AMP-activated protein kinase and the regulation of Ca²⁺ signalling in O₂-sensing cells

Abstract

All cells respond to metabolic stress. However, a variety of specialized cells, commonly referred to as O₂-sensing cells, are acutely sensitive to relatively small changes in P_O2. Within a variety of organisms such O₂-sensing cells have evolved as vital homeostatic mechanisms that monitor O₂ supply and alter respiratory and circulatory function, as well as the capacity of the blood to transport O₂. Thereby, arterial P_O2 may be maintained within physiological limits. In mammals, for example, two key tissues that contribute to this process are the pulmonary arteries and the carotid bodies. Constriction of pulmonary arteries by hypoxia optimizes ventilation–perfusion matching in the lung, whilst carotid body excitation by hypoxia initiates corrective changes in breathing patterns via increased sensory afferent discharge to the brain stem. Despite extensive investigation, the precise mechanism(s) by which hypoxia mediates these responses has remained elusive. It is clear, however, that hypoxia inhibits mitochondrial function in O₂-sensing cells over a range of P_O2 that has no such effect on other cell types. This raised the possibility that AMP-activated protein kinase might function to couple mitochondrial oxidative phosphorylation to Ca²⁺ signalling mechanisms in O₂-sensing cells and thereby underpin pulmonary artery constriction and carotid body excitation by hypoxia. Our recent investigations have provided significant evidence in support of this view.

Introduction

Although the mechanism of chemotransduction by hypoxia is keenly debated, a general consensus is developing as to the processes that underpin pulmonary artery constriction and carotid body excitation by hypoxia, respectively.

Hypoxic pulmonary vasoconstriction (HPV) is the critical and distinguishing characteristic of the arteries that feed the lung (von Euler & Liljestrand, 1946). In marked contrast, systemic arteries dilate in response to hypoxia to meet the metabolic demands of the tissues they supply (Roy & Sherrington, 1890). Physiologically, HPV contributes to ventilatation–perfusion matching in the lung by diverting blood flow to O₂-rich areas (Fishman, 1976; Voelkel, 1986). However, when alveolar hypoxia is global, as in diseases such as emphysema and cystic fibrosis, HPV leads to hypoxic pulmonary hypertension (HPH) and right heart failure (Voelkel, 1986). HPV is driven by the response to hypoxia of two different cell types, namely the pulmonary arterial smooth muscle and endothelial cells.

HPV had been presumed to be mediated by inhibition of voltage-gated K⁺ channel (Kv) currents, subsequent membrane depolarization and voltage-gated Ca²⁺ influx (Michelakis et al. 2004). However, it is now clear that constriction by hypoxia of pulmonary artery rings may be induced in the absence of extracellular Ca²⁺, i.e. under conditions where voltage-gated influx is abolished (Dipp et al. 2001). Consistent with this fact, a growing body of evidence now suggests that acute HPV is primarily initiated and maintained by the release of Ca²⁺ from smooth muscle sarcoplasmic reticulum (SR) stores via ryanodine receptors (RyRs; Salvaterra & Goldman, 1993; Jabr et al. 1997; Dipp & Evans, 2001; Dipp et al. 2001; Morio & McMurtry, 2002) and by consequent Ca²⁺ influx via a store-refilling current (Kang et al. 2003; Ng et al. 2005; Wang et al. 2005). This process of SR Ca²⁺ release by hypoxia is promoted and maintained by cyclic adenosine diphosphate ribose (cADPR; (Dipp & Evans, 2001; Wilson et al. 2001; Aley et al. 2005, 2006), a Ca²⁺ mobilizing pyridine nucleotide that activates RyRs (Lee et al. 1989; Rusinko & Lee, 1989; Galione et al. 1991; Walseth et al. 1991), and is a prerequisite for the full expression of HPV (Dipp & Evans, 2001). Subsequently, smooth muscle constriction is augmented by myofilament Ca²⁺ senstization that is likely evoked by an endothelium-derived vasoconstrictor (Demiryurek et al. 1991; Kovitz et al. 1993; Robertson et al. 1995; Gaine et al. 1998; Dipp & Evans, 2001; Dipp et al. 2001; Robertson et al. 2001; Wilson et al. 2001), the release of which appears dependent on transmembrane Ca²⁺ influx into pulmonary artery endothelial cells (Dipp et al. 2001).

Carotid body excitation by hypoxia serves a discrete physiological function and is initiated by a different mechanism. Here carotid body type I cells, otherwise referred to as glomus cells, monitor systemic arterial P_O2 and initiate corrective changes in breathing patterns (Heymans et al. 1930). Upon exposure to hypoxia, voltage-gated Ca²⁺ influx into type I cells initiates neurosecretion (Fidone et al. 1988; Gonzalez et al. 1994; Chen et al. 1997) and thereby increases sensory afferent discharge to the brain stem. This process is driven by the inhibition of what have been termed O₂-sensitive K⁺ channels (Lopez-Barneo et al. 1988; Peers, 1990; Stea & Nurse, 1991; Lopez-Lopez & Gonzalez, 1992; Wyatt & Peers, 1992, 1995; Buckler, 1997, 1999; Buckler et al. 2000). Thus, membrane depolarization is triggered, and in turn activates voltage-gated Ca²⁺ channels rather than ER Ca²⁺ release (Buckler & Vaughan-Jones, 1994). The molecular identity of the O₂-sensitive K⁺ channel has, however, been contested and there are clear differences between species (Peers, 1990; Ganfornina & Lopez-Barneo, 1991, 1992; Buckler et al. 2000; Sanchez et al. 2002; Lopez-Lopez et al. 2003). Despite this fact, it is now generally accepted that hypoxia selectively inhibits TASK-like K⁺ channels (Buckler et al. 2000) and large conductance Ca²⁺-activated (BK_Ca) K⁺ channels (Wyatt & Peers, 1995) in rat carotid body type I cells and thereby triggers carotid body excitation, whilst inhibition of Kv channels is thought to underpin depolarization in mouse and rabbit carotid body type I cells (Ganfornina & Lopez-Barneo, 1992; Sanchez et al. 2002; Perez-Garcia et al. 2004).

Mitochondrial oxidative phosphorylation and the regulation by hypoxia of O₂-sensing cells

The precise mechanism(s) by which hypoxia elicits the aforementioned responses represents a more contentious issue. It has been suggested that O₂-sensitive signal transduction pathways independent of mitochondria may play a role (Youngson et al. 1993; Prabhakar, 1998; Prabhakar & Overholt, 2000; Williams et al. 2004). However, the only consistent finding is that relatively mild hypoxia inhibits mitochondrial oxidative phosphorylation and it has been suggested that this may underpin, at least in part, cell activation (Mills & Jobsis, 1972; Rounds & McMurtry, 1981; Duchen & Biscoe, 1992a, b; Wyatt & Buckler, 2004). Thus, depolarization of the mitochondrial membrane potential and/or an increase in β-NAD(P)H levels has been reported in all O₂-sensing cells studied to date (Archer et al. 1986; Duchen & Biscoe, 1992a, b; Youngson et al. 1993; Shigemori et al. 1996; Leach et al. 2001) and over a range of P_O2 that elicits no such response in other cell types (Duchen & Biscoe, 1992b). Others have argued that the affinity of cytochrome c oxidase for O₂ is too high to allow for the inhibition of mitochondrial oxidative phosphorylation by physiological hypoxia, despite the above and the fact that respiratory control by O₂ may be determined by cell- and tissue-specific differences in (1) the metabolic environment of mitochondria, (2) the O₂ affinity of cytochrome c oxidase, and (3) the dependence of the O₂ affinity of cytochrome c oxidase on metabolic state and rate (Gnaiger et al. 1998). Such controversy has been fuelled by the fact that the mechanism by which inhibition of mitochondrial oxidative phosphorylation may couple to discrete and cell-specific Ca²⁺ signalling mechanisms has remained elusive. Previous investigations have focused on the possible role of the cellular energy status (ATP; Rounds & McMurtry, 1981; Buescher et al. 1991; Leach et al. 1998, 2000), reduced redox couples (Archer et al. 1993; Wilson et al. 2001) and reactive O₂ species (Killilea et al. 2000; Leach et al. 2001; Waypa et al. 2001; Prabhakar & Kumar, 2004), respectively, but extensive investigation of these hypotheses has delivered conflicting data and failed to unite the field since its inception in 1930 (Gonzalez et al. 2002; He et al. 2005). However, previous assessments of the role of the cellular energy status may have been limited by knowledge of the identity of the energy variable that might signal the response (Buescher et al. 1991); significantly, the role of the AMP/ATP ratio had not been considered until recently.

Regulation of AMP-activated protein kinase by metabolic stresses

Over the past decade AMP-activated protein kinase (AMPK), a serine threonine kinase comprising a catalytic α subunit and regulatory β and γ subunits, has come to prominence as a metabolic fuel gauge which monitors the cellular AMP/ATP ratio as an index of metabolic stress. AMPK is activated in response to a variety of metabolic stresses that either increase cellular ATP consumption or reduce ATP supply via mitochondrial oxidative phosphorylation, for example heat shock and metabolic poisons in hepatocytes (Corton et al. 1995; Shaw et al. 2005), exercise in skeletal muscle (Winder & Hardie, 1996; Sakamoto et al. 2005, 2006) and ischaemia in the heart (Marsin et al. 2000; Sakamoto et al. 2006). Activation of AMPK is consequent to the action of adenylate kinase, which catalyses the reaction that converts 2 molecules of ADP to ATP + AMP in order to counter, in the short term, any reduction in ATP supply. Consequently, the cellular ADP/ATP ratio is converted into a much larger increase in the AMP/ATP ratio, which varies as the square of the ADP/ATP ratio (Hardie & Hawley, 2001). Subsequent binding of AMP to two Bateman domains on the γ subunit of AMPK (Hawley et al. 1995; Scott et al. 2002; Kemp, 2004) triggers activation of the kinase by: (1) allosteric regulation via the γ subunit, (2) permitting phosphorylation of the α subunit at Thr-172 by an upstream kinase that is a complex between the tumour suppressor kinase LKB1 and two accessory proteins STRAD and MO25 (Hawley et al. 2003; Woods et al. 2003; Shaw et al. 2004, 2005), and (3) inhibiting dephosphorylation of AMPK. In the absence of metabolic stress, each of these processes is antagonized by high concentrations of ATP, for which the Bateman domains on the γ subunit have lower affinity than they do for AMP. Thus, AMPK is regulated by a triple mechanism that is exquisitely sensitive to very small changes in the AMP/ATP ratio (Hardie & Hawley, 2001). It should be noted, however, that other upstream kinases may regulate AMPK. For example, recent studies have shown that calmodulin-dependent kinase kinases, especially CaMKKβ, may also activate AMPK via the phosphorylation of Thr-172. This does not require an increase in the AMP/ATP ratio, but is instead promoted by a rise in intracellular Ca²⁺ concentration (Anderson et al. 1998; Hawley et al. 2005; Hurley et al. 2005; Woods et al. 2005). It is possible therefore that CaMKKβ may provide positive feedback in response to the rise in intracellular Ca²⁺ induced by AMPK in O₂-sensing cells. In this respect it may be significant that expression of the two CaMKK isoforms (α and β) is mainly restricted to cells of neural origin (Anderson et al. 1998), such as carotid body type I cells.

Upon activation, AMPK promotes catabolic pathways in order to maintain ATP supply, whilst switching off non-essential ATP-consuming (anabolic) pathways (see Hardie et al. 2006). Thus, the primary targets for AMPK had been presumed to be genes and proteins involved in energy metabolism, particularly lipid and carbohydrate metabolism, and little consideration was given to the role of AMPK as a mediator of physiological processes, despite the fact that AMPK is a serine threonine kinase. However, it is now recognized that AMPK can also target non-metabolic processes (Hardie, 2005), and hence the focus of this issue of The Journal of Physiology. In this respect it is important to note that while AMPK is ubiquitously expressed throughout eukaryotic cells, at least 12 different heterotrimers may be formed from multiple isoforms (Cheung et al. 2000) of the catalytic α subunit (α1 and α2) and regulatory β (β1 and β2) and γ (γ1–3) subunits, splice variants of which may add to the diversity. Thus, the selective expression of a particular AMPK isozyme(s) could determine, at least in part, both cell-and system-specific responses to metabolic stresses. Taking the above into account and the fact that inhibition of mitochondrial oxidative phosphorylation by hypoxia will promote a rise in the AMP/ATP ratio, I proposed that AMPK might act as a mediator of chemotransduction by hypoxia in O₂-sensing cells (Evans, 2004). This was a teleological argument based on the premise that (1) the evolution of complex physiological systems would utilize successful processes in different ways as long as they were fit for purpose, and (2) the observation that discrete populations of cells have evolved to serve as O₂ sensors in a variety of species. For example, in common with its effects on mammalian pulmonary arteries and carotid body type I cells, hypoxia constricts systemic arteries that lead from the gills of cyclostomes such as the Pacific hagfish (Olson et al. 2001), constricts arteries that feed the skin of the amphibia (Malvin & Walker, 2001), and activates neuroepithelial cells in the gill of teleosts (Jonz et al. 2004).

Is a specific AMPK subunit isoform combination expressed in pulmonary arterial smooth muscle?

In 1946, von Euler and Liljestrand demonstrated that hypoxia without hypercapnia induced constriction within the pulmonary circulation and proposed that HPV might aid ventilation–perfusion matching, by diverting blood flow away from poorly ventilated areas of the lung (von Euler & Liljestrand, 1946). This is now recognized as the critical and distinguishing characteristic of pulmonary arteries; systemic arteries dilate in response to tissue hypoxaemia (Roy & Sherrington, 1890). Clearly therefore one would expect a mediator of HPV to offer in some way the pulmonary selectivity required to elicit this response. One determining factor could be the nature of the AMPK isozyme(s) expressed in pulmonary arterial smooth muscle. Thus, we characterized the subunit isoforms present. These investigations showed that the α1 and α2 catalytic subunit isoforms and the β2, γ1, γ2 and γ3 regulatory subunit isoforms are expressed in smooth muscle from the second and third order branches of the pulmonary arterial tree, although the γ3 isoform accounts for such a small fraction of the total AMPK activity that its functional relevance is open to question. Thus, it is most likely that four AMPK subunit isoform combinations contribute to the regulation of pulmonary arterial smooth muscle, namely α1β2γ1, α1β2γ2, α2β2γ1 and α2β2γ2. However, the catalytic activity was mainly accounted for by the α1 subunit (> 80%) whilst 60% was associated with the regulatory γ1 subunit, consistent with the α1β2γ1 heterotrimer being predominant (Evans et al. 2005, 2006a, b). Intriguingly, the level of AMPK activities in the second and third order branches of the pulmonary arterial tree contrasted markedly with the activities in the main pulmonary artery that feeds the lung and systemic (mesenteric) arteries. Briefly, in smooth muscle from the main pulmonary artery the activity associated with the α1 catalytic subunit isoform alone declined by 50%. Thus, AMPK-α1 activity is inversely related to pulmonary artery diameter, as is the magnitude of pulmonary artery constriction by hypoxia (Kato & Staub, 1966) and, significantly, the enzyme activities for the synthesis and metabolism of cADPR, a primary trigger for SR Ca²⁺ release by hypoxia and HPV (Dipp & Evans, 2001; Wilson et al. 2001). Most importantly, perhaps, AMPK-α1 activity was at least fourfold higher in second and third order branches of the pulmonary arterial tree when compared with systemic (mesenteric) arteries, which dilate rather than constrict in response to hypoxia. Once more this correlated with the distribution of the enzyme activities for the synthesis and metabolism of cADPR. Thus, the differential distribution of AMPK-α1 catalytic activity in arterial smooth muscle together with that of the enzyme activities for the synthesis and metabolism of cADPR could provide, via signal amplification, the degree of pulmonary selectivity required for HPV. Furthermore, the predominance of the α1β2γ1 heterotrimer in smooth muscle from second and third order branches of the pulmonary arterial tree suggests that this isozyme may play a prominent role in determining the response of pulmonary arteries to hypoxia.

Does hypoxia increase the AMP/ATP ratio and thereby activate AMPK-α1 activity in pulmonary arterial smooth muscle?

If inhibition of mitochondrial oxidative phosphorylation and consequent activation of AMPK were to mediate chemotransduction by hypoxia, we would expect a prerequisite increase in the AMP/ATP ratio on exposure of pulmonary arterial smooth muscle to hypoxia. This is exemplified by previous studies on the action of adenylate kinase, which have shown that any increase in the cellular ADP/ATP ratio is converted into a rise in the AMP/ATP ratio and consequent activation of AMPK. Consistent with this fact, we found that the AMP/ATP ratio in pulmonary arterial smooth muscle rose twofold under hypoxic conditions (16–21 mmHg; Evans et al. 2005). As one would expect (Hardie & Hawley, 2001; Hardie, 2003) the adenylate kinase reaction appeared close to equilibrium under normoxia and hypoxia, because the AMP/ATP ratio varied approximately as the square of the ADP/ATP ratio. Perhaps most significantly, the rise in the AMP/ATP ratio was associated with a concomitant, 2-fold increase in AMPK activity and phosphorylation of a classical AMPK substrate, acetyl CoA carboxylase. Notably, acute hypoxia was found to increase AMPK activity associated with the α1 catalytic subunit isoform to a greater extent than that associated with the α2 catalytic subunit isoform. This finding, when combined with the fact that AMPK-α1 activity was at least fourfold higher in pulmonary versus systemic arterial smooth muscle, provides further support for the view that the AMPK-α1 associated catalytic activity may play a prominent role in the response of pulmonary arterial smooth muscle cells to hypoxia.

AMPK, hypoxia and the regulation of the metabolic state in pulmonary versus systemic arteries

Previous investigations have shown that the energy state of pulmonary arterial smooth muscle is lower than that of systemic arteries, as defined by the phosphocreatine (PCr)/ATP and P_i/ATP ratios (Buescher et al. 1991; Leach et al. 1998, 2000). Upon exposure to hypoxia a fall in PCr levels and intracellular pH (pH_i) was observed, whilst ATP levels were not altered significantly. These findings are significant because a lower energy state indicates that the resting AMP/ATP ratio may be higher in pulmonary arterial smooth muscle under normoxic conditions when compared to systemic arterial smooth muscle. The combined effect of variations in these metabolic markers could therefore be to provide a lower threshold for activation of AMPK upon exposure of pulmonary arteries to hypoxia. Consequently, a given level of hypoxia may facilitate AMPK-dependent signalling mechanisms to a greater extent in pulmonary when compared to systemic arterial smooth muscle, and the margin of differential signalling may be increased still further by the greater than fourfold higher levels of AMPK-α1 activity in pulmonary arterial smooth muscle.

It is also notable that upon exposure to hypoxia metabolic markers other than ATP (pH_i, phosphocreatine) decline to a lesser degree in pulmonary arteries when compared to systemic arteries (Leach et al. 1998, 2000). This led to the suggestion that pulmonary arterial smooth muscle contains specific mechanisms that allow for better maintenance of the cellular energy state. Clearly, the ability of pulmonary arterial smooth muscle to maintain the energy state and phosphorylation potential could be conferred, at least in part, by the at least fourfold higher levels of AMPK-α1 activity relative to that found in systemic arterial smooth muscle, and by a lower threshold for AMPK activation by hypoxia. Indirect support for this view may be taken from previous investigations on pulmonary arterial smooth muscle, which have suggested that glucose uptake (governed by GLUT 4 in arterial smooth muscle; Banz et al. 1996) and glycolysis are accelerated by hypoxia, and that glycolysis plays a vital role not only in maintaining the energy status, but in maintaining pulmonary artery constriction by hypoxia (Ohe et al. 1986; Wiener & Sylvester, 1991; Zhao et al. 1996; Leach et al. 2000, 2001). This is because the primary role of AMPK in eukaryotic cells is to maintain ATP supply by activating catabolic processes and by inhibiting non-essential ATP consuming processes. Significantly, this is achieved in part by (1) accelerating glycolysis, (2) increasing glucose transport via the translocation of glucose transporters (including GLUT4) to the plasma membrane, and (3) inactivation of acetyl CoA carboxylase and consequent acceleration of β-oxidation of fatty acids by mitochondria (for further details see Hardie et al. 2006).

On the basis of the ability of pulmonary arterial smooth muscle to maintain the energy state and phosphorylation potential under hypoxia, it has also been argued that HPV is unlikely to be mediated by changes in metabolic variables (Buescher et al. 1991; Leach et al. 1998, 2000). It is vitally important therefore to recognize that AMPK is activated in response to a rise in the AMP/ATP ratio and that this can occur without a measurable fall in the cellular ATP levels. Furthermore, the maintenance of the phosphorylation potential will support the regulation of processes by AMPK-dependent phosphorylation of target proteins, such as those that contribute to stimulus–contraction coupling by hypoxia.

Does AMPK activation mobilize ryanodine-sensitive sarcoplasmic reticulum Ca²⁺ stores in pulmonary arterial smooth muscle?

Our investigations on the cellular distribution of the AMPK-α1 subunit isoform in isolated pulmonary arterial smooth muscle cells revealed that associated catalytic activity is likely to be targeted to sites throughout the cytoplasm, whilst being excluded from the nucleus and from the plasma membrane (Fig. 1A). Thus, a specific subunit isoform combination may target AMPK activity to SR compartments in order to support a role in the regulation of Ca²⁺ mobilization by AMPK in response to hypoxia. To determine whether or not AMPK did indeed mediate Ca²⁺ signalling by hypoxia, we employed the now classical method of studying AMPK activation. This requires the use of two drugs, phenformin and AICAR (5-aminoimidazole-4-carboxamide riboside), each of which activate AMPK via discrete mechanisms. Phenformin (a drug formerly used in the treatment of type 2 diabetes) inhibits Complex I of the mitochondrial respiratory chain (El-Mir et al. 2000; Owen et al. 2000) and thereby activates AMPK by increasing the cellular AMP/ATP ratio (Hawley et al. 2005). By contrast, AICAR is metabolized to yield the AMP mimetic ZMP (AICAR monophosphate) and thereby selectively activates AMPK without affecting the cellular AMP/ATP ratio (Corton et al. 1995; Owen et al. 2000). Despite their different modes of action, hypoxia (16–21 Torr), phenformin and AICAR activated AMPK and evoked an increase in intracellular Ca²⁺ concentration in isolated pulmonary arterial smooth muscle cells. In each case the increase in intracellular Ca²⁺ concentration was resistant to removal of extracellular Ca²⁺ (Dipp et al. 2001; Evans et al. 2005). However, prior block of SR Ca²⁺ release via RyRs by preincubation of cells with ryanodine and caffeine abolished the increase in intracellular Ca²⁺ concentration induced by each stimulus. Most significantly, SR Ca²⁺ release in response to AMPK activation was also abolished upon blocking the Ca²⁺ mobilizing messenger cADPR with 8-bromo-cADPR, a selective cADPR antagonist (Sethi et al. 1997; Walseth et al. 1997). Thus, AMPK activation triggers cADPR-dependent SR Ca²⁺ release via RyRs in isolated pulmonary arterial smooth muscle cells, as does hypoxia (Dipp & Evans, 2001; Dipp et al. 2001; Wilson et al. 2001). This finding was significant, because we had previously shown that the enzyme activities for the synthesis and metabolism of cADPR confer a degree of pulmonary selectivity, that cADPR accumulation in pulmonary arterial smooth muscle is augmented by hypoxia and that cADPR-dependent SR Ca²⁺ release is a prerequisite for HPV (Dipp & Evans, 2001; Wilson et al. 2001).

Figure 1. AMPK-dependent regulation of hypoxic pulmonary vasoconstriction.

A, from left to right: brightfield image of an isolated and fixed pulmonary arterial smooth muscle cell; z section showing immuno-labelling for the AMPK-α1 catalytic subunit isoform (green) and the DAPI labelled nucleus (blue); 3D reconstruction. B, schematic diagram describing the role of AMPK in HPV: hypoxia inhibits mitochondrial oxidative phosphorylation, which leads to an increase in the AMP/ATP ratio, AMPK activation, consequent cADPR-dependent SR Ca²⁺ release in pulmonary arterial smooth muscle and HPV.

We therefore sought to determine whether or not AMPK acts as a primary mediator of HPV. To this end we carried out a detailed comparison of the effects of hypoxia and AMPK activation by AICAR on isolated pulmonary arteries. Consistent with the time course of the maintained phase of pulmonary artery constriction by hypoxia (Dipp & Evans, 2001), AICAR induced a slow, sustained and reversible constriction of pulmonary artery rings (Evans et al. 2005). Removal of the pulmonary artery endothelium reduced the constriction in response to hypoxia and AICAR by approximately 30%. Furthermore, the endothelium-dependent component of constriction by AICAR and hypoxia was abolished upon removal of extracellular Ca²⁺, and therefore requires Ca²⁺ influx into the endothelium. In contrast, constriction mediated by mechanisms intrinsic to the smooth muscle was not so abolished. However, smooth muscle constriction was abolished by blocking the mobilization of SR Ca²⁺ stores via RyRs with ryanodine and caffeine and by blocking Ca²⁺ mobilization in response to cADPR with 8-bromo-cADPR. Thus, both AMPK activation and hypoxia mediate maintained constriction of pulmonary artery smooth muscle by cADPR-dependent mobilization of SR Ca²⁺ stores via RyRs (Fig. 1B).

It should be noted, however, that maintained smooth muscle constriction by AICAR and hypoxia, respectively, exhibited a partial dependence on transmembrane Ca²⁺ influx. In this respect it is of major significance that depletion of SR Ca²⁺ stores with caffeine and ryanodine or blockade of cADPR with 8-bromo-cADPR completely abolished the constriction of pulmonary arteries, with or without endothelium, by both AICAR and hypoxia. Thus, the partial dependence of smooth muscle constriction on extracellular Ca²⁺ must be determined by SR Ca²⁺ release-activated Ca²⁺ influx, as suggested by the investigations of others (Kang et al. 2003; Ng et al. 2005; Wang et al. 2005). However, it seems likely that this process of store-release activated Ca²⁺ influx is consequent to the mobilization of SR Ca²⁺ stores rather than being directly regulated by hypoxia (Kang et al. 2003; Ng et al. 2005) or AMPK activation per se (Evans et al. 2005).

The aforementioned findings provide strong support for our assertion that AMPK activation by hypoxia underpins HPV, and our most recent studies have provided further support for this view. Thus, the AMPK antagonist compound C (Zhou et al. 2001) inhibits the increase in acetyl CoA carboxylase phosphorylation in pulmonary arterial smooth muscle evoked by hypoxia and AICAR, and inhibits both HPV and pulmonary artery constriction in response to AICAR (A. M. Evans et al. unpublished results).

Does AMPK activation elicit voltage-gated Ca²⁺ influx into the carotid body type I cell and increased sensory afferent discharge from the carotid body?

In light of our observations on the role of AMPK in pulmonary artery constriction by hypoxia, we sought to determine the contribution of AMPK to chemotransduction by hypoxia in other O₂-sensing cells, and focused our attention on the carotid body type I cell. These cells are also stimulated by hypoxia and, at least in part, via inhibition of mitochondrial oxidative phosphorylation. In this case, however, excitation is primarily mediated by voltage-gated Ca²⁺ influx, rather than by Ca²⁺ release from intracellular stores (Buckler & Vaughan-Jones, 1994). Surprisingly, immuno-fluorescence imaging revealed that the AMPK-α1 catalytic subunit isoform expressed in carotid body type I cells was almost entirely (∼75%) restricted to a volume within 1 μm of the plasma membrane (Fig. 2A; Evans et al. 2005). Thus, the spatial localization of AMPK-α1 in carotid body type I cells is consistent with it targeting plasma membrane delimited processes such as voltage-gated Ca²⁺ influx. This was contrary to our observations on pulmonary arterial smooth muscle cells, in which AMPK-α1 appeared absent from the plasma membrane and targeted to structures throughout the cytoplasm. One explanation for these conflicting observations is that AMPK-α1 forms a heterotrimer with different regulatory subunit isofoms (β and/or γ) in each of these cell types, and that the nature of the subunit combination determines the cellular compartment to which it is targeted. Consistent with this view, it has been suggested that the subunit isoform combination may determine the subcellular targeting of AMPK (Salt et al. 1998), and it has been proposed that post-translational modification of the regulatory β1 subunit isoform may target AMPK to the plasma membrane (Warden et al. 2001). The cellular distribution of AMPK may also be determined by the deployment of isozyme-specific anchoring proteins similar to the ‘A-kinase anchoring proteins’ (AKAPs) that underpin the cellular distribution of protein kinase A (Wang et al. 2006). One such candidate is plectin, a cytoskeleton linker protein that may play a role in the compartmentalization of intracellular signalling cascades (Janmey, 1998), which has been shown to bind γ1 regulatory subunits that are in combination with AMPK-α1 but not AMPK-α2 catalytic subunit isoforms, and in differentiated but not undifferentiated myotubes (Gregor et al. 2006). Thus, whilst the precise mechanisms remain to be determined it is clear that different AMPK isozymes may be targeted to discrete cellular compartments and in a cell-specific manner that may be tailored to the function of AMPK in a given cell type.

Figure 2. AMPK-dependent regulation of carotid body excitation by hypoxia.

A, from left to right: brightfield image of an isolated and fixed carotid body type I cell; z section showing immuno-labelling for the AMPK-α1 catalytic subunit isoform (green) tyrosine hydroxylase (red) and the DAPI labelled nucleus (blue); 3D reconstruction. B, schematic diagram describing the proposed mechanism of carotid body excitation by hypoxia: hypoxia inhibits mitochondrial oxidative phosphorylation in carotid body type I cells, which leads to an increase in the AMP/ATP ratio, AMPK activation, AMPK-dependent inhibition of K⁺ channels in the plasma membrane, voltage-gated Ca²⁺ influx, neurosecretion and a consequent increase in afferent fibre discharge.

Consistent with the effects of hypoxia and the distribution of AMPK-α1, AMPK activation by AICAR induced a reversible depolarization of the membrane potential of acutely isolated rat carotid body type I cells, which resulted from the selective inhibition of the O₂-sensitive K⁺ currents in these cells, namely the TASK-like leak K⁺ current and BK_Ca current (Wyatt et al. 2006). Consequently, an increase in the intracellular Ca²⁺ concentration was evoked via Ni²⁺- and Cd²⁺-sensitive voltage-gated Ca²⁺ influx pathways, which ultimately led to an increase in sensory afferent discharge from the isolated carotid body (Evans et al. 2005; Wyatt et al. 2006). Thus, AMPK is likely to mediate the excitatory effects of hypoxia on isolated carotid body type I cells and on the carotid body in vitro (Fig. 2B). Once more this view gains strong support from our most recent investigations, which have shown that the AMPK antagonist compound C reverses the increase in intracellular Ca²⁺ concentration evoked by hypoxia and AICAR in isolated carotid body type I cells and the increase in sensory afferent discharge elicited from the isolated carotid body by each of these stimuli (A.M. Evans et al. unpublished results).

Could AMPK play a role in the development of hypoxic pulmonary hypertension and the effect of chronic intermittent hypoxia on hypoxic ventilatory responses?

Chronic and intermittent hypoxia precipitate (1) changes in hypoxic ventilatory responses, (2) remodelling of the carotid body, (3) pulmonary vascular remodelling associated with the development of hypoxic pulmonary hypertension, and (4) increased erythropoietin production and consequently increased red blood cell development. The transcription factor HIF-1 has been proposed to be pivotal to each of these pathophysiological alterations (Fandrey, 2004; Prabhakar & Jacono, 2005; Semenza, 2005). HIF-1 is a heterodimer comprising α and β subunits, the activity of which may be regulated by varying the expression of HIF-1α and thereby its association with the constitutively expressed β subunit. Therefore, It may be significant that AMPK-α2, which is preferentially targeted to the nucleus (Salt et al. 1998), but not the AMPK-α1 catalytic subunit isoform may increase HIF-1α expression (Hwang et al. 2004) and play a critical role in the regulation of associated transcriptional activity and target gene expression in response to prolonged hypoxia (Lee et al. 2003; Neurath et al. 2006). This raises the possibility that heterotrimeric subunit combinations containing AMPK-α1 may underpin the regulation of primary cell function by hypoxia as described above, whilst AMPK-α2 may play a prominent role in the paradigm of O₂-dependent gene expression within O₂-sensing cells in response to chronic and intermittent hypoxia.

Summary

In summary, previous studies have established that hypoxia promotes pulmonary artery constriction and carotid body sensory afferent discharge in part by inhibiting oxidative phosphorylation by mitochondria. Our findings now suggest that inhibition of mitochondrial oxidative phosphorylation by hypoxia leads to a rise in the cellular AMP/ATP ratio, consequent AMPK activation and the initiation of cell-specific Ca²⁺ signalling mechanisms in pulmonary arterial smooth muscle cells and carotid body type I cells. Thus, the characteristic response of each tissue type to AMPK activation precisely mirrors that seen in response to hypoxia: (1) constriction of pulmonary arteries by cADPR-dependent SR Ca²⁺ release in the smooth muscle cells, with a secondary component of constriction driven by the pulmonary arterial endothelial cells, and (2) inhibition of TASK-like and BK_Ca K⁺ currents, leading to depolarization, voltage-gated Ca²⁺ influx into carotid body type I cells and a consequent increase in afferent fibre discharge. We therefore proposed that in addition to maintaining the cellular energy state, AMPK acts as the primary metabolic sensor and effector of Ca²⁺ signalling by hypoxia in O₂-sensing cells. This view unites for the first time the mitochondrial and Ca²⁺ signalling hypotheses for chemotransduction by hypoxia. Our data also suggest that high levels of AMPK-α1 expression, together with a reliance on mitochondrial oxidative phosphorylation for ATP production, may explain why O₂-sensing cells are particularly sensitive to hypoxia. It is also possible that a lower resting energy state in O₂-sensing cells relative to other cell types may alter the metabolic setting in such a way that may increase the sensitivity of these cells to hypoxia by lowering the threshold for activation of AMPK.

Given the ubiquitous expression of AMPK in eukaryotes, it is intriguing to speculate on the possibility that the targeting of AMPK-dependent Ca²⁺ signalling mechanisms to specialized cells that function to monitor O₂ supply may have been a relatively early evolutionary development. This may be revealed by future studies on the role of AMPK as a metabolic sensor and effector of chemotransduction by hypoxia in O₂-sensing cells and tissues from primitive species; for example, systemic arteries of cyclostomes, arteries that feed the skin of amphibia and neuroepithelial cells from the gills of teleosts. If this proves to be the case, it would suggest that AMPK-dependent signalling mechanisms have likely been adapted to serve a variety of cell- and tissue-specific functions for which they are fit for purpose.