Arteriolar oxygen reactivity: where is the sensor and what is the mechanism of action?

Abstract

Arterioles in the peripheral microcirculation are exquisitely sensitive to changes in $P_{{\mathrm{O}}_2}$ in their environment: increases in $P_{{\mathrm{O}}_2}$ cause vasoconstriction while decreases in $P_{{\mathrm{O}}_2}$ result in vasodilatation. However, the cell type that senses O₂ (the O₂ sensor) and the signalling pathway that couples changes in $P_{{\mathrm{O}}_2}$ to changes in arteriolar tone (the mechanism of action) remain unclear. Many (but not all) ex vivo studies of isolated cannulated resistance arteries and large, first‐order arterioles support the hypothesis that these vessels are intrinsically sensitive to $P_{{\mathrm{O}}_2}$ with the smooth muscle, endothelial cells, or red blood cells serving as the O₂ sensor. However, in situ studies testing these hypotheses in downstream arterioles have failed to find evidence of intrinsic O₂ sensitivity, and instead have supported the idea that extravascular cells sense O₂. Similarly, ex vivo studies of isolated, cannulated resistance arteries and large first‐order arterioles support the hypotheses that O₂‐dependent inhibition of production of vasodilator cyclooxygenase products or O₂‐dependent destruction of nitric oxide mediates O₂ reactivity of these upstream vessels. In contrast, most in vivo studies of downstream arterioles have disproved these hypotheses and instead have provided evidence supporting the idea that O₂‐dependent production of vasoconstrictors mediates arteriolar O₂ reactivity, with significant regional heterogeneity in the specific vasoconstrictor involved. Oxygen‐induced vasoconstriction may serve as a protective mechanism to reduce the oxidative burden to which a tissue is exposed, a process that is superimposed on top of the local mechanisms which regulate tissue blood flow to meet a tissue's metabolic demand.

Abbreviations

BK_Ca channels, large‐conductance

Ca²⁺‐activated K⁺ channels

Ca_L channels

L‐type Ca²⁺ channels

Cl_Ca channels

Ca²⁺‐activated Cl⁻ channels

CYP450

cytochrome P450

CysLTs

cysteinyl leukotrienes

CysLTRs

CysLT receptors

DDMS

N‐methylsulfonyl‐12,12‐dibromododec‐11‐enamide

DIDS

4,4′‐diisothiocyano‐2,2′‐stilbenedisulfonic acid

ETYA

eicosatetraynoic acid

FLAP

5‐lipoxygenase‐activating protein

20‐HETE

20‐hydroxyeicosatetraenoic acid

5‐LO

5‐lipoxygenase

LTA₄

leukotriene A₄

LTC₄

leukotriene C₄

LTD₄

leukotriene D₄

LTE₄

leukotriene E₄

17‐ODYA

17‐octadecynoic acid

$P_{{\mathrm{O}}_2}$

partial pressure of oxygen

PSS

physiological salt solution

Introduction

Oxygen has been implicated in the local control of blood flow for more than a century (Sparks, 1980; Renkin, 1984; Kontos & Wei, 1985; Golub & Pittman, 2013). In tissues such as skeletal muscle, heart, brain, kidneys and the gut, blood flow (O₂ supply) is directly proportional to tissue oxygen consumption (O₂ demand) (Sparks, 1980; Renkin, 1984; Kontos & Wei, 1985; Golub & Pittman, 2013). Oxygen as one link between O₂ demand and O₂ supply provides a compact control system that would help ensure that O₂ supply meets the tissues’ demand for O₂. For O₂ to participate in the local regulation of blood flow, arterioles in the microcirculation must be able to sense changes in the $P_{{\mathrm{O}}_2}$ of the surrounding tissue, and then respond appropriately to either constrict, if O₂ supply exceeds O₂ demand and tissue $P_{{\mathrm{O}}_2}$ increases, or dilate, if O₂ supply is less than O₂ demand and tissue $P_{{\mathrm{O}}_2}$ decreases.

Consistent with a possible role for O₂ in the local regulation of blood flow, there is consensus that O₂ is vasoactive. Arterioles in the peripheral microcirculation constrict when exposed to increases in $P_{{\mathrm{O}}_2}$ produced either by increases in inspired $P_{{\mathrm{O}}_2}$ (Hutchins et al. 1974; Zhu et al. 1998; Demchenko et al. 2000; Vucetic et al. 2004; Sakai et al. 2007; Kisilevsky et al. 2008; Justesen et al. 2010) or increases in the $P_{{\mathrm{O}}_2}$ of solutions flowing over microvascular beds (Duling, 1972; Prewitt & Johnson, 1976; Lombard & Duling, 1977, 1981; Tuma et al. 1977; Kontos et al. 1978; Lindbom et al. 1980; Proctor et al. 1981; Sullivan & Johnson, 1981 a,b; Proctor & Duling, 1982 a,b; Jackson & Duling, 1983; Jackson, 1986, 1987, 1988, 1989, 1993; Lombard et al. 1986, 1999, 2004; Kaul et al. 1995; Welsh et al. 1998; Hungerford et al. 2000; Kunert et al. 2001 a,b, 2009; Frisbee & Lombard, 2002; Drenjancevic‐Peric et al. 2003, 2004; Wang et al. 2009; Ngo et al. 2010, 2013). Conversely, decreases in $P_{{\mathrm{O}}_2}$, produced by lowering inspired $P_{{\mathrm{O}}_2}$, cause vasodilatation (see Marshall, 2000 for references). What remains unclear are the cellular location of the process that detects changes in $P_{{\mathrm{O}}_2}$ (the O₂ sensor) and the mechanism that transduces the changes in $P_{{\mathrm{O}}_2}$ to changes in arteriolar tone (the mechanism of O₂ action). This review will focus on the site and mechanism of action of O₂ on arterioles in peripheral tissues at rest. Changes in arteriolar tone induced by systemic hypoxia or hyperoxia produced by changes in inspired $P_{{\mathrm{O}}_2}$ will not be considered because the mechanisms involved may be different. Systemic hypoxia and hyperoxia are complicated by changes in neural outflow to arterioles (Marshall, 1994; Guyenet, 2000), changes in circulating hormones and vasoactive substances (Mazzeo et al. 1991), and changes in circulating cytokines released from macrophages in the lungs (Shah et al. 2003). Oxygen reactivity of microvessels in active tissue (contracting striated muscle, the heart, etc.) will also not be considered as there are probably interactions among mechanisms that have not been fully explored, and to maintain the focus on arteriolar O₂ reactivity per se. Evidence will be presented that: (1) O₂ acts as a vasoconstrictor on arterioles in the peripheral microcirculation, (2) there is regional heterogeneity in the specific vasoconstrictor pathway that mediates arteriolar O₂ reactivity, and (3) arterioles may be tuned to detect changes in tissue $P_{{\mathrm{O}}_2}$, whereas upstream feed arteries may be tuned to respond to changes in blood $P_{{\mathrm{O}}_2}$.

Setting the stage

Microvascular architecture and the study of arteriolar O₂ reactivity

The microcirculation is the business end of the cardiovascular system. It is here that exchange of O₂, CO₂, substrates, hormones, waste products, etc. occurs. This exchange function depends on a continuous, regulated flow of blood through the microvasculature. The microcirculation originates when a feed artery (100–200 μm internal diameter) that is external to the tissue branches into a first‐order arteriole (50–100 μm). Feed arteries provide up to half of the precapillary vascular resistance to blood flow in tissues such as skeletal muscle and importantly contribute to the regulation of total blood flow to the tissue (Segal & Duling, 1986). Arterioles are embedded in the parenchyma that they perfuse and have 1–2 layers of smooth muscle in the media of their walls. Arterioles contribute the remaining 50% of the precapillary vascular resistance to blood flow and determine the distribution of blood flow within the tissue. However, the primary effectors of local blood flow control to and within a tissue are the arterioles. Upstream feed arteries are external to the tissue, and are recruited to the process of blood flow regulation by indirect mechanisms such as conducted vasomotion (Segal & Duling, 1986) or flow‐dependent responses (Davis et al. 2011) that are initiated by changes in tone of the downstream arterioles. First‐order arterioles usually have ∼two layers of smooth muscle, while smaller arterioles are invested with only a single layer of smooth muscle. In striated muscle, for example, second‐ to fifth‐order arterioles divergently branch from the first‐order arterioles, ending in terminal arterioles that supply 10–20 capillaries each. Post‐capillary venules collect blood flow from two or more capillaries. These then drain into higher order venules, which converge on veins draining a tissue or organ. While capillaries are a major site of exchange of materials and heat between tissues and blood because of their large surface area, O₂ and CO₂ also readily diffuse across the wall of arterioles and venules contributing to the exchange of these respiratory gases in the microcirculation (Renkin, 1984).

Methods and approaches used to study arteriolar O₂ reactivity

Two general approaches have been used to explore microvascular O₂ reactivity: (1) pressure myography (Duling et al. 1981; Schjorring et al. 2015) and other techniques for the ex vivo study of isolated vessels, and (2) intravital microscopy of arterioles, in situ, in exteriorized, autoperfused tissues (Baez, 1973; Duling, 1973). Pressure myography involves the surgical removal of a vessel from a tissue. The vessels are then cannulated onto micropipettes and pressurized to a physiological pressure. The vessels are superfused with a physiological salt solution (PSS; to allow treatment with drugs, altering the $P_{{\mathrm{O}}_2}$, and temperature control), and imaged with a microscope to measure internal diameter. Intravital microscopy usually involves the exteriorization of a tissue from an anaesthetized animal retaining the tissue's blood and nerve supply. The tissue is superfused with PSS (to control the $P_{{\mathrm{O}}_2}$ and temperature), and the vessels imaged using a microscope. Both approaches allow the $P_{{\mathrm{O}}_2}$ in the environment around vessels to be measured, controlled and changed while the diameters of the vessels are measured, the main requirements to study vascular O₂ reactivity.

The study of isolated vessels, ex vivo, offers the advantage of precise control of experimental conditions: the composition of the bathing solutions, perfusion solutions, pressure, etc. can be precisely controlled. In addition, the cell type responsible for O₂‐mediated responses can be identified (Busse et al. 1983; Fredricks et al. 1994 a). However, this approach is suitable mainly for the study of feed arteries and first‐order arterioles due to the technical difficulties associated with the dissection and cannulation of vessels smaller than 50 μm. A major assumption of the ex vivo study of feed arteries is that the reactivity of these larger upstream vessels accurately models the reactivity of the smaller downstream arterioles, an assumption that has not been adequately tested. In addition, loss of input from hormones, extravascular cells and other vessels in the microcirculation may alter the function of isolated vessels. Trauma due to dissection of the vessels can also be problematic.

The primary advantage of the intravital microscopy approach is the ability to study arterioles of any size in their native environment. However, most studies have focused on smaller arterioles (third‐ to fifth‐order arterioles < 40 μm). As will be pointed out repeatedly in sections below, this difference between ex vivo and in situ studies (the study of arteries vs. arterioles) may contribute to the lack of consensus on the site and mechanism of action of O₂ in the microcirculation. The main disadvantage of the in situ approach is the lack of control of the environment. The arterioles are embedded in a connected microvascular network that allows conduction of signals along the vessel wall. They are perfused with systemic blood and surrounded by parenchymal cells, mast cells, nerves, etc. Therefore, the signals to which a given vessel is exposed are not always apparent. As will be outlined below, there are strategies to circumvent some of these issues, but these approaches are technically challenging. Because of the requirement to control tissue $P_{{\mathrm{O}}_2}$ and to observe vessel diameter, intravital microscopy is limited to the study of thin tissues such as the hamster cheek pouch or hamster, mouse, or rat cremaster muscles, although surface vessels can be studied in thicker tissues. The control of $P_{{\mathrm{O}}_2}$ in superfused exteriorized tissues can also be problematic. Care must be taken to limit the depth of the superfusion solution to 0.5 mm or less, and to use relatively high superfusate flow rates (5–10 ml min⁻¹ for cheek pouches or cremaster muscles) to obviate diffusional boundary layers and lack of the ability to accurately control $P_{{\mathrm{O}}_2}$ at the surface of the preparations. As with the ex vivo study of isolated vessels, surgical trauma and the resultant inflammation can also alter microvascular reactivity in intravital preparations.

What is the relevant $P_{{\mathrm{O}}_2}$ range over which arterioles are O₂ sensitive?

Intravital microscopy studies have defined the $P_{{\mathrm{O}}_2}$ range over which arterioles display O₂ reactivity as illustrated in Fig. 1 A. Arterioles respond to changes in $P_{{\mathrm{O}}_2}$ over a range from ∼10 to >70 mmHg in the tissue (Klitzman et al. 1982) and ∼30 to 150 mmHg at the surface of arterioles (Gorczynski & Duling, 1978; Jackson & Duling, 1983; Jackson, 1987) (Fig. 1 A). In Fig. 1 A, several $P_{{\mathrm{O}}_2}$ scales are provided, because the cell type in which changes in $P_{{\mathrm{O}}_2}$ are sensed, that will be termed the sensor, has yet to be firmly established. Thus, it is not clear what $P_{{\mathrm{O}}_2}$ (tissue $P_{{\mathrm{O}}_2}$ or arteriolar $P_{{\mathrm{O}}_2}$) is particularly relevant to the vasomotor effect of O₂ on the arterioles. However, these values (tissue $P_{{\mathrm{O}}_2}$ and arteriolar $P_{{\mathrm{O}}_2}$) should provide the lower and upper bounds, respectively, for defining the O₂ sensitivity of the system.

Figure 1. Arteriolar O₂ reactivity in superfused microvascular preparations .

A, data from Klitzman et al. (1982). Data are mean ± SEM diameters of arterioles in hamster cremaster muscles when exposed to solutions equilibrated with gases containing different $P_{{\mathrm{O}}_2}$values and 5% CO₂ (solution $P_{{\mathrm{O}}_2}$ values, upper x‐axis labels). The $P_{{\mathrm{O}}_2}$ values shown were measured with O₂ microelectrodes in the free solution flowing over the preparations. The tissue $P_{{\mathrm{O}}_2}$ values shown on the lower x‐axis were measured at the midpoint between the venous ends of two capillaries. The arteriolar $P_{{\mathrm{O}}_2}$ values (top x‐axis labels) were estimated from haemoglobin oxygen saturation measurements in hamster cheek pouch (Jackson & Duling, 1983). B, typical O₂‐induced vasoconstriction in a hamster cheek pouch preparation exposed to superfusion solutions with approximate $P_{{\mathrm{O}}_2}$values as indicated using methods as described (Jackson, 1987). C, same vessel as in B after exposure to ruthenium red (0.001%) to label mast cells (Shepherd & Duling, 1995). Scale bar in B, 25 μm.

Several points should be made regarding the relationship between O₂ and arteriolar diameter (tone) in superfused microvascular preparations as depicted in Fig. 1. First, arterioles in intravital preparations display substantial myogenic tone (diameter at rest = 50–80% of the maximum diameter obtained with a vasodilator) with low $P_{{\mathrm{O}}_2}$ in the superfusion solution (the solutions shown with $P_{{\mathrm{O}}_2}$ = 12 mmHg were equilibrated with gases containing 0% O₂ and 5% CO₂). Arterioles in microvascular preparations such as hamster (Jackson, 1986), mouse (Hungerford et al. 2000; Figueroa et al. 2003) or rat cremaster muscle (Jackson, 1986), hamster cheek pouch (Jackson & Duling, 1983; Jackson, 1987), hamster retractor muscle (Lombard et al. 1999) rat spinotrapezius muscle (Marvar et al. 2007) or mouse gluteus maximus muscle (Sinkler & Segal, 2014) all retain substantial tone; they are not maximally dilated at the low $P_{{\mathrm{O}}_2}$ values experienced by the preparations (note the right y‐axis scale in Fig. 1 A). Second, at rest, arterioles and the tissue surrounding them experience relatively ‘low’ $P_{{\mathrm{O}}_2}$ values due to diffusional loss of O₂ from precapillary vessels and consumption of O₂ by the tissue as well as diffusion of O₂ from the arterioles into the superfusion solution (for low $P_{{\mathrm{O}}_2}$ equilibrated solutions) (Duling & Berne, 1970). Early studies using recessed‐tip O₂ microelectrodes (Whalen et al. 1974) or multi‐point surface O₂ electrodes (Lund et al. 1980 a,b) indicated that there was a broad range of $P_{{\mathrm{O}}_2}$ values measured in the microcirculation of skeletal muscle, from < 1 to 100 mmHg. The $P_{{\mathrm{O}}_2}$ distributions recorded by both methods were dominated by relatively low $P_{{\mathrm{O}}_2}$ values with the means of the $P_{{\mathrm{O}}_2}$ distributions being in the order of 17 mmHg in the anaesthetized cat (Whalen et al. 1974), 33 mmHg in the anaesthetized rat (Lund et al. 1980 b) and 15 mmHg in the conscious human (Lund et al. 1980 a). More recent measurements support the findings of these early studies (Smith et al. 2002, 2004; Johnson et al. 2005). The average minimum resting tissue $P_{{\mathrm{O}}_2}$ (defined as the $P_{{\mathrm{O}}_2}$ midway between the venous end of two capillaries away from arterioles or venules) in a relatively undisturbed striated muscle microvascular bed, in vivo, is ∼25 mmHg as measured with phosphorescence probes (Smith et al. 2002, 2004; Johnson et al. 2005). Thus, a superfused microvascular preparation allows the study of both the response to mild decreases in tissue $P_{{\mathrm{O}}_2}$ (from 25 down to ∼10 mmHg, with the superfusate serving as a sink for O₂) and increases in $P_{{\mathrm{O}}_2}$ (>25 mmHg, with the superfusate serving as a source for O₂). An additional weakness of the superfused intravital microvascular preparations is the lack of ability to study more severe hypoxia (below 10 mmHg). Tissue $P_{{\mathrm{O}}_2}$ values below 10 mmHg can be attained during increases in metabolic activity (e.g. striated muscle contraction) (Lash & Bohlen, 1987; Boegehold & Bohlen, 1988; Smith et al. 2002), occlusion of blood flow (Lombard & Duling, 1977) or breathing gases with reduced O₂ content (Shah et al. 2003). However, each of these approaches adds additional complexity beyond simple effects of O₂ on the arterioles, and thus, will not be considered in this review.

Finally, it is important to recognize that elevated $P_{{\mathrm{O}}_2}$ causes profound, maintained vasoconstriction of arterioles in well‐prepared microvascular beds (preparations with substantial resting arteriolar tone at physiological $P_{{\mathrm{O}}_2}$values as presented above and without leukocyte adherence to arteriolar endothelium or other signs of inflammation) (see Fig. 1 B in this review, Fig. 2 in Ngo et al. 2010 and Fig. 1 in Ngo et al. 2013, for examples). This point is particularly important when considering the data obtained from the ex vivo study of isolated arteries and arterioles.

Figure 2. Schematic diagram of O₂ signalling in the microcirculation .

Oxygen in the environment of arterioles can act directly on the arteriolar wall or on cells in the lumen to produce a vasomotor effect (dilatation in the case of reduced $P_{{\mathrm{O}}_2}$, or constriction in the case of elevated $P_{{\mathrm{O}}_2}$). Alternatively, changes in $P_{{\mathrm{O}}_2}$ may be sensed by extravascular cells (parenchymal cells, mast cells, nerves, etc.), a mediator produced, which then acts on the vessel wall to produce the appropriate vasomotor effect. The $P_{{\mathrm{O}}_2}$ values shown below the cross section of the arterioles refer to tissue $P_{{\mathrm{O}}_2}$ in a superfused, intravital preparation measured at the midpoint between the venous ends of two capillaries as reference values, only.

Location of the sensor: what cell type senses changes in $P_{{\mathrm{O}}_2}$ relevant to arteriolar O₂ reactivity?

At least four cellular sites have been postulated to sense changes in $P_{{\mathrm{O}}_2}$ and signal vascular smooth muscle cells to contract (elevated $P_{{\mathrm{O}}_2}$) or relax (reduced $P_{{\mathrm{O}}_2}$) in response to this change. These include the arteriolar wall (vascular smooth muscle or endothelial cells), red blood cells in the lumen of the vessels, and extravascular cells (parenchymal cells, nerves, mast cells, etc., see Fig. 2).

Arteriolar smooth muscle cells as an O₂ sensor

Studies testing the hypothesis that arteriolar smooth muscle cells are intrinsically sensitive to changes in $P_{{\mathrm{O}}_2}$ within the range of 10–150 mmHg have lead to equivocal results. Early organ bath studies of preparations from conduit arteries (as model systems), ex vivo, indicated that smooth muscle cells are intrinsically sensitive to changes in bath $P_{{\mathrm{O}}_2}$ (Carrier et al. 1964; Detar & Bohr, 1968; Coburn et al. 1979; Chang & Detar, 1980). However, this intrinsic O₂ sensitivity appears to result from the formation of an anoxic core in the tissue due to O₂ consumption by the vascular smooth muscle cells and the multicellular thickness of these conduit artery preparations (Pittman & Duling, 1973). Thus, these studies do not appear to be relevant to arteriolar O₂ reactivity except, perhaps, under conditions when the $P_{{\mathrm{O}}_2}$ in the microcirculation falls to very low levels ($P_{{\mathrm{O}}_2}$ ≪ 10 mmHg).

Studies of the O₂ reactivity of smooth muscle cells of resistance arteries and arterioles have lead to conflicting results. Endothelium‐denuded rat gracilis muscle feed arteries studied ex vivo by pressure myography display O₂ reactivity in the appropriate $P_{{\mathrm{O}}_2}$ range (10–150 mmHg), but the response is only 15–25% of that observed in vessels with intact endothelium (Frisbee et al. 2002). Low $P_{{\mathrm{O}}_2}$ (∼12 mmHg) also inhibits vasoconstriction of first‐order rat cremaster arterioles constricted by activation of α₂‐adrenoreceptors (Tateishi & Faber, 1995 a,b). In addition, noradrenaline‐induced contraction of isolated hamster cremaster arteriolar smooth muscle cells is inhibited by exposure to low $P_{{\mathrm{O}}_2}$ (Jackson, 2000 b; Cohen & Jackson, 2003). In the latter studies, the low O₂‐induced inhibition of noradrenaline‐induced contraction of hamster cremaster arteriolar smooth muscle cells was not due to the activation of K⁺ channels (Jackson, 2000 b), the inhibition of L‐type Ca²⁺ channels (Cohen & Jackson, 2003), nor diminution of noradrenaline‐induced Ca²⁺ transients (Cohen & Jackson, 2003). These data suggested that the low $P_{{\mathrm{O}}_2}$ reduced the Ca²⁺ sensitivity of the contractile machinery, consistent with several studies of macrovascular smooth muscle cells (Gebremedhin et al. 1994; Aalkjaer & Lombard, 1995; Shimizu et al. 2000). On the other hand, patch clamp studies of isolated vascular smooth muscle cells from systemic arteries have demonstrated activation of ATP‐sensitive K⁺ (K_ATP) channels (Dart & Standen, 1995), activation of large‐conductance Ca²⁺‐activated K⁺ (BK_Ca) channels (Gebremedhin et al. 1994), or inhibition of L‐type voltage‐gated Ca²⁺ (Ca_L) channels (Franco‐Obregon et al. 1995; Franco‐Obregon & Lopez‐Barneo, 1996 a,b) when the cells are exposed to low $P_{{\mathrm{O}}_2}$ (15–20 mmHg), any of which could contribute to O₂ reactivity.

In contrast to the studies described above, removal of the endothelium from first‐order rat cremaster arterioles abolished O₂ reactivity in ex vivo pressure‐myography studies (Messina et al. 1992, 1994). These data exclude smooth muscle as the O₂ sensor in these isolated rat cremaster arterioles. Furthermore, several additional ex vivo studies of intact, endothelium‐replete rat first‐order cremaster arterioles have failed to demonstrate effects of changes in $P_{{\mathrm{O}}_2}$ between 10 and 150 mmHg on myogenic tone (Tateishi & Faber, 1995 a,b; Kerkhof et al. 1999), again not supporting smooth muscle as the O₂ sensor. In the studies by Tateishi and Faber (Tateishi & Faber, 1995 a) reduced O₂ tensions ($P_{{\mathrm{O}}_2}$ < 6 mmHg) inhibited myogenic tone suggesting intrinsic O₂ reactivity at much lower $P_{{\mathrm{O}}_2}$ values.

Intravital microscopy studies performed in hamster cheek pouches have also failed to support smooth muscle, or other components of the vascular wall, as the O₂ sensor (Duling, 1974; Jackson, 1987). Altering the $P_{{\mathrm{O}}_2}$ of the wall of arterioles through the use of fluid‐filled micropipettes (Duling, 1974; Jackson, 1987) (Figs 3 and 4), or by in situ perfusion of segments of arterioles in the hamster cheek pouch (Jackson, 1987) (Fig. 5) failed to demonstrate arteriolar O₂ reactivity intrinsic to the arteriolar wall. In contrast to these findings, elevation of the $P_{{\mathrm{O}}_2}$ of the superfusion solution flowing over the entire cheek pouch produced typical O₂‐induced vasoconstriction (Duling, 1974; Jackson, 1987) (Figs 3, 4, 5). The lack of effect of local changes in $P_{{\mathrm{O}}_2}$ on arteriolar diameter (Figs 3, 4, 5) do not support the hypothesis that arteriolar smooth muscle cells are intrinsically sensitive to changes in $P_{{\mathrm{O}}_2}$ between 10 and 150 mmHg.

Figure 3. Local increases in ${\mathit{}\mathit{P}}_{{\mathit{}\mathrm{O}}_{\mathit{}2}}$ have no effect on arteriolar tone .

Data shown are modified from Jackson (1987). A, schematic diagram of a segment (1–8 mm) of an arteriole in a hamster cheek pouch from which the parenchyma has been surgically removed (aparenchymal arteriole) to obviate effects of local $P_{{\mathrm{O}}_2}$ changes on parenchymal and other extravascular cells. A Whalen‐type recessed tip O₂ microelectrode was inserted through the wall of the vessel into the lumen as shown to monitor luminal $P_{{\mathrm{O}}_2}$. A temperature‐controlled micropipette filled with PSS equilibrated with varied $P_{{\mathrm{O}}_2}$ (fluid‐filled micropipette) was positioned opposite the O₂ microelectrode. Pressurization of the fluid‐filled micropipette ejected the O₂‐equilibrated solution onto the surface of the arteriole to produce a local change in $P_{{\mathrm{O}}_2}$ _. The entire cheek pouch preparation was superfused with PSS, the $P_{{\mathrm{O}}_2}$ of which could be varied to produce global changes in $P_{{\mathrm{O}}_2}$ that affected both the aparenchymal arteriole and all other vessels and cell types in the cheek pouch. The diameter of the arteriole was measured by intravital video microscopy. B, results from a typical experiment in which either local increases in $P_{{\mathrm{O}}_2}$ were produced using the fluid‐filled pipette (Local) or global increases in $P_{{\mathrm{O}}_2}$ were produced by changing the $P_{{\mathrm{O}}_2}$ of the entire superfusate (Global) (replotted data are from Fig. 3 in Jackson, 1987). Local increases in $P_{{\mathrm{O}}_2}$ that effectively changed the $P_{{\mathrm{O}}_2}$ across the wall of the arteriole had no significant effect on arteriolar diameter, whereas a global increase in $P_{{\mathrm{O}}_2}$ produced sustained vasoconstriction. These data suggest that components of the arteriolar wall (endothelial cells, smooth muscle cells or perivascular nerves) are not the sensor cells responsible for arteriolar O₂ reactivity in the hamster cheek pouch. See Jackson (1987) for details.

Figure 4. Local increases in ${\mathit{}\mathit{P}}_{{\mathrm{O}}_2}$ have no effect on arteriolar tone in occluded arterioles .

A, a preparation similar to that depicted in Fig. 3, but with the inclusion of an occluding pipette that was pressed down on the arteriole to eliminate blood flow through the aparenchymal segment. B, summary data from these experiments (data are means ± SEM, n = 5). Local changes in $P_{{\mathrm{O}}_2}$ across the arteriolar wall produced by the fluid‐filled pipette (Local) were ineffective in producing arteriolar constriction. However, raising the $P_{{\mathrm{O}}_2}$ of the superfusate over the entire preparation (Global) produced consistent, sustained arteriolar constriction. These data, along with those shown in Fig. 3, suggest that components of the arteriolar wall (endothelial cells, smooth muscle cells or perivascular nerves) are not the sensor cells responsible for arteriolar O₂ reactivity in the hamster cheek pouch. Data are replotted from Fig. 4 B in Jackson (1987); see this reference for more details.

Figure 5. Perfusion of arterioles in situ with solutions equilibrated with high ${\mathit{}\mathit{P}}_{{\mathit{}\mathrm{O}}_{\mathit{}2}}$ has no effect on arteriolar tone .

A and B are reproduced from Jackson (1987). A, schematic diagram of a hamster cheek pouch arteriole in which a sharpened fluid‐filled pipette has been inserted through the wall of an arteriole allowing perfusion of the arteriole with PSS equilibrated with varied $P_{{\mathrm{O}}_2}$. The entire cheek pouch preparation was superfused with PSS to allow global changes in $P_{{\mathrm{O}}_2}$. B, summary data (means ± SEM). Perfusion of the arterioles with solutions equilibrated with high or low $P_{{\mathrm{O}}_2}$ had no significant effect on arteriolar diameter. Only when the global $P_{{\mathrm{O}}_2}$ was elevated via the superfusate did the arterioles constrict (compare low $P_{{\mathrm{O}}_2}$ superfusate points with high $P_{{\mathrm{O}}_2}$ superfusate points). These data suggest that components of the arteriolar wall (endothelial cells, smooth muscle cells or perivascular nerves) are not the sensor cells responsible for arteriolar O₂ reactivity in the hamster cheek pouch. See Jackson (1987) for details.

Thus, smooth muscle per se, in some vessels (resistance arteries and first‐order arterioles), under some conditions, may indeed sense changes in $P_{{\mathrm{O}}_2}$ in the range of 10–150 mmHg. However, there does not appear to be significant support for smooth muscle cells as the O₂ sensor, particularly based on in situ studies of arterioles in the intact microcirculation.

Arteriolar endothelial cells as an O₂ sensor

There is considerable evidence from ex vivo, pressure‐myography studies of isolated vessels that endothelial cells may serve as an O₂ sensor, at least in small arteries and large arterioles. Busse and coworkers first demonstrated endothelium‐dependent O₂ reactivity in rat tail arteries, canine femoral artery branches and canine epicardial coronary arteries (Busse et al. 1983, 1984). These findings were subsequently confirmed in rabbit aortas and rabbit femoral arteries (Pohl & Busse, 1989), hamster carotid arteries (Jackson et al. 1987), first‐order rat cremaster arterioles (Messina et al. 1992, 1994), rat middle cerebral arteries (Fredricks et al. 1994 b), large arterioles from rat diaphragm (Ward, 1999), and rat gracilis muscle feed arteries (Fredricks et al. 1994 a; Frisbee et al. 2001 a,b,c, 2002).

In contrast, several pressure‐myography studies of endothelium‐intact first‐order rat cremaster arterioles have failed to demonstrate effects of $P_{{\mathrm{O}}_2}$ changes on myogenic tone (Tateishi & Faber, 1995 a,b; Kerkhof et al. 1999), as noted above. The apparent difference between the studies showing endothelium‐dependent O₂ reactivity (Busse et al. 1983, 1984; Messina et al. 1992, 1994; Fredricks et al. 1994 a,b; Ward, 1999; Frisbee et al. 2001 a,b,c, 2002) and those not displaying O₂ reactivity (Tateishi & Faber, 1995 a,b; Kerkhof et al. 1999) is the presence of luminal flow in the former and not in the latter. The exception to this relationship is a pressure‐myography study by Dietrich et al. (2000) using endothelium‐intact rat cerebral arterioles. These investigators found that PSS or dextran–PSS‐perfused arterioles did not display O₂ reactivity in the absence of luminal red blood cells.

Lack of evidence for endothelium‐dependent O₂ reactivity also can be gleaned from the hamster cheek pouch studies already cited in the previous section (Duling, 1974; Jackson, 1987). In those studies, either blood (Duling, 1974; Jackson, 1987) or O₂‐equilibrated PSS (Jackson, 1987) was flowing through the arterioles. In these studies, despite effectively changing the $P_{{\mathrm{O}}_2}$ of endothelial cells, O₂ reactivity was not observed when only the vessel wall was exposed to changes in $P_{{\mathrm{O}}_2}$. These data do not support an endothelial O₂ sensor in arterioles in the hamster cheek pouch. They, along with the studies by Dietrich et al. (2000), suggest that shear stress is not the key to promoting endothelial cell O₂ sensing.

Thus, there is evidence both for and against an endothelial cell O₂ sensor. However, the extant data supporting this hypothesis stem from pressure‐myography studies of isolated arteries and large arterioles, ex vivo. No data from intravital preparations support endothelial cells as the sensors mediating arteriolar O₂ reactivity.

As noted in the previous paragraph, all of the ex vivo studies of vascular O₂ reactivity have been performed on arteries or large, first‐order arterioles (diameters > 50 μm). In contrast, most in situ intravital microscopy studies of arteriolar O₂ reactivity have focused on small, third‐ to fifth‐order arterioles (diameters < 40 μm). Given the known differences in mechanisms regulating myogenic tone between small arteries and downstream arterioles (Westcott & Jackson, 2011; Westcott et al. 2012), it may be that there are different mechanisms (and hence, sites of action) coupling changes in $P_{{\mathrm{O}}_2}$ with changes in vascular tone in different parts of the vascular tree. In addition, as noted above, in situ, elevated $P_{{\mathrm{O}}_2}$ produces dramatic, long lasting constriction of arterioles (Fig. 1). This degree of O₂ reactivity has not been described in the ex vivo study of isolated vessels, perhaps reflecting that mostly feed arteries and first‐order arterioles have been studied ex vivo. However, these data may also indicate that sites other than the arteriolar wall are more important for O₂ reactivity of arterioles in the intact, living microcirculation.

Red blood cells as an O₂ sensor

Red blood cells also have been proposed as the O₂‐sensitive cell type responsible for arteriolar O₂ reactivity. Three lines of evidence support this hypothesis. First, in vitro, red blood cells release ATP as $P_{{\mathrm{O}}_2}$ is reduced from 85 mmHg down to 17 mmHg (Ellsworth et al. 1995). Second, at concentrations consistent with the amount of ATP released from red blood cells, intraluminal ATP produces conducted dilatation of arterioles in the hamster retractor muscle in situ, through a mechanism mediated by NO (McCullough et al. 1997). Intraluminal application of ATP into collecting venules also produces dilatation of upstream arterioles suggesting conduction of vasodilatation initiated by ATP from venules to arterioles through capillaries (Collins et al. 1998) (see below for more on conducted responses). Third, inclusion of red blood cells in the solution perfusing rat cerebral arterioles in pressure‐myography studies instills O₂ reactivity to isolated arterioles that do not respond to changes in $P_{{\mathrm{O}}_2}$ in their absence (Dietrich et al. 2000).

In contrast, the studies outlined in Figs 3, 4, 5 (Jackson, 1987) suggest that red blood cells may not be the O₂ sensor that mediates arteriolar O₂ reactivity in situ. In those studies, local changes in $P_{{\mathrm{O}}_2}$ that effectively altered blood $P_{{\mathrm{O}}_2}$ had no effect on arteriolar tone (Figs 3 and 4). Oxygen reactivity was not observed whether blood was flowing through the lumen of the arterioles (Fig. 3) or not (Fig. 4). In addition, if red blood cells were major O₂ sensors coupling changes in $P_{{\mathrm{O}}_2}$ with changes in vasomotor tone, then perfusion of arterioles in situ with PSS should eliminate arteriolar O₂ reactivity. This was not the case in the hamster cheek pouch (Fig. 5) (Jackson, 1987) or in the mouse cremaster muscle (Ngo et al. 2010); in both tissues, arteriolar O₂ reactivity was retained when the tissues were perfused with physiological salt solution and no red blood cells were present. Thus, in situ, red blood cells may not be a major site that senses changes in $P_{{\mathrm{O}}_2}$ values greater than 10–15 mmHg.

Extravascular cells as an O₂ sensor

To test the hypothesis that extravascular cells (parenchymal cells, etc.) are the location of the O₂ sensor involved in arteriolar O₂ reactivity in the hamster cheek pouch, Jackson and Duling surgically removed the parenchyma along long segments of blood‐perfused second‐order arterioles (Jackson & Duling, 1983) (Figs 3, 4 and 6). Surprisingly, this manoeuvre had no significant effect on O₂ reactivity: the aparenchymal arteriolar segments continued to respond to changes in superfusate $P_{{\mathrm{O}}_2}$ as the solution flowed over the entire preparation (Jackson & Duling, 1983). Furthermore, sealing the aparenchymal arteriolar segments under glass (to prevent superfusate access) and occluding the arterioles (to obviate convection of O₂ in the blood to the segments) did not eliminate O₂ reactivity (Fig. 6). To complicate matters, we found that some isolated, cannulated cheek pouch arterioles retained O₂ reactivity. However, the responses of these isolated arterioles were ephemeral. Taken as a whole, the data from this study suggested that: (1) changes in $P_{{\mathrm{O}}_2}$ are probably sensed at extravascular sites that are non‐uniformly distributed along the arterioles (to account for the lack of O₂ reactivity in some isolated vessels), and (2) O₂‐dependent vasoconstriction, once initiated, can be conducted along the arteriolar wall (to account for the maintained O₂ reactivity of aparenchymal segments sealed under glass: Fig. 6) (see below for more on the topic of conducted O₂ responses). Subsequent studies, outlined above (Fig. 4), in which segments of cheek pouch arterioles were perfused in situ with solutions with varied $P_{{\mathrm{O}}_2}$ supported and extended these conclusions (Jackson, 1987).

Figure 6. Covering aparenchymal segments with glass and occluding them to eliminate blood flow does not eliminate arteriolar O₂ reactivity .

A, schematic diagram of an aparenchymal segment in which the parenchyma has been removed from a long segment of hamster cheek pouch arteriole as described by Jackson & Duling (1983). Elevation of the $P_{{\mathrm{O}}_2}$ of the solution flowing over the preparation from 12 to 150 mmHg resulted in arteriolar constriction as depicted in D. As shown in B, subsequent covering of the aparenchymal segment with a piece of glass coverslip (sealed in place with silicone grease), to eliminate contact of the arteriole with the superfusate, had no effect on O₂‐induced constriction as shown in D. To eliminate blood flow through the covered aparenchymal segments, an occluding pipette was used as shown in C. Despite the lack of access to the superfusate and flowing blood, these covered and occluded aparenchymal segments retained significant O₂ reactivity as shown in D. These data suggest that the constriction induced by elevated $P_{{\mathrm{O}}_2}$ can be conducted along the arteriolar wall. See Jackson & Duling (1983) for details.

Mast cells: a possible O₂ sensor in the hamster cheek pouch

Arterioles in the hamster cheek are intimately associated with mast cells as shown in Fig. 1 C. These cells cluster, non‐uniformly, near the arterioles and they produce cysteinyl leukotrienes (CysLTs) (Storch et al. 2015), vasoconstrictors that mediate arteriolar O₂ reactivity in the hamster cheek pouch (see below for more on this topic). Mast cells as the location of the O₂ sensor may also explain why some isolated cannulated cheek pouch arterioles retained O₂ reactivity in a previous study (Jackson & Duling, 1983). Dissection of these arterioles may remove or damage some/all of these perivascular cells accounting for the loss and the labile nature of O₂ reactivity in the ex vivo study of cheek pouch arterioles.

Conduction of O₂‐induced vasoconstriction: a complicating factor

As noted above, we (Jackson & Duling, 1983) found that even when arterioles in the hamster cheek pouch were sealed under glass (to prevent superfusion solution access) and occluded (to obviate convection of O₂ into the segment by blood flow) the vessels retained O₂ reactivity (Fig. 6). We hypothesized that O₂‐induced vasoconstriction could be conducted along the arterioles over considerable distance (at least several millimetres). Mouse cremaster muscle arterioles also display conducted O₂‐induced vasoconstriction (Ngo et al. 2010; Riemann et al. 2011), suggesting that this phenomenon is not restricted to the hamster cheek pouch.

Conduction of O₂‐induced arteriolar responses implies that responses can be initiated away from the site of observation and, importantly, that responses initiated at several sites may summate, as has been observed for conducted vasomotion in response to other vasoactive substances (Segal et al. 1989). Based on conducted vasomotor responses induced by other vasoactive substances (Delashaw & Duling, 1991; Xia & Duling, 1995), the conduction of O₂‐induced vasomotor responses implies that changes in vascular smooth muscle membrane potential are probably involved in the mechanism of action of O₂ on arterioles. As will be discussed below, this prediction was verified in the hamster cheek pouch where O₂‐induced vasoconstriction was preceded by depolarization of the smooth muscle cells (Welsh et al. 1998).

Conduction of O₂‐induced changes in vascular tone along arterioles complicates identification of the location of the cell type that actually senses changes in $P_{{\mathrm{O}}_2}$ because the sensor cells can be remote from the site of observation. Changes in $P_{{\mathrm{O}}_2}$, could, for example, be sensed in parenchyma around terminal arterioles or even capillaries or venules downstream from arterioles, a response initiated, and this response conducted upstream to produce the appropriate change in arteriolar tone as suggested by Jackson (1987) and later by Collins et al. (1998) and Ellsworth et al. (2016). Oxygen sensing in the vicinity of terminal arterioles, capillaries or post‐capillary venules might explain why, in situ, changes in $P_{{\mathrm{O}}_2}$ that are restricted to segments of second‐ to third‐order arterioles, do not produce a change in arteriolar tone, whereas global changes in $P_{{\mathrm{O}}_2}$ consistently produce an upstream arteriolar response (Duling, 1974; Jackson, 1987). However, in the hamster cheek pouch, conduction of vasoconstrictor‐induced depolarization occurs along the smooth muscle layer (Welsh & Segal, 1998). This constraint suggests that, in the cheek pouch, the O₂‐sensing cells are probably adjacent to arteriolar smooth muscle cells and not in downstream capillaries or venules (see below for more on this topic). The pathway for conduction in skeletal muscle arterioles has not been as firmly established.

The hypothesis that changes in $P_{{\mathrm{O}}_2}$ can be sensed in capillaries was tested in rat extensor digitorum longus muscles using a microfluidic device approach (Ghonaim et al. 2011). The $P_{{\mathrm{O}}_2}$ of regions overlying capillary beds ranging from 100 μm circles (affecting 1 capillary and surrounding skeletal muscle fibres) to 200 μm × 1000 μm rectangles (affecting ∼3 capillaries and surrounding muscle fibres) were changed (Ghonaim, 2013), with capillary red blood cell haemoglobin O₂ saturation (%Sat) and red blood cell flux measured. The authors found that while changes in $P_{{\mathrm{O}}_2}$ applied through 100 or 200 μm circles (affecting only 1–2 capillaries) produced appropriate changes in red blood cell %Sat, they were without effect on red blood cells flux. In contrast, changes in $P_{{\mathrm{O}}_2}$ applied to a 200 μm × 1000 μm area (affecting ∼3 capillaries and associated muscle fibres) produced changes in red blood cell flux through the affected capillaries. These results suggest that there may be summation of responses to altered $P_{{\mathrm{O}}_2}$ that are required to initiate a conducted response sufficient to affect upstream arterioles that control red blood cell flux through the capillaries. Although the authors interpreted these findings as evidence for capillaries as the site where changes in $P_{{\mathrm{O}}_2}$ are sensed (in the context of red blood cells as the O₂ sensors), effects of changes in $P_{{\mathrm{O}}_2}$ on striated muscle fibres, which are also putative O₂ sensors (see below), were not excluded. Striated muscle fibres are much longer than capillaries, and they will contact arterioles in the network. Thus, it is also possible that the striated muscle fibres sensed changes in $P_{{\mathrm{O}}_2}$, a response initiated and transmitted to upstream arterioles to modulate arteriolar smooth muscle tone and hence capillary perfusion.

Striated muscle fibres as an O₂ sensor

Frisbee and Lombard compared the O₂ reactivity of first‐order arterioles in rat cremaster muscles, in situ, with similar arterioles studied by pressure myography, ex vivo (Frisbee & Lombard, 2002). They found that while these large arterioles retained O₂ reactivity when studied ex vivo, the maximal O₂‐induced diameter responses of the isolated vessels were only 50% of the responses observed in situ. These data support a role for striated muscle fibres as an O₂ sensor in rat cremaster muscle. As will be outlined below, the location of cytochrome P450_4A (CYP450_4A) ω‐hydroxylase, a key enzyme involved in arteriolar O₂ reactivity in striated muscle fibres (Kunert et al. 2001 a), also lends support for these cells as O₂ sensors in striated muscle.

Summary of the location of the sensor

Ex vivo studies of isolated feed arteries and first‐order arterioles support the hypothesis that these vessels are intrinsically sensitive to changes in $P_{{\mathrm{O}}_2}$ within the appropriate range (10–150 mmHg), with smooth muscle, endothelium, or red blood cells as the location of the sensor (see references above). In contrast, intravital microscopy studies (Duling, 1974; Jackson & Duling, 1983; Jackson, 1987), all in the hamster cheek pouch, and all directed at smaller, downstream arterioles, have failed to support the hypothesis that these smaller vessels are intrinsically sensitive to O₂ in situ. Furthermore, local changes in $P_{{\mathrm{O}}_2}$ that effectively changed the $P_{{\mathrm{O}}_2}$ of the blood flowing through the arterioles in the cheek pouch were also without effect on arteriolar tone (Jackson, 1987). Microvascular beds perfused with PSS retain O₂ reactivity (Jackson, 1987; Ngo et al. 2010). These data do not support the hypothesis that red blood cells are a major sensor mediating arteriolar O₂ reactivity in the intact microcirculation of the hamster cheek pouch or mouse cremaster muscle. Instead, the data from intravital preparations suggest that changes in $P_{{\mathrm{O}}_2}$ are sensed at some extravascular site. These conflicting conclusions may indicate that vascular O₂ sensing varies along the vascular tree with upstream feed arteries possessing intrinsic mechanisms to respond to changes in blood $P_{{\mathrm{O}}_2}$, and downstream arterioles being tuned to detect local changes in tissue $P_{{\mathrm{O}}_2}$. However, tissue‐dependent heterogeneity also cannot be excluded as studies to evaluate the intrinsic O₂ sensitivity of small arterioles (<30 μm), in situ, have not been evaluated in tissues other than the hamster cheek pouch. Conduction of O₂‐induced changes in arteriolar tone also complicates identification of the microvascular O₂ sensor that has not been adequately explored in preparations other than the hamster cheek pouch.

Mechanism of action

Several signalling pathways have been proposed to explain arteriolar O₂ reactivity. These can be separated into two groups: (1) mechanisms in which low O₂ results in increased production or decreased destruction of a vasodilator substance, and (2) mechanisms by which increased O₂ results in production of a vasoconstrictor. As will be evident in the sections below, there appears to be significant regional heterogeneity that has complicated the study of the mechanisms responsible for arteriolar O₂ reactivity. As discussed above, conduction of O₂‐initiated responses also adds to the complexity of defining the mechanism of action of O₂ on arterioles in the microcirculation.

Oxygen‐dependent inhibition of production of prostanoids

A number of ex vivo studies of isolated vessels have supported a role for prostaglandins in mediating dilatation of resistance arteries and large arterioles to a reduction in $P_{{\mathrm{O}}_2}$ from 150 mmHg to 20–50 mmHg (Busse et al. 1983, 1984; Fredricks et al. 1994 a,b; Frisbee et al. 2001 b,c, 2002) or constriction of these vessels to increases in $P_{{\mathrm{O}}_2}$ from 20 to 150–600 mmHg (Messina et al. 1994; Frisbee et al. 2001 a).

However, the relevance of these ex vivo studies to in situ arteriolar O₂ reactivity remains unclear. Effective inhibition of cyclooxygenase has no effect on arteriolar O₂ reactivity in intravital preparations of the hamster cheek pouch, hamster and rat cremaster muscles (Jackson, 1986), rat spinotrapezius muscle (Pries et al. 1995) and mouse cremaster muscle (Ngo et al. 2013). A weakness in these intravital studies is that only large changes in $P_{{\mathrm{O}}_2}$ (from ∼12 to 150 mmHg) were examined. Thus, subtle effects of cyclooxygenase inhibition on the $P_{{\mathrm{O}}_2}$–response relation could have been missed. Nonetheless, the intravital studies do not support a sole, major role for prostaglandins in mediating arteriolar O₂ reactivity in the intact microcirculation. However, the intravital studies do not exclude a role for prostaglandins in larger feed arteries outside of the tissue proper.

The mechanism by which elevated $P_{{\mathrm{O}}_2}$ inhibits, and lowered $P_{{\mathrm{O}}_2}$ enhances, vasodilator prostanoid formation also is not known. The enzymes catalysing oxygenation of arachidonic acid, the cyclooxygenases (COXs), are half‐maximally activated at a $P_{{\mathrm{O}}_2}$ of about 3 mmHg (Lands et al. 1978; Mukherjee et al. 2007), with maximal activity at $P_{{\mathrm{O}}_2}$ values greater than 18 mmHg (Lands et al. 1978; Mukherjee et al. 2007). Thus, formation of prostanoids should be $P_{{\mathrm{O}}_2}$ independent over much of the $P_{{\mathrm{O}}_2}$ range to which arterioles are O₂ sensitive (Fig. 1). However, the COXs can undergo auto‐oxidation, which inhibits enzyme activity (Tsai & Kulmacz, 2010). Oxygen‐dependent oxidation and inhibition of COXs might explain O₂‐dependent inhibition of production of vasodilator prostanoids, but this has not been established.

Oxygen‐dependent destruction of nitric oxide

Reducing the $P_{{\mathrm{O}}_2}$ of solutions perfusing rabbit aortas and rabbit femoral arteries (Pohl & Busse, 1989) or hamster carotid arteries (Jackson et al. 1987), ex vivo, from 150 to 20–30 mmHg produces endothelium‐dependent vasodilatation mediated by NO. Nitric oxide has also been proposed to mediate arteriolar O₂ reactivity in rat spinotrapezius muscle (Pries et al. 1995) and the O₂ reactivity of rat intestinal arterioles (Nase et al. 2003) in situ. Ex vivo studies of rat gracilis muscle feed arteries also support a role for NO, but only at relatively high $P_{{\mathrm{O}}_2}$ values (between 100 and 150 mmHg) (Frisbee et al. 2002). Pressure‐myography studies of rat first‐order cremaster muscle arterioles, ex vivo, found that NO mediated O₂ reactivity, but only after application of the CYP450_4A ω‐hydroxylase inhibitor, 17‐octadecynoic acid (17‐ODYA) (Kerkhof et al. 1999). In the absence of 17‐ODYA, these authors found that isolated rat cremaster arterioles did not respond to changes in $P_{{\mathrm{O}}_2}$ suggesting that endogenous production of 20‐hydroxyeicosatetraenoic acid (20‐HETE) by CYP450_4A ω‐hydroxylation of arachidonic acid inhibited O₂‐dependent signalling by NO.

In contrast to the studies supporting a role for NO in arteriolar O₂ reactivity, intravital studies in the hamster cheek pouch (Jackson, 1991) and in mouse cremaster muscle (Ngo et al. 2010, 2013; Riemann et al. 2011) found that effective inhibition of NO synthesis had no effect on arteriolar O₂ reactivity. An earlier ex vivo study of rat gracilis feed arteries also did not support a role for NO in the reactivity of these vessels to changes in $P_{{\mathrm{O}}_2}$ between 35 and 150 mmHg $P_{{\mathrm{O}}_2}$ (Frisbee et al. 2001 a). These findings may indicate that there is heterogeneity in the mechanism of action of O₂, which is dependent on the region, location in the vascular tree and possibly on the species studied. In addition, another study in the rat spinotrapezius muscle suggests a major role for CYP450_4A ω‐hydroxylase and 20‐HETE in arteriolar O₂ reactivity (see below) (Marvar et al. 2007). This observation is difficult to reconcile with the findings of Pries et al. cited above, who reported that inhibition of NO synthase abolished arteriolar O₂ reactivity in the same tissue, from the same rat strain (Pries et al. 1995). It could be that multiple mechanisms are in play, as suggested by ex vivo studies of first‐order rat cremaster arterioles noted above (Kerkhof et al. 1999). Subtle methodological differences also cannot be excluded. Thus, the contribution of NO to arteriolar O₂ reactivity remains unclear.

Oxygen‐dependent production of superoxide anion

If NO mediates arteriolar O₂ reactivity, the relationship between O₂ and NO production cannot be direct, because O₂ is a substrate for NO synthase, with half‐maximal activation at about 2 mmHg for the isolated enzyme (Abu‐Soud et al. 2000), or as high as 38 mmHg in cell‐based assays (Whorton et al. 1997). Thus, NO production should increase or remain unchanged as O₂ increases. This is opposite to what would be required for direct effects of O₂ on NO production to mediate arteriolar O₂ reactivity.

It has been suggested that O₂‐dependent production of superoxide anion (O₂ ^−•) and the subsequent destruction of NO may explain arteriolar O₂ reactivity in situ (Golub & Pittman, 2013). However, previous direct tests of this hypothesis in striated muscle preparations using exogenous superoxide dismutase (SOD) both failed (Proctor & Duling, 1982 b; Pries et al. 1995). Golub & Pittman (2013) argue that the kinetics of the reaction between O₂ ^−• and NO is so much faster than that between O₂ ^−• and SOD, and that the tissue content of extracellular SOD is so high in striated muscle, the use of exogenous SOD does not adequately test for a role for O₂ ^−•. In the rat brain, where there is lower expression of extracellular SOD, hyperbaric O₂‐induced constriction of resistance arteries indeed does appear to be mediated by O₂‐dependent production of O₂ ^−• and destruction of NO, and can be inhibited by application of exogenous SOD (Demchenko et al. 2000). Similarly, superfusion with SOD and catalase or the SOD mimetic tempol abolishes O₂‐induced constriction of rat sciatic epineural arterioles, supporting a role for O₂ ^−• in these vessels (Sakai et al. 2007). Inhibitors of NADPH oxidase or xanthine oxidase also eliminated O₂ reactivity, suggesting that these enzymes may be the O₂ sensors in rat epineural arteriolar O₂ reactivity (Sakai et al. 2007). However, unlike the studies in the brain (Demchenko et al. 2000), inhibition of NO synthase did not inhibit O₂ reactivity of rat sciatic epineural arterioles suggesting that alterations in NO bioavailability do not explain O₂ reactivity in this tissue. In addition, earlier studies demonstrated that O₂ ^−• is a dilator in cat (Wei et al. 1996) and rabbit (Didion & Faraci, 2002) cerebral arterioles, and human coronary arterioles (Sato et al. 2003), supporting the idea that the O₂ ^−•–NO destruction pathway is not a general mechanism accounting for arteriolar O₂ reactivity. The lack of effect of effective inhibition of NO production on O₂ reactivity in other preparations (Jackson, 1991; Ngo et al. 2010, 2013; Riemann et al. 2011) also argues that this is not a general mechanism explaining arteriolar O₂ reactivity in all tissues and every species.

All of the mechanisms discussed above involve O₂‐dependent modulation of a vasodilator: as $P_{{\mathrm{O}}_2}$ increases, the dilator signal would decrease and as $P_{{\mathrm{O}}_2}$ decreases, the dilator signal would increase. Recall that in the normal resting microcirculation of skeletal muscle, for example, the $P_{{\mathrm{O}}_2}$ in the environment of arterioles is low. Thus, if arteriolar O₂ reactivity involves modulation of a vasodilator, then when arterioles are isolated and cannulated for ex vivo study and exposed to ambient O₂ conditions (21% O₂), these arterioles should display substantially greater tone than they displayed in situ under low $P_{{\mathrm{O}}_2}$ (∼10–15 mmHg) conditions. This simply has not been observed: the degree of myogenic tone developed by cannulated, pressurized arterioles ex vivo exposed to room air (21% O₂, $P_{{\mathrm{O}}_2}$ ∼150 mmHg) is similar to the tone observed when these same vessels are studied in situ by intravital microscopy under low $P_{{\mathrm{O}}_2}$ conditions ($P_{{\mathrm{O}}_2}$ in the order of 10–15 mmHg) (for example, compare Jackson & Blair, 1998 with Burns et al. 2004). This observation alone suggests that arteriolar O₂ reactivity in situ does not involve a vasodilator and also that it is unlikely that O₂ is sensed directly by cells that comprise the wall of the arterioles.

Oxygen as a vasoconstrictor

In contrast to the mechanisms discussed thus far involving the O₂‐dependent modulation of vasodilators, there is a substantial body of literature that O₂‐dependent production of vasoconstrictors mediates arteriolar O₂ reactivity in the peripheral microcirculation (see below for references). These vasoconstrictor mechanisms largely have been ignored when considering the role played by O₂ in the local regulation of blood flow (Casey & Joyner, 2011; Marshall & Ray, 2012; Golub & Pittman, 2013; Reglin & Pries, 2014).

Oxygen‐dependent production of leukotrienes

The 5‐lipoxygenase and CysLTs appear to mediate arteriolar O₂ reactivity in the epithelial portion of the hamster cheek pouch (Jackson, 1988, 1989, 1993) (Fig. 7). General lipoxygenase inhibitors (Jackson, 1988), selective 5‐lipoxygenase inhibitors (Jackson, 1989, 1993), a selective inhibitor of the 5‐lipoxygenase‐activating protein (FLAP) (Jackson, 1993), as well as antagonists of CysLT receptors (Jackson, 1989, 1993) all selectively and effectively inhibit O₂‐induced arteriolar constriction in the cheek pouch. Importantly, these same inhibitors have no effect on the O₂ reactivity of arterioles in hamster cremaster muscles, supporting their vascular bed selectivity (Jackson, 1993). These latter findings also strongly support the hypothesis that there are regional differences in the mechanisms coupling changes in $P_{{\mathrm{O}}_2}$ with changes in arteriolar tone.

Figure 7. The site and mechanism of action of O₂ in the hamster cheek pouch .

Schematic diagram depicting a mast cell (the proposed sensor site in the cheek pouch), a smooth muscle cell replete with receptors for cysteinyl leukotrienes (CysLTRs) and ion channels involved in arteriolar O₂ reactivity in the cheek pouch. Elevated $P_{{\mathrm{O}}_2}$ is sensed by the 5‐lipoxygenase (5‐LO) in the nuclear membrane of periarteriolar mast cells that decorate arterioles in this tissue (see Fig. 1). This results in conversion of arachidonic acid to cysteinyl leukotrienes (CysLTs) such as LTC₄, LTD₄ and LTE₄ through a process that involves presentation of the arachidonic acid to the 5‐LO by the 5‐LO‐activating protein (FLAP). This process can be inhibited by drugs such as MK 866, that blocks interaction of FLAP with the 5‐LO, or SC 43251, U 60257, nordihydroguaiaretic acid (NDGA), eicosatetraynoic acid (ETYA) or phenidone, inhibitors of the 5‐LO. The CysLTs then bind to and activate CysLTRs on vascular smooth muscle cells to induce vasoconstriction. CysLTR antagonists such as SKF 102922 or FPL 55712 can inhibit this step in the process. Activation of CysLTRs results in activation of L‐type Ca²⁺ channels (Ca_L), Ca²⁺ influx, an increase in intracellular Ca²⁺ and vasoconstriction, which can be antagonized by Ca_L blockers such as diltiazem or nifedipine. The increase in Ca²⁺ activates Ca²⁺‐activated Cl⁻ channels (Cl_Ca). The resulting efflux of Cl⁻ through these channels causes membrane depolarization, further activating Ca_L and amplifying the initial signal. Blockers of Cl_Ca channels such as niflumic acid or DIDs can inhibit this step in the process. The increase in intracellular Ca²⁺ and the membrane depolarization due to activation of Ca_L and Cl_Ca activates large conductance, Ca²⁺‐activated K⁺ channels (BK_Ca). The efflux of K⁺ through BK_Ca channels blunts the depolarizing effects of activation of Ca_L and Cl_Ca providing a degree of negative feedback, and limiting membrane depolarization. This step in the process can be inhibited by iberiotoxin or tetraethylammonium ions (TEA). Oxygen‐induced smooth muscle depolarization can be conducted along the vessel wall by gap junctions (GJ) supporting the conduction of O₂‐induced vasoconstriction that has been observed experimentally.

Leukotrienes are synthesized from arachidonic acid released from membrane phospholipid stores by a multistep process (Peters‐Golden, 1998; Storch et al. 2015). In activated cells, FLAP presents arachidonic acid to the 5‐lipoxygenase that catalyses the first step in this reaction sequence to form an unstable epoxide, leukotriene A₄ (LTA₄). The epoxide LTA₄ is then conjugated with glutathione to form the CysLT leukotriene C₄ (LTC₄), which is converted to leukotriene D₄ (LTD₄) and leukotriene E₄ (LTE₄) (also CysLTs) by consecutive cleavage of peptides from the added glutathione. The K _m for O₂ of the 5‐lipoxygenase purified from porcine leukocytes is about 9 mmHg $P_{{\mathrm{O}}_2}$ (Ibe & Raj, 1992). Given the nuclear membrane localization of the 5‐lipoxygenase (Peters‐Golden & Brock, 2001; Woods et al. 1995) and the steep intracellular gradients in O₂ that arise from O₂ consumption by mitochondria, oxidases and oxygenases within cells (Jones, 1981), leukotriene production by intact cells will be O₂ dependent well within the physiological range (10–150 mmHg) required for a microvascular O₂ sensor (see Fig. 1). Oxygen‐dependent synthesis of leukotrienes has been reported in several in vitro systems (Paterson, 1986; Ohwada et al. 1990; Ibe & Raj, 1992; Martin et al. 1992) providing support for the hypothesis that this pathway is involved in O₂ sensing, at least in the hamster cheek pouch.

Oxygen‐dependent production of 20‐HETE

In contrast to the data from the epithelial portion of the cheek pouch, arteriolar O₂ reactivity in striated muscle appears to be mediated by 20‐HETE produced by CYP450_4A ω‐hydroxylase (Harder et al. 1996; Lombard et al. 1999; Hungerford et al. 2000; Kunert et al. 2001 a,b, 2009; Drenjancevic‐Peric et al. 2003, 2004; Marvar et al. 2007; Wang et al. 2009; Ngo et al. 2013) (Fig. 8). This pathway also has been implicated in O₂‐induced cerebral vasoconstriction in fetal sheep (Ohata et al. 2010) and contributes to O₂‐induced constriction of retinal arterioles (Zhu et al. 1998) produced by increases in inspired $P_{{\mathrm{O}}_2}$. Importantly, inhibitors of CYP450_4A ω‐hydroxylase have no effect on the O₂ reactivity of arterioles in the epithelial portion of the hamster cheek pouch (Lombard et al. 1999) indicating that the inhibition observed in striated muscle preparations is not simply some non‐specific effect. The lack of effect of CYP450_4A ω‐hydroxylase inhibitors on O₂ reactivity in the epithelial portion of the cheek pouch also further supports the idea that there are regional differences in the mechanism of action of O₂. The K _m for O₂ for formation of 20‐HETE by CYP450_4A ω‐hydroxylase by renal microvessels is in the order of 50 mmHg $P_{{\mathrm{O}}_2}$ (Harder et al. 1996), well within the physiological $P_{{\mathrm{O}}_2}$ range experienced by the microcirculation (see Fig. 1).

Figure 8. The site and mechanism of action of O₂ in cremaster muscle .

Schematic diagram of a striated muscle fibre (the proposed sensor cell in this tissue) and a smooth muscle cell replete with ion channels that may be involved in arteriolar O₂ reactivity in this tissue. Elevated $P_{{\mathrm{O}}_2}$ is sensed by cytochrome P450_4A/4F ω‐hydroxylase (CYP4A/4F) located in the endoplasmic reticulum of striated muscle fibres, resulting in conversion of arachidonic acid into 20‐HETE, a process that can be inhibited by 17‐ODYA or DDMS. 20‐HETE then acts on smooth muscle cells to induce Ca²⁺ influx through L‐type Ca²⁺ channels (Ca_L), an increase in intracellular Ca²⁺ and vasoconstriction. As indicated by the ‘?’ next to the arrow connecting 20‐HETE and the smooth muscle cell, the precise receptor for 20‐HETE that is responsible for O₂ reactivity is unclear because 20‐HETE has been proposed to close large conductance, Ca²⁺‐activated K⁺ channels (BK_Ca), which would lead to membrane depolarization activating Ca_L. In contrast, other studies suggest that BK_Ca serve a negative feedback role as they do in the cheek pouch. As in the cheek pouch it is proposed that O₂‐induced depolarization of smooth muscle cells can be conducted along the vessel wall through gap junctions (GJ), consistent with the observed conducted vasoconstriction that has been observed experimentally. 6(Z),15(Z)‐20‐HEDE, 20‐hydroxy‐6Z,15Z‐eicosadienoic acid.

Studies in rat cremaster muscle have demonstrated expression of CYP450_4A ω‐hydroxylase in both arteriolar smooth muscle cells and in the surrounding striated muscle (Kunert et al. 2001 a). This observation suggests that O₂ may be sensed either by smooth muscle cells or by striated muscle fibres to mediate arteriolar O₂ reactivity. In first‐order arterioles within the cremaster muscle, it appears that striated muscle cells serve as an important source of 20‐HETE that is responsible for the bulk of O₂ reactivity in those large arterioles in situ (Frisbee & Lombard, 2002). Isolated first‐order rat cremaster arterioles studied ex vivo were demonstrated to display intrinsic O₂ reactivity that is also mediated, in part, by O₂‐dependent production of 20‐HETE by the smooth muscle (Frisbee & Lombard, 2002). However, this is not a universal finding as other investigators have failed to demonstrate intrinsic O₂ reactivity in isolated, cannulated first‐order arterioles from the rat (Tateishi & Faber, 1995 a,b; Kerkhof et al. 1999), as noted above. This is in contrast to the large body of evidence from intravital preparations demonstrating 20‐HETE‐mediated O₂ reactivity in small arterioles in a variety of striated muscle microvascular preparations (see references above).

Thus, it seems likely that striated muscle fibres and the CYP450_4A ω‐hydroxylase contained within are the site where changes in $P_{{\mathrm{O}}_2}$ are primarily sensed, with 20‐HETE serving as the mediator linking changes in $P_{{\mathrm{O}}_2}$ with changes in arteriolar tone in the microcirculation of striated muscle. Additional experiments designed to critically test the hypothesis that vascular smooth muscle cell CYP450_4A ω‐hydroxylase contributes to arteriolar O₂ reactivity in situ are needed; perhaps conditional, cell‐specific knock‐out of this enzyme family in smooth muscle cells might help to resolve this issue. In addition, the location of CYP450_4A ω‐hydroxylase in striated muscle fibres (and smooth muscle cells) favours O₂ sensing in the muscle fibres rather than capillaries downstream from the arterioles as has been proposed (Ghonaim et al. 2011). Additional experiments will be required to positively identify where O₂ is sensed in striated muscle to mediate arteriolar O₂ reactivity.

While there is considerable evidence that CYP450_4A ω‐hydroxylase serves as the main O₂ sensor mediating arteriolar O₂ reactivity in the microcirculation of striated muscle, this system is not immutable. For example, the CYP450_4A ω‐hydroxylase antagonist N‐methylsulfonyl‐12,12‐dibromododec‐11‐enamide (DDMS) (Kroetz & Xu, 2005) reduces arteriolar O₂ reactivity in cremaster muscles of Wistar–Kyoto (WKY) rats and spontaneously hypertensive rats (SHR) with established hypertension (Kunert et al. 2001 b). However, DDMS is without significant effect on O₂ reactivity during the development of hypertension in the SHR (Kunert et al. 2001 b). Similarly, DDMS inhibited O₂ reactivity in cremaster arterioles in sham‐operated controls and in rats with established Goldblatt hypertension, but was without effect during the early phase of hypertension development in this model (Kunert et al. 2009). Spinotrapezius arterioles of Sprague–Dawley rats fed a high salt diet lose CYP450_4A ω‐hydroxylase‐mediated O₂ reactivity, with no change in CYP450_4A ω‐hydroxylase protein expression (Marvar et al. 2007). Conversely, in Dahl salt‐sensitive rats, high salt diet increases arteriolar O₂ reactivity and the sensitivity to inhibition by DDMS that is associated with an increased expression of CYP450_4A ω‐hydroxylase isoforms (Wang et al. 2009). These data indicate that the mechanism of action of O₂ on arterioles can change and support the idea that there are multiple mechanisms coupling changes in $P_{{\mathrm{O}}_2}$ with changes in arteriolar diameter. The fact that O₂ reactivity remains but the mechanism of action of O₂ can change suggests that this is an important process that is ‘protected’ by mechanism redundancy.

As demonstrated by Kerkhof et al. (1999), there may be substantial interactions among signalling pathways that complicate interpretation of data based on inhibition of single pathways. While such interactions have been explored in isolated vessels ex vivo (Kerkhof et al. 1999; Frisbee et al. 2001 a,b, 2002), there are no similar studies examining the interaction among all of the putative O₂‐related signalling pathways in small arterioles in situ in the intact microcirculation.

Ionic basis of arteriolar O₂ reactivity

Despite the differences in chemical mediators of O₂ reactivity between the hamster cheek pouch and striated muscle preparations (CysLTs vs. 20‐HETE, respectively), a common downstream element in the signalling cascade appears to be activation of L‐type Ca²⁺ channels. Blockers of these channels eliminate O₂ reactivity in both the cheek pouch and cremaster muscle models (Welsh et al. 1998; Jackson, 2012; Ngo et al. 2013) (Figs 7 and 8). However, the signalling pathways leading to activation of these Ca²⁺ channels may be different in the two tissues. In the hamster cheek pouch in situ, elevated $P_{{\mathrm{O}}_2}$ causes arteriolar smooth muscle cell depolarization (Welsh et al. 1998) and a subsequent increase in intracellular Ca²⁺ (Brekke et al. 2006). Consistent with a role for CysLTs in arteriolar O₂ reactivity, we have found that LTD₄ depolarizes smooth muscle cells, constricts isolated, cannulated hamster cheek pouch arterioles and induces conducted vasoconstriction of arterioles in cheek pouches in situ (Jackson & Segal, 2005). Blockade of L‐type Ca²⁺ channels in situ obliterates both O₂‐induced constriction and depolarization (Welsh et al. 1998). There are at least two, non‐mutually exclusive mechanisms that could be responsible for the O₂‐dependent depolarization in the cheek pouch and that are consistent with the findings that L‐type Ca²⁺ channel blockers prevent both constriction and depolarization of the arteriolar muscle cells (Welsh et al. 1998). First, it is possible that the inward current of Ca²⁺ through L‐type Ca²⁺ channels causes the depolarization. However, this seems unlikely given the high resting K⁺ permeability of vascular muscle cells (Jackson et al. 1997, 1998, 2000 c), the density of voltage‐gated K⁺ (K_V) and BK_Ca channels (which will activate as the membrane is depolarized) (Jackson et al. 1997, 1998, 2000 c), and the small currents through L‐type Ca²⁺ channels at physiological membrane potentials and Ca²⁺ gradients (Nelson et al. 1990; Gollasch et al. 1992). More likely, this depolarizing Ca²⁺ signal must be amplified. Calcium‐activated Cl⁻ (Cl_Ca) channels could provide a means of amplifying the Ca²⁺ signal and promoting depolarization. These channels are activated by elevation of intracellular Ca²⁺ and, given the electrochemical gradient for Cl⁻ ions in most vascular muscle cells, result in Cl⁻ efflux from the cells and hence depolarization (Large & Wang, 1996). Calcium‐activated Cl⁻ channels mediate depolarization associated with vasoconstrictor‐induced contraction of vascular muscle in other systems (Large & Wang, 1996). Preliminary studies showed that O₂‐induced constriction and membrane depolarization of smooth muscle cells in hamster cheek pouch arterioles could be reversed by the Cl_Ca channel blockers niflumic acid and 4,4′‐diisothiocyano2,2′‐stilbenedisulfonic acid (DIDS) (Jackson, 2000 a), supporting this hypothesis. Additional research will be needed to fill in the gaps in this signalling pathway and to exclude alternative hypotheses.

The details of the signalling pathway by which elevated $P_{{\mathrm{O}}_2}$ causes constriction of arterioles in striated muscle, between 20‐HETE and activation of L‐type Ca²⁺ channels, remain unclear (Fig. 8). It has been proposed that O₂‐induced constriction of arterioles in rat spinotrapezius muscle involves 20‐HETE‐dependent closure of BK_Ca channels, membrane depolarization and subsequent opening of L‐type Ca²⁺ channels (Marvar et al. 2007). In preparations exposed to low $P_{{\mathrm{O}}_2}$, they found that the BK_Ca channel blocker, iberiotoxin, produced substantial vasoconstriction that was similar in magnitude to that produced by elevation of $P_{{\mathrm{O}}_2}$ (Marvar et al. 2007). Iberiotoxin blunted both 20‐HETE‐ and O₂‐induced vasoconstriction, while arteriolar constriction induced by noradrenaline remained intact. Studies of isolated rat gracilis arteries ex vivo also support this signalling pathway (Frisbee et al. 2001 a). In contrast, it was previously shown in hamster cremaster muscle that arteriolar BK_Ca channels are silent at rest under low $P_{{\mathrm{O}}_2}$ conditions and that inhibition of these channels potentiated, rather than inhibited, O₂‐induced vasoconstriction (Jackson, 1998) even though O₂ reactivity in this preparation is also mediated by CYP450_4A ω‐hydroxylase (Lombard et al. 1999). These data suggest that in the hamster microcirculation, BK_Ca channels participate in the negative feedback regulation of arteriolar tone and actually blunt O₂‐induced vasoconstriction (Jackson & Blair, 1998). Blockade of BK_Ca channels also potentiates O₂‐induced constriction of hamster cheek pouch arterioles (W. F. Jackson, personal observation) (Fig. 7). Thus, in hamster cremaster muscle and in the cheek pouch, BK_Ca channels appear to lie downstream from L‐type Ca²⁺ channels in the O₂‐dependent signalling cascade, rather than upstream as was proposed by Marvar et al. (2007). In addition to inhibiting BK_Ca channels (Zou et al. 1996), 20‐HETE augments currents through L‐type Ca²⁺ channels (Gebremedhin et al. 1998), possibly through activation of protein kinase C (Lange et al. 1997) or tyrosine kinases (Sun et al. 1999). In porcine coronary resistance arteries, 20‐HETE‐induced vasoconstriction appears to involve activation of Rho kinase (Randriamboavonjy et al. 2003). Thus, alternative signalling pathways are possible that could reconcile these data. Additional research in which both membrane potential and diameter are measured before and after blockade of L‐type Ca²⁺ channels, BK_Ca channels and/or CYP450_4A ω‐hydroxylase in striated muscle arterioles, in situ, will be required to resolve these apparent differences. It is also not known if 20‐HETE can induce conducted vasoconstriction, as does O₂ in striated muscle arterioles. Nonetheless, as stated above, the observation that O₂‐induced vasomotor responses can be conducted in mouse cremaster muscle (Ngo et al. 2010; Riemann et al. 2011) strongly suggests that changes in arteriolar smooth muscle membrane potential are part of the mechanism of action of O₂ in this tissue and presumably other striated muscle beds.

Summary and conclusions

The location of the cell type responsible for arteriolar O₂ reactivity, in situ, remains unclear, because there is regional heterogeneity in the site and mechanism of action of O₂ in the peripheral microcirculation. As a target for future investigation, it is assumed that the models shown in Figs 7 and 8 apply to arteriolar O₂ reactivity over the tissue $P_{{\mathrm{O}}_2}$ range of 10–70 mmHg for tissues such as the hamster cheek pouch (Fig. 7), or striated muscle preparations such as the hamster, rat or mouse cremaster muscles (Fig. 8). Based largely (but not exclusively) on studies in the hamster cheek pouch, it is proposed that changes in $P_{{\mathrm{O}}_2}$ are sensed, extravascularly, by O₂‐dependent enzymes located in the membrane of the nucleus or endoplasmic reticulum in parenchymal or other cell types (e.g. mast cells in the case of the hamster cheek pouch). This results in the O₂‐dependent production of a vasoconstrictor (CysLTs in hamster cheek pouch or 20‐HETE in striated muscle) that acts on arteriolar smooth muscle cells in the wall of arterioles. The vasoconstrictors produce membrane depolarization and activation of L‐type, voltage‐gated Ca²⁺ channels, leading to an increase in intracellular Ca²⁺ and local vasoconstriction at the site of action of the mediator. The membrane depolarization is conducted along the vessel via gap junctions to produce conducted vasoconstriction along the arteriole allowing for summation and integration of O₂‐generated signals throughout the microvascular network. In the hamster cheek pouch, vasoconstrictors that depolarize smooth muscle cells produce conducted constriction transmitted along the smooth muscle cell layer (Welsh & Segal, 1998; Bartlett & Segal, 2000). These data suggest that, in this tissue, O₂‐induced vasomotor effects are initiated near smooth muscle‐containing arterioles, rather than in downstream capillaries or venules, to provide an arteriolar smooth muscle pathway for conduction of the O₂‐initiated response. A similar scenario is assumed in striated muscle with the understanding that the tremendous length of striated muscle fibres (relative to the length of arterioles, capillaries and venules) may also contribute to transmission of O₂‐induced signals within a microvascular network. However, it is also recognized that additional research is required to exclude alternatives in the microcirculation of striated muscle. It should also be noted that detailed studies of arteriolar O₂ reactivity have only been studied in rodents (hamsters, mice and rats). Whether similar mechanisms are functional in larger mammals (including humans) remains to be established.

While there is a plethora of ex vivo studies suggesting that smooth muscle cells, endothelial cells, red blood cells or some combination contribute to O₂ reactivity, it is not clear that these ex vivo findings accurately represent the arteriolar O₂ reactivity that is consistently observed, in situ. This may partly be due to the use of relatively large vessels (small arteries and larger arterioles) for the majority of ex vivo studies. Given the regional heterogeneity of mechanisms modulating myogenic tone between arteries and arterioles (Westcott & Jackson, 2011; Westcott et al. 2012), it seems likely that different mechanisms, and hence different O₂ sensors and mechanisms, may exist for sensing and responding to changes in $P_{{\mathrm{O}}_2}$ in different segments of the vascular tree. Resistance arteries upstream from the microcirculation may be tuned to respond primarily to changes in blood $P_{{\mathrm{O}}_2}$ (systemic hypoxia/hyperoxia), whereas downstream arterioles may be more tuned to respond to changes in local tissue $P_{{\mathrm{O}}_2}$.

It is also possible that different mechanisms operate over different $P_{{\mathrm{O}}_2}$ ranges, as suggested by the ex vivo studies of Frisbee et al. (Frisbee et al. 2002). Most studies have examined responses of vessels to relatively large changes in $P_{{\mathrm{O}}_2}$ and have not carefully examined the concentration–response relationship between O₂ and vessel tone. Nor have detailed concentration–response relationships been performed for antagonists and inhibitors to define their efficacy and potency on O₂ reactivity. Thus, when a single concentration of an inhibitor is used and only partial inhibition is observed, is this because there are multiple mechanisms involved, or because of a lack of efficacy of the drug on its target? Interactions among signalling pathways (Kerkhof et al. 1999) also complicate the interpretation of inhibitor data. Signalling pathway interactions have not been adequately explored in the O₂ reactivity of small arterioles in situ. Conduction of O₂‐induced vasomotor responses also complicates the study of arteriolar O₂ reactivity, and should be further examined.

It is clear that there are regional differences in the mechanism of action of O₂ on arterioles in the peripheral microcirculation (e.g. hamster cheek pouch epithelium vs. hamster striated muscle). However, there is a remarkably consistent result: elevated $P_{{\mathrm{O}}_2}$ in the environment of small arterioles in situ results in profound vasoconstriction. Oxygen‐induced arteriolar constriction may be a protective mechanism to defend against oxidative stress and excessive production of O₂‐derived reactive oxygen species (ROS) as suggested by Golub and Pittman (Golub & Pittman, 2013). Oxygen‐induced arteriolar constriction may be part of a negative feedback mechanism to limit tissue exposure to O₂ to help reduce the oxidative burden that is superimposed on the mechanisms that are involved in the regulation of blood flow to meet the metabolic demands of tissues. That different tissues have different mechanisms (e.g. cheek pouch vs. striated muscle) to accomplish the same end‐result (O₂‐induced arteriolar constriction) suggests that this may be an important response to maintain homeostasis.

Additional information

Competing interests

None declared.

Funding

The author is supported by the National Heart, Lung and Blood Institute of the National Institutes of Health (NIH) under award numbers RO1 HL32469 and P01 HL070687. The content is solely the responsibility of the author and does not necessarily represent the official views of the NIH.

Biography

William F. Jackson is Professor of Pharmacology and Toxicology at Michigan State University. He has had a long‐term interest in identifying the site in tissues where changes in oxygen tension are sensed and how such changes alter the contractile function of arteriolar muscle cells. His research focuses on identifying the ion channels expressed by both arteriolar smooth muscle and endothelial cells, how these ion channels function in the regulation of myogenic tone and oxygen signalling in the microcirculation, and how ageing and disease states such as hypertension and obesity impact the expression and function of ion channels in the microcirculation.