muscle fibres have also been shown to release ATP through pannexin-1 hemichannels activated by membrane depolarization, intracellular Ca2+ or mechanical stress (Buvinic et al. 2009). This evidence suggests that the adenosine that contributes to exercise hyperaemia originates from ATP that is released by skeletal muscle fibres in proportion to muscle metabolism and exercise intensity.

Based on the pharmacological evidence of adenosine receptor antagonism, adenosine has been credited with 10–20% of the exercise hyperaemia evoked in forearm or leg muscles of humans by rhythmic exercise, but as much as 20–40% of the hyperaemia evoked in limb muscles of dogs, cats and rats by stimulation of the motor nerve (see Rådegren & Hellsten, 2000; Marshall, 2007). The finding that adenosine seems to make a smaller contribution in humans is probably explained by the fact that the only adenosine receptor antagonist approved for use in humans is theophylline/aminophylline, which, when given at the doses routinely used, attenuates muscle vasodilatation evoked by infused adenosine by only ∼50% (see Martin et al. 2006). Theophylline also inhibits phosphodiesterase activity and so may raise the level of cAMP in vascular smooth muscle and distort the balance of cAMP and cGMP that determines dilatation (see de Wit et al. 1994; Marshall, 2007). The effect of the adenosine receptor antagonist 8-phenyltheophylline, which has a higher affinity as an antagonist than theophylline and does not inhibit phosphodiesterase, indicated that adenosine contributes to the maintained phase of exercise hyperaemia from about the third minute of contraction, that this contribution persists for the remainder of contraction and that it continues through postcontraction hyperaemia (see Poucher, 1996; Ray & Marshall, 2009a).

Mechanism of action of adenosine Of the four subtypes of adenosine receptors, adenosine induces vasodilatation mainly by acting on the A1 and A2A receptors, which were originally differentiated on the basis that A1 receptor stimulation decreases and A2A receptor stimulation increases cAMP levels (Ralevic & Burnstock, 1998). The availability of selective antagonists allowed Poucher (1996) to show that the adenosine component of exercise hyperaemia induced in the cat by a 20 min period of contraction is mediated by A2A receptors rather than A1 or other subtypes; the attenuating effect of 8-phenyltheophylline was fully matched by the selective A2A receptor antagonist ZM241385. Likewise, the hyperaemia evoked by a 5 min period of twitch or tetanic contractions in the rat was attenuated by ZM241385, and subsequent administration of 8-sulphophenyltheophylline, an antagonist that is no selective between adenosine receptor subtypes had no effect (Ray & Marshall, 2009a; see Fig. 1A).

Figure 1.

Involvement of adenosine and NO in the vasodilator responses evoked by isometric twitch contractions in rat hindlimb muscle 
A–C shows femoral vascular conductance (FVC; –▪–) for the control response in the 1 min before (baseline), 5 min during (S1–5) and 7 min after sciatic nerve stimulation at 40 Hz (R1–7). In A,FVC is shown after ZM241385 (–▴–) and after ZM241385 + 8-sulphophenyltheophylline (8-SPT)(…♦…). §P < 0.05, control S1–5 and/or R1–7 vs. stimulation after ZM241385 and stimulation after ZM241385 + 8-SPT. There was no significant difference between stimulation after ZM241385 and stimulation after ZM241385 + 8-SPT at any time point. In B, FVC is shown after l-NAME (–○–) and after l-NAME + ZM241385 (…▴…). *P < 0.05 vs. control; §P < 0.05 vs.l-NAME. In C, FVC is shown after l-NAME (–○–), after l-NAME + SNAP to restore baseline FVC (…▵…) and after l-NAME + SNAP + ZM241385 (–▴–). *P < 0.05 vs. control; §P < 0.05 vs.l-NAME; and †P < 0.05 vs.l-NAME + SNAP. All values are shown as means ± SEM. In A, n = 10; in B and C, n = 12 rats. Modified from Ray & Marshall (2009a,b). In D, the effects on FVC of breathing air or 40% O2 without or with 8-SPT are compared. The FVC response to sciatic nerve stimulation (40 Hz) is shown when breathing air (–▪–) or 40% O2 (–○–), as well as after 8-SPT when breathing air (–♦–) and after 8-SPT during 40%O2 (…♦…). *P < 0.05 vs. control. All values are shown as means ± SEM; n = 10 rats. (CJ Ray, L Hargreaves, AM Coney, JM Marshall, unpublished observations).

We had previously shown that adenosine contributes to muscle vasodilatation induced by systemic hypoxia (Ray et al. 2002). Our evidence indicated that adenosine is released from endothelial cells as a consequence of the competitive interaction between NO and O2 on cytochrome oxidase (cyta3), such that even tonic levels of NO increase the sensitivity of endothelial cyta3 to falls in inline image, decreasing ATP synthesis and releasing adenosine (Clementi et al. 1999; Edmunds et al. 2003). In systemic hypoxia, adenosine stimulates endothelial A1 receptors to activate cyclo-oxygenase (COX), generating PGI2 as an intermediate, which acts back on the endothelial cells to generate cAMP. This presumably overcomes the decrease in cAMP that A1 receptor stimulation would otherwise produce, allowing activation of NO synthase (NOS) by phosphorylation and synthesis of NO, which produces dilatation (Ray et al. 2002; Ray & Marshall, 2005, 2006). We were, therefore, particularly interested to establish whether the adenosine component of exercise hyperaemia depends on NO.

In fact, NOS inhibition attenuated exercise hyperaemia in the rat, but had no effect on hindlimb inline image, while the A2A contribution of adenosine persisted and showed no sign of being blunted (Fig. 1B). Furthermore, when an NO donor or cell-permeant cGMP was infused after NOS inhibition so as to restore the tonic dilator effect of NO and cGMP levels in the vascular smooth muscle, exercise hyperaemia was fully restored and A2A receptor blockade had the same effect as before NOS inhibition (Ray & Marshall, 2009b; Fig. 1C). Thus, release of adenosine in exercise cannot be attributed to the interaction of NO and O2 at cyta3 in endothelial cells or muscle fibres, consistent with the evidence discussed above that adenosine is generated extracellularly from ATP. Furthermore, in the rat, NO does not mediate the A2A component of exercise hyperaemia (Ray & Marshall, 2009b). This is consistent with evidence that NOS inhibition in humans has little effect on the maintained phase of exercise hyperaemia or inline image (Frandsen et al. 2001; Schrage et al. 2004; Mortensen et al. 2007).

In contrast, it has been argued that adenosine acts on extraluminal receptors in exercise, to induce dilatation via interdependent actions of PGI2 and NO (Mortensen et al. 2009). This is based on evidence from studies on young humans that single inhibition of NOS or COX had no effect on hyperaemia evoked by knee-extensor exercise, but that dual blockade of NOS and COX, or single blockade of adenosine receptors with theophylline, or all three antagonists given together had similar attenuating effects (Mortensen et al. 2007, 2009). More recently, it was proposed that adenosine acts on extraluminal receptors on capillary endothelial cells to generate PGI2 and NO and that dilatation then spreads to arterioles by conduction through endothelial cells (Nyberg et al. 2010). In support of this, Nyberg et al. (2010) showed that infusion of adenosine into the interstitium via microdialysis probes generated both PGI2 and NO, while in vitro, on cells isolated from rats, adenosine released PGI2 and NO from microvascular endothelial cells, but released only NO from skeletal muscle myotubes. They also cited evidence that localized contraction of muscle fibres underlying capillaries evoked dilatation of upstream arterioles (Berg et al. 1997) and that extraluminal application of adenosine to arterioles evoked dilatation that was conducted upstream via a rise in endothelial Ca2+ (Duza & Sarelius, 2003). Furthermore, it was shown by using ACh that a component of conducted dilatation was inhibited by dual NOS and COX blockade (Domeier & Segal, 2007).

Exactly how these two sets of apparently contrasting findings can be reconciled is not clear. If it is assumed, until proved otherwise, that the mechanisms are essentially the same in rats and humans then it is likely that the adenosine receptors that generated PGI2 in response to interstitial infusion of adenosine in humans (Nyberg et al. 2010) included A1 receptors on capillary endothelium, because A1 receptor stimulation generated NO and PGI2, whereas A2A receptor stimulation generated only NO (Ray et al. 2002). The contribution made by adenosine to conducted dilatation and thereby to the increase in gross blood flow that is the usual record of exercise hyperaemia is not known. However, this contribution seems unlikely to involve PGI2 acting as an intermediate for NO, which is the mechanism that operates in systemic hypoxia (Ray et al. 2002), because NOS inhibition did not attenuate the adenosine-mediated component of exercise hyperaemia in the rat and had little or no effect on exercise hyperaemia in humans (see above). It may be that the apparent interdependence of the three mediators arises in exercise because dual NOS and COX blockade decreases the levels of cGMP and cAMP in vascular smooth muscle so much that dilatation induced by adenosine acting via A2A and cAMP is severely blunted (see de Wit et al. 1994). It could also be that PGs contribute independently of adenosine to exercise hyperaemia (see below), but the dilator action of PGs depends on a tonic level of NO. These possibilities could be explored in rats and humans by testing the effect on exercise hyperaemia of adenosine receptor blockade before or after blockade of COX, and the effect of COX inhibition after NOS inhibition with and without restoration of a tonic level of NO. Such experiments would also help to reveal whether there are species differences between rats and humans in the functioning of these pathways.

Is exercise hyperaemia O2 dependent?

It was established many years ago, by using O2-sensitive microelectrodes, that the inline image immediately external to arteriole walls decreases gradually from proximal to terminal arterioles because O2 diffuses along the length of arterioles as well as from capillaries (Duling & Berne, 1970). This finding has still not made its way into the majority of textbooks. The same technique showed that during muscle contraction, periarteriolar inline image hardly changed, but inline image measured at the mid-region of the capillary bed fell for 1–2 min from the onset of contraction and then recovered. In contrast, tissue inline image estimated at the venous end of capillaries fell substantially to levels that depended on contraction frequency and O2 demand, from ∼14–25 or 28 mmHg at rest in the cremaster and spinotrapezius muscle, respectively, to ∼5 mmHg during submaximal contraction (Gorzynski & Duling, 1978; Lash & Bohlen, 1987). Perivenular inline image showed a more maintained fall from ∼30 mmHg at rest to ∼15 mmHg during contraction (Lash & Bohlen, 1987). These results correspond reasonably well with estimates made in humans, from myoglobin O2 saturation measured with proton nuclear magnetic resonance spectroscopy, which suggest that muscle intracellular inline image falls from ∼34 mmHg at rest to ∼3 mmHg at exercise intensities ranging from 50% maximum to maximum (see Richardson et al. 2006).

Taken together, these findings indicate that exercise hyperaemia is extremely unlikely to reflect a direct effect of a fall in inline image on arteriolar smooth muscle and, likewise, is unlikely to be caused by substances released from arteriolar endothelium in response to a fall in inline image. However, they do allow the possibility that exercise hyperaemia is due, at least in part, to substances generated in an O2-dependent manner by skeletal muscle fibres that diffuse to arterioles, or are generated by capillaries or venules and produce a response that is conducted to arterioles. They would also allow the possibility that substances are released by red blood cells when haemoglobin unloads O2.

In agreement with these ideas, when the superfusate inline image for cremaster was raised so as to prevent the fall in tissue inline image during contraction, there was arteriolar constriction at rest, but the size of the arteriolar dilatation evoked by muscle contraction also decreased. Indeed, it was estimated that between 38 and 55% of the arteriolar dilatation could be explained by the fall in tissue inline image and would serve to keep inline image constant (Gorzynski & Duling, 1978). However, subsequent studies showed that tonic vasoconstriction induced by topically applied noradrenaline also decreased the arteriolar dilator response to muscle contraction. It was therefore suggested that tissue inline image is not precisely regulated and that the feedback between tissue inline image and the arterioles simply serves to buffer the fall in tissue inline image (Klitzman et al. 1982). Experiments in which attempts were made to ameliorate the fall in tissue inline image by breathing O2 at high concentration have also produced equivocal results. Thus, breathing 100% O2 caused tonic vasoconstriction in limb muscles of healthy human subjects, and either attenuated hyperaemia evoked by dynamic one-leg exercise by ∼10% (Welch et al. 1977; Gonzalez-Alonso et al. 2002) or had no effect in whole-body, cycling exercise (Knight et al. 1993).

In contrast, more modest hyperoxia, breathing 40% O2, had no effect on baseline blood flow and attenuated by ∼35% the postcontraction hyperaemia evoked by hand grip at 60% maximal voluntary contraction (MVC; Win & Marshall, 2005). This was consistent with the earlier finding that breathing 60% O2 had no effect on resting blood flow and attenuated hyperaemia that occurred during dynamic leg exercise, by ∼12% (Pedersen et al. 1999). Furthermore, when 40% O2 was breathed only during hand grip at 100% maximal voluntary effort, postcontraction hyperaemia was substantially attenuated, whereas when it was breathed only during the recovery period, there was no effect on the hyperaemia (Fig. 2). Given that breathing 40% O2 during hand grip also decreased venous efflux of lactate and H+ ions, it can be concluded that sufficient additional O2 reached the contracting muscle fibres to affect muscle metabolism, even during strong static contraction that limits muscle blood flow (Fordy & Marshall, 2012). Thus, taken together, these results indicate that O2-dependent factor(s) generated during static contraction by the fall in tissue inline image contribute significantly to postcontraction hyperaemia. The results of Pedersen et al. (1999) allow a similar conclusion to be drawn for the maintained phase of hyperaemia associated with dynamic exercise. In other words, O2-dependent factor(s) released from muscle fibres apparently represent one or more of the error signals discussed above that are released when O2 supply is inadequate and which cause dilatation of arterioles, so helping to increase O2 delivery. When arterial inline image is raised by breathing 40–60% O2, the diffusion of O2 to muscle fibres increases sufficiently to attenuate release of these O2-dependent factors.

Figure 2.

Effects of breathing 40% O2 during contraction (A and C) or during recovery (B) on responses evoked by isometric forearm contraction at 100% maximal voluntary effort (MVE) to exhaustion 
A, effects of breathing 40% O2 during contraction on forearm vascular conductance (FVC). B, FVC when 40% O2 was breathed during recovery from contraction, recorded at 1 min intervals at rest and after contraction and, in addition, immediately (0) and at 15 s after contraction. Continuous and dashed lines join values recorded when air was breathed throughout and when 40% O2 was breathed, respectively, for periods indicated by the bars below. C, effects of breathing 40% O2 during contraction on venous lactate concentration and pH. Values were recorded when air was breathed throughout (open columns) and when 40% O2 was breathed only during contraction (filled columns). All values are shown as means ± SEM. *,† Difference from baseline in the air and 40% O2 conditions, respectively (P < 0.05). § Difference between values recorded when air and 40% O2 were breathed (P < 0.05). Modified from Fordy & Marshall (2012).

The nature of the O2-dependent substances  The adenosine hypothesis proposed by Berne in the 1960s for the regulation of coronary blood flow envisaged that when the delivery of O2 is not sufficient to match O2 demand, ATP hydrolysis outweighs ATP synthesis, leading to generation and release of adenosine from cardiac myocytes, which by its dilator actions restores O2 delivery until it matches O2 demand (Berne, 1963). Clearly, from the discussion above, recent studies on skeletal muscle indicate that ATP itself is released from contracting muscle and that breakdown of ATP to adenosine occurs extracellularly. However, there is evidence that the activity of 5′-nucleotidease is increased by a fall in inline image (Hunsucker et al. 2005) or a fall in pH (Cheng et al. 2000), while the release of ATP that has been associated with cystic fibrosis transmembrane conductance regulator channels has been attributed to a fall in intracellular pH (Tu et al. 2010). Thus, ameliorating the fall in tissue inline image during muscle contraction might be expected to decrease interstitial adenosine accumulation. Accordingly, we recently found that when rats were given 40% O2 instead of air to breathe, exercise hyperaemia was attenuated to a similar extent as that achieved with the adenosine receptor antagonist 8-SPT. Moreover, when 40% O2 was given in the presence of 8-SPT there was no further attenuation of the hyperaemia (CJ Ray, L Hargeaves, AM Coney & JM Marshall, unpublished observations; see Fig. 1D).

Other factors released from skeletal muscle fibres that might be O2-dependent dilators include Pi, H+ and lactate. Levels of intracellular Pi increase in skeletal muscle fibres during contraction as a result of ATP hydrolysis; the rise in ADP is limited by rephosphorylation by phosphocreatine, but Pi continues to increase unless mitochondrial ATP synthesis matches ATP hydrolysis. Raised intracellular Pi leads to Pi efflux, and its accumulation in interstitial fluid is exacerbated because Pi uptake via Na+–Pi transporters is inhibited during contraction (Abraham & Terjung, 2004). Microdialysis showed that interstitial Pi concentrations increased in cat muscles during contractions and in human leg and forearm muscles during static or rhythmic contractions (Bouschel et al. 1998; MacLean et al. 1998, 2000; Lott et al. 2001). Moreover, in cats, venous efflux of Pi correlated well with the magnitude of exercise hyperaemia, while in rats, topical Pi induced substantial arteriolar dilatation within 5 s (Hilton et al. 1978). Whether Pi makes an O2-dependent contribution to exercise hyperaemia remains to be tested.

Although lactate and H+ ions also leave skeletal muscle fibres during contraction, the time course and magnitude of their accumulation in interstitial fluid did not correlate well with the changes in muscle blood flow (see MacLean et al. 1998, 2000; Lott et al. 2001), consistent with the general view that they make little contribution to exercise hyperaemia (see Clifford & Hellsten, 2004). Thus, even though their release would be expected to be O2 dependent, as our results indicate (see above; Fordy & Marshall, 2012), neither is likely to make an important O2-dependent contribution to exercise hyperaemia.

In contrast, there is evidence that PGs make an O2-dependent contribution to postcontraction hyperaemia. Thus, COX inhibition with oral aspirin had a very similar attenuating effect on postcontraction hyperaemia evoked by static hand grip at 60% MVC to that of breathing 40% O2, and the combination of modest hyperoxia and COX inhibition caused no further attenuation (Win & Marshall, 2005).

These findings are consistent with previous reports that COX inhibition attenuated exercise hyperaemia evoked in forearm muscle during and after static hand grip at 15% MVC, postcontraction hyperaemia following rhythmic forearm exercise at maximal force or treadmill exercise in calf muscles, and hyperemia evoked during rhythmic forearm exercise at 10% MVC (Kilbom & Wennmalm, 1976; Cowley et al. 1985; Duffy et al. 1999; Schrage et al. 2004). Furthermore, graded rhythmic exercise of knee extensors produced graded increases in interstitial concentrations of PGI2 and PGE2 (Karamouzis et al. 2001), while rhythmic forearm and leg exercise increased venous efflux of PGI2 and PGE2 (Nowak & Wennmalm, 1978; Wilson & Kapoor, 1993).

Considering the source of PGs, skeletal muscle fibres of rodents and humans express COX-1 and COX-2 and generate PGE2 in the presence of the COX substrate, arachidonic acid (Testa et al. 2007). In experiments on cats, intramuscular generation of PGE2 was increased by even 30 s of static contraction, and the increase was augmented when contraction was evoked during arterial occlusion (Symons et al. 1991). On the other hand, adenosine did not release PGI2 from skeletal myotubes in vitro, although interstitial infusion of adenosine did release PGI2 into the interstitium, presumably from capillary endothelial cells (Nyberg et al. 2010). Furthermore, endothelial synthesis of PGI2 is regulated by O2 tension (Messina et al. 1992, 1994; Michiels et al. 1993; Frisbee et al. 2002). As discussed above, it is unlikely that arteriolar inline image falls sufficiently during exercise to release PGI2 from arteriolar endothelium, but the fall in capillary inline image could facilitate release of PGI2 into the interstitium. In addition, the fall in venular inline image may release PGI2 that diffuses to adjacent arterioles to dilate them (see Hammer et al. 2001). Thus, it is reasonable to propose that PGE2 and PGI2 are released into the interstitium from skeletal muscle fibres and from capillaries and/or venules, respectively, in response to a fall in tissue inline image during exercise. They dilate arterioles from the extraluminal surface (Messina & Kaley, 1980) and, possibly, by a Ca2+ signal conducted from capillary endothelium to upstream arterioles that generates a wave of PGI2 and NO release (Domeier & Segal, 2007).

Given that the combination of COX inhibition with breathing 40% O2 had no greater attenuating effect on postcontraction hyperaemia than hyperoxia alone (Win & Marshall, 2005), it seems that modest hyperoxia abolishes the contribution of PGI2 and PGE2. In other words, at 60% MVC, the contribution of PGs to postcontraction hyperaemia is totally O2 dependent. Clearly, we do not know whether supplying O2 at concentrations >40% would have produced further attenuation of the hyperaemia beyond that induced by COX inhibition. Thus, our results do not exclude the possibility that O2-dependent factors in addition to PGs contribute to exercise hyperaemia. Notably, we cannot deduce whether the O2-dependent roles of adenosine and PGs are dependent on one another, or distinct.

Drawing together the findings discussed above, it seems likely that accumulations of adenosine and PGE2 in the interstitium, generated in an O2-dependent manner by skeletal muscle fibres, contribute in a rather generalized way to exercise hyperaemia by inducing dilatation throughout the arteriolar tree, thereby increasing gross vascular conductance and muscle blood flow. However, release of PGI2 from capillaries and venules in response to a local fall in inline image has the potential, by propogation up the arteriolar tree and diffusion to nearby arterioles, to produce localized dilatation of arterioles supplying specific capillary networks that is graded with the local fall in inline image. Nevertheless, when the concentration of PGI2 in the interstitium builds up as a consequence of a more generalized fall in inline image, then PGI2 would also be expected to contribute to the increase in gross vascular conductance. Likewise, ATP released from red blood cells in proportion to the unloading of O2 from haemoglobin, but acting on the intraluminal surface of endothelial cells, would be expected to induce localized dilatation that matches capillary flow to the local fall in tissue inline image, as reviewed by Ellsworth & Sprague (2012). Moreover, the fact that PGI2 released from endothelium facilitates ATP release from red blood cells (Ellsworth & Sprague, 2012) suggests that interaction between the two mediators may ultimately contribute to the increase in gross conductance. In addition, NO that escapes from red blood cells when haemoglobin unloads O2 may have the potential to exert local or more generalized dilator effects (see Owuso et al. 2012).

Vasoconstrictor influences of O2 in exercise

Although studies in which 40% O2 (Win & Marshall, 2005; Fordy & Marshall, 2012) or 60%O2 was breathed (Pedersen et al. 1999) showed no effect on resting blood flow in forearm or leg muscles or systemic haemodynamics, in several studies on humans, breathing 100% O2 or breathing 100% O2 at 2 Atmospheres. Hyperoxia reduces exercising forearm evoked tonic vasoconstriction in limb muscle and a rise in arterial pressure (Bird & Telfer, 1966; Mak et al. 2002). Thus, it seems that O2 induces overt vasoconstriction when the inspired O2 concentration is >60%. Furthermore, on the basis of previous studies, the effect of 100% O2 on exercise hyperaemia is equivocal, with some studies showing a small attenuation of 10% (see Welch et al. 1977; González-Alonso et al. 2002) while others reported no effect (Knight et al. 1993). Indeed, it has been argued that a change in dissolved O2 in plasma, which is the main effect of breathing high O2 concentrations in a healthy individual with normal haemoglobin, plays a minor regulatory role in exercise hyperaemia compared with O2 bound to haemoglobin (Roach et al. 1999; Calbet, 2000; González-Alonso et al. 2001, 2002). In other words, arterial inline image is of minor importance relative to arterial O2 content.

Certainly, studies involving dynamic one-leg exercise in which arterial O2 content and/or arterial inline image were decreased by inducing anaemia, breathing carbon monoxide or a hypoxic mixture (10% O2), alone or in combination, conclusively showed good inverse relationships between the magnitude of exercise hyperaemia and arterial O2 content, but no obvious relationship with arterial inline image when it was changed from ∼40 to ∼540 mmHg (see Roach et al. 1999; Calbet 2000; Gonzalez-Alonso et al. 2001). We would not want to argue against the importance of arterial O2 content, but the question arises as to why these studies did not reveal the effect of arterial inline image that we have observed with more modest hyperoxia (40% O2).

The answer may lie in part with the vasoconstrictor effect of O2 at high concentration, which has been attributed to generation of reactive oxygen species (Rubanyi & Vanhoutte, 1986). Thus, in healthy young subjects, breathing 100% O2 not only increased baseline forearm vascular resistance, but also attenuated dose-dependent forearm vasodilatation induced by intra-arterial infusion of the endothelium-dependent dilator ACh (Mak et al. 2002) and attenuated cutaneous vasodilatation evoked by iontophoresis of ACh (Rousseau et al. 2010). These effects of hyperoxia were reversed by vitamin C, which scavenges intracellular and, potentially, extracellular, superoxide (O2), which inactivates NO by forming ONOO (see May, 2000). The effect of breathing 100% O2 on ACh-induced cutaneous dilatation was also reversed by COX inhibition, which led to the proposal that reactive oxygen species generated by hyperoxia inhibited endothelial synthesis of PGI2 (Rousseau et al. 2010), a proposal compatible with earlier evidence that hyperoxia decreased endothelial synthesis of dilator PGs in skeletal muscle arterioles (Messina et al. 1994).

Thus, it is reasonable to suggest that breathing 100% O2 generates O2 and blunts the dilator influences of NO and PGI2, not only in skeletal muscle, but also in other body tissues, which would explain the pressor effect generally reported with 100% O2. This systemic haemodynamic effect may be the underlying explanation for previous reports that breathing 100% O2 decreases cardiac output during exercise or changes the distribution of cardiac output (Ekblom et al. 1975; Pedersen et al. 1999). Thus, even though the O2 generated by breathing 100% O2 would be expected to produce a similar attenuating effect on exercise hyperaemia to combined COX and NOS blockade (see Mortensen et al. 2009), this may be difficult to demonstrate against a background of systemic haemodynamic change. Clearly, it cannot be assumed that breathing 40 or 60% O2 does not generate O2, but it seems likely that with more modest hyperoxia, the effect of O2 on systemic haemodynamics is minor and overwhelmed by the influence of additional O2 diffusing to the contracting muscle fibres.

Our recent findings agree with this proposal. Vitamin C given orally to healthy young male subjects, at a comparable dose to that used by Rousseau et al. (2010), augmented postcontraction hyperaemia evoked by hand grip at 60% MVC, but breathing 40% O2 still attenuated the hyperaemia by a similar extent to that observed in control conditions (H Caruanas & JM Marshall, unpublished observations). In view of the findings discussed above, it seems that even in healthy individuals sufficient reactive oxygen species are generated during submaximal contraction to limit postcontraction hyperaemia and that the antioxidant, vitamin C, decreases their effect on NO and PGI2 availability, thereby accentuating the hyperaemia. Breathing 40% O2 in the presence of vitamin C must therefore attenuate the hyperaemia by a mechanism that is independent of reactive oxygen species, presumably by limiting the O2-dependent generation of PGs, adenosine etc., as discussed above. To put this another way, if 40% O2 attenuated exercise hyperaemia by generating reactive oxygen species and limiting NO and PGI2 availability, then it would be expected to have little effect in the presence of vitamin C.

Interestingly, there have been sporadic reports over many years that breathing 100% O2 is less beneficial during exercise than more modest concentrations; the deleterious effects were attributed to its vasoconstrictor influence and consequent effects on cardiac output (see Bannister & Cunningham, 1954; Ekblom et al. 1975; Waring et al. 2003). The results discussed above indicate that breathing 40–60% O2 may be just enough to promote the beneficial effects and limit the deleterious effects.


The evidence discussed in this review indicates that substances released into the interstitium and acting directly on arteriolar smooth muscle, rather than via the endothelium, make a major contribution to exercise hyperaemia (Fig. 3). The initial phase of hyperaemia is attributable to dilatation caused by K+ released from the muscle fibres, but hyperpolarization of vascular smooth muscle and endothelium induced by K+ may facilitate the vasodilator action of substances that contribute to the maintained phase directly, as well as by conduction of signals upstream. Both ATP and NO released from red blood cells into plasma when haemoglobin offloads O2 may contribute to the dilatation by acting from the intraluminal surface. However, we argue that dilator substances are also released by a decreased inline image in the vicinity of muscle fibres, capillaries and venules, but not arterioles. Notably, ATP is released from skeletal muscle fibres via channels associated with cystic fibrosis transmembrane conductance regulator, in response to a fall in intracellular pH. Adenosine is generated extracellularly by nucleotidases, facilitated by the reductions in extracellular pH and in inline image, and acts via extraluminal A2A receptors to produce vasodilatation in an NO-independent manner. Furthermore, PGE2, released mainly from muscle fibres, and PGI2, released mainly from capillaries and venular endothelium, act via extraluminal receptors on arterioles to cause dilatation. We propose that modest hyperoxia (breathing 40–60% O2) attenuates the contribution of these O2-dependent substances to exercise hyperaemia by alleviating the fall in tissue inline image. By contrast, breathing 100% O2 generates sufficient reactive oxygen species to cause vasoconstriction by limiting endothelium-dependent dilatation, offsetting the beneficial effects otherwise produced by additional O2.

Figure 3.

Schematic diagram showing contribution to exercise hyperaemia of substances released into interstitial fluid during contraction 
Shear stress, acting on endothelium, causes tonic activation of endothelial nitric oxide synthase (NOS), generating NO, which produces tonic dilatation of arterioles, upon which other influences are superimposed. Nitric oxide also competes with O2 for the same binding site on cytochrome oxidase (cyta3) in mitochondria, thereby regulating endothelial ATP synthesis. During exercise, K+ released from skeletal muscle fibres during their action potentials dilates arterioles by causing hyperpolarization of vascular smooth muscle. Even at rest, O2 diffuses outwards along length of arterioles, but during exercise the periarteriolar inline image stays virtually constant because arterial inline image is well maintained. In contracting muscle fibres, capillaries and venules, however, inline image falls due to increase in muscle oxygen consumption (inline image). The fall in muscle inline image and pH leads to release of ATP from muscle fibres through regulated channels; ATP is metabolized extracellularly to adenosine, which dilates arterioles via adenosine A2A receptors. The fall in muscle inline image and capillary and venular inline image also leads to release of the prostaglandins PGE2 and PGI2 synthesized by cyclo-oxygenase (COX), which act on extraluminal EP and IP receptors for PGE2 and PGI2, respectively, to dilate arterioles. The release of ATP from red blood cells caused by haemoglobin unloading O2 is also shown; ATP can act locally on P2 receptors to cause dilatation. Breathing 40% O2 during exercise limits the fall in tissue inline image, which attenuates the generation of adenosine and prostaglandins, thereby attenuating exercise hyperaemia. The A1 receptors on arterioles and A1 and A2A receptors on endothelium make little direct contribution to exercise hyperaemia.