Divergent roles of glycolysis and the mitochondrial electron transport chain in hypoxic pulmonary vasoconstriction of the rat: identity of the hypoxic sensor

Abstract

1. The mechanisms responsible for sensing hypoxia and initiating hypoxic pulmonary vasoconstriction (HPV) are unclear. We therefore examined the roles of the mitochondrial electron transport chain (ETC) and glycolysis in HPV of rat small intrapulmonary arteries (IPAs).

2. HPV demonstrated a transient constriction (phase 1) superimposed on a sustained constriction (phase 2). Inhibition of complex I of the ETC with rotenone (100 nm) or complex III with myxothiazol (100 nm) did not cause vasoconstriction in normoxia, but abolished both phases of HPV. Rotenone inhibited the hypoxia-induced rise in intracellular Ca²⁺ ([Ca²⁺]_i). Succinate (5 mm), a substrate for complex II, reversed the effects of rotenone but not myxothiazol on HPV, but did not affect the rise in NAD(P)H fluorescence induced by hypoxia or rotenone. Inhibition of cytochrome oxidase with cyanide (100 μm) potentiated phase 2 constriction.

3. Phase 2 of HPV, but not phase 1, was highly correlated with glucose concentration, being potentiated by 15 mm but abolished in its absence, or following inhibition of glycolysis by iodoacetate or 2-deoxyglucose. Glucose concentration did not affect the rise in [Ca²⁺]_i during HPV.

4. Depolarisation-induced constriction was unaffected by hypoxia except in the absence of glucose, when it was depressed by ∼50 %. Depolarisation-induced constriction was depressed by rotenone during hypoxia by 23 ± 4 %; cyanide was without effect.

5. Hypoxia increased 2-deoxy-[³H]glucose uptake in endothelium-denuded IPAs by 235 ± 32 %, and in mesenteric arteries by 218 ± 38 %.

6. We conclude that complex III of the mitochondrial ETC acts as the hypoxic sensor in HPV, and initiates the rise in smooth muscle [Ca²⁺]_i by a mechanism unrelated to changes in cytosolic redox state per se, but more probably by increased production of superoxide. Additionally, glucose and glycolysis are essential for development of the sustained phase 2 of HPV, and support an endothelium-dependent Ca²⁺-sensitisation pathway rather than the rise in [Ca²⁺]_i.

Neither the identity of the hypoxic sensor that initiates hypoxic pulmonary vasoconstriction (HPV) nor its signalling pathways have been precisely determined. It is, however, generally accepted that hypoxia causes an increase in intracellular [Ca²⁺] ([Ca²⁺]_i) of the pulmonary vascular smooth muscle (VSM). It has been variously suggested that this is related to inhibition of K⁺ channels and subsequent Ca²⁺ entry via L-type Ca²⁺ channels (Post et al. 1992; Weir & Archer, 1995), release of Ca²⁺ from stores (Gelband & Gelband, 1997; Jabr et al. 1997; Wilson et al. 2001), and/or activation of voltage-independent Ca²⁺ entry pathways (Robertson et al. 2000b). However, there is also evidence that the endothelium is necessary for sustained HPV (Demiryurek et al. 1993; Kovitz et al. 1993; Leach et al. 1994; Lazor et al. 1996; Ward & Aaronson, 1999), and we have proposed that the endothelium releases a factor during hypoxia that causes Ca²⁺ sensitisation of VSM (Robertson et al. 1995, 2000a; Ward & Aaronson, 1999).

Most current theories regarding the hypoxic sensor involve changes in cytosolic redox state or reactive oxygen species (ROS). These fall broadly into two apparently opposed camps, one proposing a more reduced redox state and fall in ROS as the key signal (Archer et al. 2000), and the other a rise in ROS (Marshall et al. 1996; Chandel & Schumacker, 2000; Weissmann et al. 2000). Confusingly, similar mechanisms are suggested as possible hypoxic sensors by both, namely an NAD(P)H oxidase (Marshall et al. 1996; Jones et al. 2000; Weissmann et al. 2000) and the mitochondrial electron transport chain (ETC) (Archer et al. 2000; Chandel & Schumacker, 2000). However, it has been shown that transgenic mice lacking the critical gp91^phox subunit of NADPH oxidase still exhibit O₂ sensing and HPV (Archer et al. 1999), and this, coupled with doubts concerning selectivity of agents used to inhibit NAD(P)H oxidases, has been used to support the case for the ETC (Archer et al. 2000; Chandel & Schumacker, 2000).

Glycolysis provides substrate for mitochondria in the form of pyruvate and reducing equivalents, but may directly account for as much as a third of ATP production in smooth muscle, a proportion that is increased during hypoxia (Wendt, 1989). Glucose utilisation in the lung is rate limited by glucose entry (Perez-Diaz et al. 1977), but hypoxia potentiates glucose uptake in many tissues (Zhang et al. 1999). Early studies in perfused lungs suggested that HPV was potentiated by inhibition of glycolysis (Stanbrook & McMurtry, 1983), but more recent reports have shown that high glucose enhances HPV (Wiener & Sylvester, 1991). The effects of glucose on HPV may not be related to the ETC, as it has been reported in rat isolated pulmonary arteries that whereas inhibition of oxidative phosphorylation with rotenone abolished the early transient phase of HPV (phase 1), the sustained late phase constriction (phase 2) depended on ATP derived from glycolysis (Zhao et al. 1996).

There is clearly considerable confusion in the literature, and we therefore investigated the relative roles of the mitochondrial ETC and glycolysis in HPV of rat small intrapulmonary arteries (IPAs), using a combination of inhibitors and substrates to dissect out components of these pathways. Our results suggest that complex III of the ETC is the key hypoxic sensor for HPV, and generates a signal that initiates the rise in VSM [Ca²⁺]_i. However, we also show that glucose and glycolysis are essential for the sustained phase 2 component of HPV, and suggest that they act by supporting endothelium-dependent Ca²⁺ sensitisation rather than by altering [Ca²⁺]_i.

METHODS

IPA mounting

Male Wistar rats (250–350g) were anaesthetised with sodium pentobarbitone (55 mg kg⁻¹i.p.) and killed by cervical dislocation. The lungs and in some cases intestines were placed in physiological salt solution (PSS) containing (mm): NaCl 118; NaHCO₃ 24; MgSO₄ 1; NaH₂PO₄ 0.435; glucose 5; CaCl₂ 1.8; and KCl 4. Small IPAs (490 ± 8 μm i.d.) and mesenteric arteries (368 ± 34 μm) were dissected free of adventitia, mounted in a small vessel myograph at 37 °C, and gassed with 95 % air-5 % CO₂ (pH 7.35). They were stretched to give an equivalent transmural pressure of 30 mmHg as previously described (Leach et al. 1994). In some experiments the endothelium was disrupted by gently rubbing the luminal surface with a 40 μm wire. The presence of an endothelium was determined with acetylcholine (1 μm) following constriction with 20 μm prostaglandin F_2α (PGF_2α).

Hypoxia protocol

IPAs were equilibrated with three exposures to 80 mm KCl-PSS (KPSS), of 2 min duration (isotonic replacement of NaCl by KCl), as previously described (Leach et al. 1994). A small degree of agonist-induced tone is required to facilitate the hypoxic response in isolated rat IPAs, and vessels were therefore exposed to ∼5 μm PGF_2α for 20 min prior to, and during, the hypoxic challenge, equivalent to ∼17 % of the response to KPSS. Pretone was stable over the experimental period (Robertson et al. 1995, 2000b). Hypoxia was induced by gassing with 0–1 % O₂-5 % CO₂, balance N₂, to obtain a chamber P_O2 of typically 15–18 mmHg, compared with a control of 135–145 mmHg. After 45–60 min the vessels were reoxygenated and washed with PSS. There were no changes in PSS pH or osmolarity over this period. Oxygen tension was monitored via a dissolved oxygen meter (Diamond General electrode, Ann Arbor, MI, USA; Strathkelvin oxygen meter, Glasgow, UK). HPV in rat IPAs is reproducible providing the vessels are allowed at least 1 h to recover between hypoxic challenges (Leach et al. 1994; Robertson et al. 2000b). A recovery period of 60–90 min was therefore used between control and experimental exposures.

Estimation of [Ca²⁺]_i

IPAs were loaded with the Ca²⁺-sensitive fluorophore fura PE-3, via incubation with the acetoxymethyl ester (3 μm) for 2 h at room temperature. The vessels were then washed with PSS, the bath temperature raised to 37 °C, and the myograph transferred to the stage of an inverted fluorescence microscope and spectrophotometer (Nikon Diaphot, Nikon Ltd, UK; Cairn Research Ltd, Newnham, Kent, UK). Changes in [Ca²⁺]_i were estimated from the ratio of light emitted through a > 500 nm emission filter when the vessel was illuminated at 340 and 380 nm, respectively (F_340/380). As the microscope is focused on the outside surface of the vessel, the measured fluorescence is almost certainly derived entirely from the smooth muscle, and not from the endothelium. Consistent with this, we have previously shown that removal of the endothelium does not alter F_340/380 in fura-2-loaded IPA during hypoxia (Ward & Aaronson, 1999).

Estimation of changes in NAD(P)H

Standard autofluorescence techniques for NAD(P)H use a single excitation wavelength, and are compromised in intact preparations by tissue movement and non-specific fluorescence. We therefore modified a ratiometric method described for beating hearts (Scott et al. 1994). IPAs were mounted as above for estimation of [Ca²⁺]_i, though without fura PE-3, with the same filter configurations. The technique makes use of the fact that NAD(P)H fluoresces at ∼500 nm more greatly when excited at 340 nm than at 380 nm; the ratio of light emitted following excitation at 340 and 380 nm is therefore proportional to changes in NAD(P)H. Movement of the preparation affects emission at both wavelengths equally, and does not affect the ratio.

The fact that the same filters are used for autofluorescence and fura PE-3 raises the problem that changes in NAD(P)H-related autofluorescence during hypoxia may compromise estimation of [Ca²⁺]_i with fura PE-3. We therefore estimated the effect that hypoxia-induced changes in autofluorescence would have on the 340 nm/380 nm ratio of preparations loaded with fura PE-3 (F_340/380) in three preparations. We recorded the hypoxia-induced increase in autofluorescence at 340 and 380 nm prior to loading each preparation with fura PE-3; following loading total fluorescence during normoxia was increased by ∼75–100 %. We then calculated the effect that the previously observed increases in autofluorescence at 340 and 380 nm during hypoxia would have on F_340/380, in other words the effective error that these changes in autofluorescence would have on the estimation of [Ca²⁺]_i. This amounted to ∼9 % of the change in ratio induced by KPSS in fura PE-3-loaded vessels; KPSS had no significant effect on autofluorescence (see Fig. 3A). This is a significant component of the rise in F_340/380 during phase 2 of HPV (e.g. Fig. 4, and Robertson et al. 2000b, Fig. 4), but is similar to the residual rise in F_340/380 that we have previously reported for phase 2 in the presence of 100 μm La³⁺ to block Ca²⁺ entry (Robertson et al. 2000b). La³⁺ had no effect on autofluorescence. Although it is likely that this analysis overestimates rather than underestimates the error in the Ca²⁺ signal, as autofluorescence decays by ∼30 % over the period of the experiment, it is clear that estimation of [Ca²⁺]_i with short wavelength fluorochromes such as fura PE-3 can only be used with care during hypoxia, and with reference to any changes in autofluorescence. We believe it is inadvisable to attempt to correct more precisely for such changes in autofluorescence, as variations in total fluorescence and efficiency of fura PE-3 loading between preparations limit the use of time matching for the procedure we describe above, and would be likely to result in large inaccuracies.

Figure 3. NAD(P)H autofluorescence.

Original traces of changes in 340 nm/380 nm fluorescence ratio under various conditions (each typical of 3–6 preparations). The dotted lines labelled r and o represent the estimated fully reduced and fully oxidised states, obtained by rapid gassing with 0 % O₂ (< 5 mmHg, ‘anoxia’) in the presence of substrate, and 95 % O₂ in the absence of substrate, respectively. [PGF_2α] was 5 μm.

Figure 4. Effect of glucose on [Ca²⁺]_i and tension during HPV in IPA.

A and B, IPAs were exposed to hypoxia in the presence of 5 μm PGF_2α and 0 (▵), 5 (○, control) and 15 mm (•) glucose. A shows tension development and B the change in F_340/380 expressed as percentage response to KPSS, an indication of changes in [Ca²⁺]_i. Each symbol represents the mean ±s.e.m. of 5 IPAs.

Effects of extracellular glucose

HPV was induced in the presence of 0–15 mm glucose, following preincubation for 20 min. Sucrose was used to maintain osmolarity for concentrations below 5 mm. Addition of 10 mm sucrose to experiments with 5 mm glucose to mimic the changes in osmolarity during high glucose had no effect. Three hypoxic responses were examined in each IPA using randomly sequenced glucose concentrations. The effects of glucose on the response of mesenteric arteries to hypoxia was also examined. In some experiments changes in [Ca²⁺]_i (F_340/380) were simultaneously recorded with tension. We additionally determined whether pyruvate, an end-product of glycolysis, or succinate, a substrate for complex II of the mitochondrial ETC, could substitute for glucose. The effect of glucose concentration on depolarisation-induced constriction was examined during normoxia and hypoxia by constructing concentration- response curves (CRCs) to K⁺ (iso-osmolar substitution for Na⁺) in IPAs.

Endothelium-denuded IPAs

Removal of the endothelium abolishes phase 2 of HPV in this preparation (Leach et al. 1994). We examined whether 15 mm glucose could restore phase 2 in endothelium-denuded IPA. As removal of the endothelium increased the response to PGF_2α, pretone was matched by adjusting the concentration of PGF_2α in denuded IPAs (∼1 μm). A control response was obtained from endothelium-intact IPAs. The endothelium was then removed in situ, matched pretone induced, and the preparation again challenged with hypoxia. Glucose concentration was increased to 15 mm for the final hypoxic challenge.

Modulation of glycolysis and mitochondrial electron transport chain

Glycolysis was inhibited with iodoacetate (100 μm); although commonly used for this purpose, iodoacetate is a non-specific alkylating agent, and we also examined 2-deoxyglucose (15 mm) in the absence of glucose. The ETC was inhibited at complex I with the specific inhibitor rotenone (100 nm), at complex III with myxothiazol (100 nm), an inhibitor of electron transfer to the Rieske iron-sulphur centre, or with the cytochrome oxidase inhibitor cyanide (100 μm). This concentration of cyanide is greatly in excess of its IC₅₀ (∼1 μm) for cytochrome oxidase (Wilson et al. 1972). Succinate is a substrate for complex II, and should therefore bypass inhibition of complex I, but not of complex III (see Fig. 10). Experiments with rotenone and myxothiazol were therefore repeated in the presence of 5 mm succinate. All agents were applied 20 min before HPV was induced and pretone was matched between experimental and control conditions. The effects of agents on basal tone and KPSS-induced constriction were also examined.

Figure 10. Potential model for role of ETC and glycolysis during HPV.

Simplified diagram of relevant cellular metabolic pathways, showing proposed actions of inhibitors, substrates and hypoxia. GluT, sarcolemmal glucose transporter; MAS, malate- aspartate shuttle (shown as single entity for clarity); I, II, III and IV, electron transport chain complexes; dotted arrows, electron transfer between complexes; Q, coenzyme Q (ubiquinone cycle); SO, superoxide.

2-Deoxy-[³H]glucose uptake

Uptake of radiolabelled 2-deoxy-[³H]glucose was examined in endothelium-denuded IPAs and mesenteric arteries, and endothelium-intact IPAs. The method was a modification of that used for skeletal muscle (Cartee et al. 1991). Two IPAs were mounted on a dual myograph, and bathed in PSS (5 mm glucose). 2-Deoxy-[³H]glucose (10 μCi ml⁻¹) was added to the bath and pretone induced. The chamber was then divided with a partition so as to allow one artery to act as a control whilst the other was challenged with hypoxia. Hypoxia was induced 20 min after addition of label. After 60 min vessels were removed from the chamber, and washed several times in ice-cold PSS containing phloretin (200 μm) to prevent further uptake and remove labelled PSS. The preparations were blotted dry, weighed, and placed in vials containing 250 μl tissue solubiliser (Solvease, BDH). Aliquots of labelled PSS from the bath (10 μl) were treated similarly. Once the tissue had dissolved, 10 ml scintillant (Pico-Fluor 15, Canberra Packard) was added before counting in a gamma scintillation counter. Counts were normalised for tissue wet weight, and glucose uptake calculated from the specific activity of the PSS and expressed as micromoles per gram wet weight per hour.

Materials and measurements

All drugs were obtained from Sigma-Aldrich Ltd (Poole, Dorset, UK), with the exception of PGF_2α (Upjohn Pharmaceuticals Ltd, Crawley, UK), insulin (human Actrapid, Novo Nordisk Pharmaceuticals, UK), and fura PE-3/AM (Calbiochem, Nottingham, UK) Other chemicals were of Analar quality (BDH, Southampton, UK). 2-Deoxy-[³H]glucose was obtained from Amersham Novoscience, UK. Drugs were prepared as stock solutions using PSS, with the exception of rotenone and myxothiazol which were dissolved in DMSO, and prepared as stock solutions by dilution in PSS. Final DMSO concentration was < 0.01 %, which had no effect on its own. Tension is expressed as a percentage of the response to KPSS (80 mm KCl). Mean changes in [Ca²⁺]_i are expressed as a percentage of the change in F_340/380 induced by KPSS; although not linearly related to [Ca²⁺]_i this provides a reasonable qualitative index. However, note that fura PE-3 will give a quantitative overestimate of [Ca²⁺]_i during hypoxia due to changes in autofluorescence (see Methods). All values are given as means ±s.e.m.; n represents a single IPA per rat except where stated. Data were compared using Student's t test or ANOVA with a Student-Newman-Keuls (SNK) post hoc test as appropriate.

RESULTS

Effect of glucose concentration

Hypoxia elicited a biphasic response in rat IPAs, consisting of a transient phase 1 constriction superimposed on a sustained phase 2 constriction (Fig. 1A). Altering glucose concentration between 0 and 15 mm had no significant effect on basal tension, pretone elicited by PGF_2α, or phase 1; the apparent reduction with 0 mm glucose did not reach significance (P = 0.14). Phase 2, however, was strongly dependent on glucose, showing potentiation with 15 mm glucose, and depression as the concentration fell below 5 mm, such that at 1 and 0 mm glucose phase 2 was essentially abolished (Fig. 1A). Pyruvate (5 mm) was unable to substitute for glucose (tension at 60 min: 5 mm glucose, 15.0 ± 3.8 % response to KPSS; 5 mm pyruvate, −1.3 ± 3.5 %; n = 6 from 4 rats, P < 0.01). This suggests that inhibition of phase 2 by low glucose was not due to substrate limitations of oxidative phosphorylation.

Figure 1. Effect of glucose concentration on the response to hypoxia in IPA and mesenteric arteries.

A, IPAs were preconstricted with 5 μm PGF_2α and challenged with hypoxia for 60 min in the presence of 0 (▵), 1 (▪), 2.5 (□), 5 (○, control) or 15 mm (•) glucose. Each experimental plot was significantly different from control (P < 0.05; ANOVA); an SNK post hoc test revealed that time points between 15 and 60 min were significant for 0, 1 and 2.5 mm glucose compared with control (not shown for clarity), whereas for 15 mm each point between 25 and 60 min was significant (*P < 0.05). The symbols represent the mean ±s.e.m. of 5–10 IPAs from 8 rats. Where error bars are not shown, the error is smaller than the symbol. B, similar to A except with 6 mesenteric arteries.

Inhibition of glycolysis with iodoacetate suppressed phase 1 (control, 57 ± 5 % KPSS; iodoacetate, 11 ± 2 %; n = 7 from 5 rats, P < 0.05), and almost abolished phase 2 (at 60 min: control, 32 ± 5 %; iodoacetate, 3 ± 2 %; P < 0.001). Similar results were observed with 2-deoxyglucose (phase 1: control, 72 ± 3 %; deoxyglucose, 13 ± 2 %, P < 0.05; phase 2: control, 24 ± 5 %; deoxyglucose,−1.4 ± 2 %; both P < 0.001; n = 5 from 3 rats). Neither iodoacetate nor 2-deoxyglucose affected basal tension or KPSS-induced constriction in normoxia (n = 4–6). In hypoxia, iodoacetate depressed KPSS-induced constriction by 70 ± 4 % (n = 4; P < 0.001).

As previously reported (Leach et al. 1994), mesenteric arteries exhibited a transient constriction to hypoxia followed by vasorelaxation (Fig. 1B). Increasing glucose to 15 mm had no significant effect on this response (Fig. 1B). Removal of the endothelium from IPAs abolished phase 2 of HPV (Fig. 2) as previously described (Leach et al. 1994; Ward & Aaronson, 1999), with the result that at 60 min there was a ∼50 % fall in tension below the level of pretone. Increasing glucose to 15 mm had no effect on either phase of HPV in denuded IPAs, suggesting that potentiation of HPV by high glucose is due to an effect on the endothelium-dependent component of the phase 2 constriction.

Figure 2. Glucose and the response to hypoxia in endothelium-denuded IPA.

IPAs were challenged with hypoxia in 5 mm glucose as previously described (○), following which the endothelium was removed and the hypoxic challenge repeated in the presence on 5 mm (□) or 15 mm (▪) glucose. Each symbol represents the mean ±s.e.m. of 6 IPAs. *P < 0.05 compared with intact IPAs.

Effect of hypoxia on NAD(P)H fluorescence

Neither constriction with KPSS nor induction of pretone with PGF_2α changed the autofluorescence 340 nm/380 nm ratio (Fig. 3). Hypoxia caused a rapid, sustained rise in the ratio that reversed on reoxygenation. Rapid bubbling with 0 % O₂–5 % CO₂ to reduce the P_O2 to < 5 mmHg (‘anoxia’) in the presence of substrate caused a further increase in the ratio, consistent with near-maximal reduction of NAD(P)⁺ to NAD(P)H (Fig. 3H). Changing glucose from 5 to 15 mm had no effect on the NAD(P)H fluorescence ratio, either during normoxia or hypoxia (Fig. 3B and C). Changing to 0 mm glucose during normoxia caused a small fall in the ratio, consistent with a fall in NAD(P)H, but had a relatively minor effect on the rise in the ratio induced by hypoxia (Fig. 3D), although the ratio then decayed during the hypoxic challenge. A similar response was observed in the presence of iodoacetate (Fig. 3E).

Effect of glucose on fura PE-3 fluorescence

The effects of glucose concentration on tension during phase 2 were not associated with any changes in F_340/380 (Fig. 4). As glucose also had little effect on the increase in NAD(P)H autofluorescence during hypoxia (Fig. 3B–D), this suggests that the effects of glucose on tension were not due to alterations in [Ca²⁺]_i.

Glycolysis and the ETC

Rotenone, myxothiazol, cyanide and succinate had no effect on basal tone in IPAs (n = 4–11; see also Fig. 5), nor on KPSS-induced constriction of IPAs in normoxia (n = 4–6). Rotenone reduced KPSS-induced constriction during hypoxia by 23 ± 4 % (n = 4, P < 0.05), whereas neither myxothiazol nor cyanide had any significant effect (n = 4–6). Cyanide (Fig. 3F) and succinate (not shown) had no effect on basal NAD(P)H autofluorescence ratio or the increase caused by hypoxia (n = 3–4). Rotenone, however, caused a rise in the ratio that was not easily reversible (Fig. 3G), and equivalent to that caused by anoxia (< 5 mmHg P_O2). Hypoxia in the presence of rotenone caused only a marginal effect, and subsequent application of succinate (5 mm) in the presence of both hypoxia and rotenone had no effect on NAD(P)H (Fig. 3G). These results are consistent with inhibition of NADH oxidation/consumption by the ETC.

Figure 5. Inhibition of the electron transport chain.

IPAs were exposed to hypoxia in the presence of PGF_2α (○), and following preincubation for 20 min with rotenone (100 nm, •), myxothiazol (100 nm, □) or cyanide (100 μm, ▴). A shows the effects of rotenone (n = 10, P < 0.001 for all points during hypoxia) and myxothiazol (n = 4, P < 0.01) on tension development during hypoxia. Note that neither had any effect on tension during normoxia. B shows the effect of rotenone on F_340/380 measured simultaneously in 4 of the above preparations (P < 0.01); note that rotenone itself increased F_340/380 (P < 0.01). C and D, as above except for cyanide (▴; A, n = 8; B, n = 4). *P < 0.05.

Rotenone and myxothiazol abolished both phases of constriction during hypoxia (Fig. 5A; P < 0.001). Succinate (5 mm) was able to reverse the inhibitory effect of rotenone (Fig. 6A), but could not reverse that of myxothiazol (not shown, n = 4). Succinate when applied to a rotenone-treated preparation after 30 min of hypoxia caused an immediate increase in tension (not shown, n = 4). This is consistent with the hypothesis that succinate can bypass inhibition of complex I by providing substrate to complex II, but cannot reverse the effects of myxothiazol, which inhibits distally to complex II (see Fig. 10). Succinate was unable, however, to substitute for glucose in supporting phase 2 of HPV, although in the presence of glucose it caused a small enhancement that only reached significance at one point (Fig. 6A). Rotenone itself caused an increase in F_340/380 when applied in normoxia (Fig. 5B); the size of this increase (∼14 % response to KPSS) suggests that it may be largely attributed to the increase in NAD(P)H fluorescence reported above (Fig. 3G) rather than an increase in [Ca²⁺]_i; note that rotenone did not cause any increase in basal tone (Fig. 5A). Following preincubation with rotenone hypoxia caused little further increase in F_340/380 (Fig. 5B; P < 0.01); the level of F_340/380 during hypoxia was significantly less than that in the absence of rotenone (P < 0.01), suggesting that rotenone blocked the hypoxia-induced rise in [Ca²⁺]_i. Succinate was also able to reverse the inhibitory effect of rotenone on the rise in F_340/380 during hypoxia (n = 3).

Figure 6. Effect of succinate on HPV.

IPAs were challenged with hypoxia in the presence of 5 mm glucose (○), 5 mm glucose, 100 nm rotenone and 5 mm succinate (•, n = 5), 5 mm glucose and 5 mm succinate (▴, n = 4) or 0 mm glucose and 5 mm succinate (□, n = 4). #P < 0.05, *P < 0.001 compared with 5 mm glucose (control).

In contrast to rotenone and myxothiazol, cyanide (100 μm) had no significant effect on phase 1, and potentiated phase 2 (Fig. 5C); after 40 min the difference between control and cyanide groups was no longer significant, and in four experiments extended to 60 min tension was the same in both groups (control, 28 ± 6 %; cyanide, 31 ± 3 %; n = 4, not significant). Cyanide had no effect on the rise in F_340/380 during hypoxia (Fig. 5D), and as cyanide did not affect NAD(P)H autofluorescence (Fig. 3F), this can be taken to mean that cyanide had no effect on the hypoxia-induced rise in [Ca²⁺]_i. The effect of cyanide on phase 2 constriction required prior incubation; addition of cyanide at 25 min from the initiation of hypoxia had no apparent effect (data not shown, n = 4).

Effect of glucose and hypoxia on depolarisation-induced constriction

Varying glucose concentration between 0 and 15 mm had no significant effect on the K⁺ CRC during normoxia (n = 8, from 6 rats). Hypoxia did not cause any significant change in the K⁺ response curves for 2.5–15 mm glucose (e.g. Fig. 7); for example, the tension induced by 60 mm[K⁺] in the presence of 2.5 mm glucose was 8.9 ± 1.2 mN mm⁻¹ during normoxia and 7.8 ± 1.0 mN mm⁻¹ during hypoxia (n = 8). In the absence of glucose, however, tension was reduced significantly during hypoxia (at 60 mm[K⁺]: normoxia, 7.6 ± 1.0 mN mm⁻¹; hypoxia, 4.2 ± 0.8 mN mm⁻¹; n = 8, P < 0.05). The relationship between glucose concentration and phase 2 of HPV was much steeper than that for either high K⁺-induced constriction or phase 1 (Fig. 7).

Figure 7. Relative dependency of phase 1 and 2 of HPV, and depolarisation-induced constriction on glucose concentration during hypoxia.

The graph shows the relative dependency on glucose of phase 1 (•) and phase 2 (○) of HPV, and constriction to high [K⁺] (□) during hypoxia, using data derived from Fig. 1A and the depolarisation-induced tension experiments. The data points have been normalised by expressing them as a percentage of developed tension in 5 mm glucose. The values for phase 2 were taken at 60 min after initiation of the hypoxic challenge. In order to match tension approximately between conditions, data for 35 mm[K⁺] were used to generate the graph (∼50 % response to KPSS), although higher concentrations show little qualitative difference. Each symbol represents the mean ±s.e.m. of 5–8 IPAs.

Effect of hypoxia on glucose uptake

Figure 8 shows the effect of hypoxia on 2-deoxy-[³H]glucose uptake in IPAs and mesenteric arteries preconstricted with PGF_2α. Hypoxia significantly increased uptake in both endothelium-denuded (235 ± 32 %, n = 5, P < 0.05) and intact IPAs (277 ± 61 %, n = 9, P < 0.05), and denuded mesenteric arteries (218 ± 38 %, n = 4, P < 0.05). There was no significant difference between endothelium-intact and denuded IPAs during either normoxia or hypoxia. Uptake was greater in IPAs than mesenteric arteries in both normoxia and hypoxia (P < 0.05).

Figure 8. 2-Deoxy-[³H]glucose uptake in IPA and mesenteric arteries during hypoxia.

Uptake of 2-deoxy-[³H]glucose was examined after 60 min of hypoxic or normoxic challenge, in intact () and endothelium-denuded (□) IPAs, and denuded mesenteric arteries (▪) as described in Methods. The columns represent the mean ±s.e.m. of 4–9 arteries. *P < 0.05 hypoxic compared with normoxic; #P < 0.05, ##P < 0.01 mesenteric denuded compared with pulmonary denuded.

Insulin enhances glucose uptake (for example see Zhang et al. 1999), and we examined its effect on HPV (preincubation with 10 mU ml⁻¹ for 20 min, and present during hypoxic challenge). Insulin did not affect phase 1 of HPV, but reduced the subsequent fall in tension, so that phase 2 was significantly increased between 15 and 55 min (Fig. 9; P < 0.05 at each point). By 60 min there was no significant difference between control and insulin-treated IPAs. Acute application of insulin at 25 min after the start of the hypoxic challenge had no significant effect (data not shown, n = 5).

Figure 9. Effect of insulin on HPV.

IPAs were exposed to hypoxia in the presence of 5 μm PGF_2α (○), and following 20 min preincubation with 10 mU ml⁻¹ insulin (•). Each symbol represents the mean ±s.e.m. of 4 IPAs. *P < 0.05.

DISCUSSION

As we have previously reported, hypoxia induced a biphasic constrictor response in rat small IPAs, with a transient phase 1 superimposed on a more slowly developing but sustained and endothelium-dependent phase 2 constriction (Leach et al. 1994; Ward & Robertson, 1995; Robertson et al. 2000b). Our current results show that the development of both the phase 1 and the phase 2 components of HPV is critically dependent on the mitochondrial ETC, but that this is not due to energy lack for contraction. However, phase 2 is also strongly related to the concentration of extracellular glucose. These findings are potentially of fundamental importance to our understanding of the mechanisms of HPV.

The role of the mitochondria

It is known that complex III of the mitochondrial ETC can generate superoxide by electron donation to O₂ via ubisemiquinone (Turrens et al. 1985), part of the ubiquinone-ubiquinol-ubisemiquinone (coenzyme Q) cycle that transfers electrons from complex I and II to complex III. There is good evidence that such production of reactive oxygen species (ROS) increases in hypoxia, which has led to the hypothesis that the ETC may act in many cell types as a hypoxic sensor (Dawson et al. 1993; Chandel & Schumacker, 2000). A preliminary report suggested that this could also be the case for HPV (Waypa et al. 2000). In the present study, we have shown that both rotenone, a specific inhibitor of complex I, and myxothiazol, an inhibitor of electron transfer to the Rieske iron-sulphur centre of complex III, completely ablated both phases of HPV. However, they had relatively little effect on depolarisation-induced constriction, indicating that there was still sufficient energy from other sources for tension development. Rotenone also inhibited the hypoxia-associated rise in [Ca²⁺]_i, as estimated from F_340/380, during both phases (Fig. 5). As complex II can also provide electrons to complex III (see Fig. 10), succinate, a substrate for complex II, should be able to bypass the inhibitory effect of rotenone but not that of myxothiazol, and this is indeed what was found (Fig. 6). The small increase in HPV with succinate and glucose alone would be consistent with additional electron transfer to complex III. The fact that HPV was abolished by rotenone and myxothiazol, but potentiated by the complex IV cytochrome oxidase inhibitor cyanide lends support to the hypothesis that complex III of the ETC acts as the hypoxic sensor for HPV, and that the resultant signal, presumably a direct or indirect action of superoxide, is responsible for initiating the hypoxia-induced rise in [Ca²⁺]_i.

Our results are broadly consistent with previous studies in perfused lungs, in that inhibition of complex I has been shown to inhibit HPV and production of ROS (Archer et al. 1993; Waypa et al. 2000). Conversely cyanide increased ROS and either did not alter (Archer et al. 1993) or potentiated HPV (Waypa et al. 2000). It should be noted, however, that Archer et al. (1993) reported that rotenone caused vasoconstriction on its own in perfused lungs, something we never observed in IPAs. Their interpretation therefore differs from our own, and this is discussed below. The one previous study on isolated pulmonary arteries reported that rotenone only abolished the transient component of HPV, but not the sustained phase (Zhao et al. 1996). There are, however, important differences between the preparation used by the latter group and our own, including the much larger degree of pretone (∼50 % of KPSS) and the use of main branch pulmonary artery rather than the physiologically relevant small IPAs as used in our experiments (Zhao et al. 1996).

Although there is strong evidence that mitochondrial production of ROS is enhanced in hypoxia (reviewed in (Chandel & Schumacker, 2000), it is less clear why this occurs. Chandel et al. (1997) have shown that hypoxia decreases the V_max of cytochrome oxidase. They hypothesised that this would reduce electron transfer into complex IV and so enhance transfer to ubisemiquinone, and thus superoxide production (Chandel et al. 1997; Chandel & Schumacker, 2000). This implies that cyanide should mimic hypoxia, and cyanide has been shown to potentiate HPV and/or ROS in perfused lungs (Archer et al. 1993; Waypa et al. 2000), and HPV in the present study. However, we have also shown that cyanide does not cause constriction of IPAs in normoxia, nor enhance HPV when applied midway through the hypoxic challenge. It is therefore possible that its potentiation of HPV may be at least partly due to alternative mechanisms (see below). Recently, Chandel et al. (2000) modified their original proposal in the light of data concerning HIF-1 stabilisation during hypoxia, and suggested that complex III may itself have intrinsic sensitivity to hypoxia. This may be more consistent with our cyanide data.

Although both the above and other reports (Marshall et al. 1996; Weissmann et al. 1998, 2000; Waypa et al. 2000) would be consistent with increased ROS production being key to HPV, there remains controversy surrounding the roles of ROS and redox state. Archer and Weir have proposed a redox state model, where during hypoxia a more reducing cytosolic environment and a fall in ROS causes inhibition of membrane K⁺ channels, depolarisation and Ca²⁺ entry via L-type channels (Archer et al. 1993, 2000). Some of the direct evidence for this model is based on the fact that in perfused lungs they found that rotenone could mimic hypoxia and cause vasoconstriction, and that hypoxia caused a fall in ROS as estimated by chemiluminescence in the whole lung (Archer et al. 1993). A perplexing aspect of this study is that antimycin A, an inhibitor of complex III distal to ubisemiquinone, also caused a decrease in chemiluminescence, even though this agent is known to increase mitochondrial superoxide generation (Chandel & Schumacker, 2000; Raha et al. 2000). In favour of the reduced redox hypothesis, it is clear that NAD(P)H does increase in hypoxia, and most studies that have examined exogenously applied H₂O₂ have reported only vasorelaxation in pulmonary arteries at moderate doses (reviewed in Jones & Morice, 2000).

Rotenone did not induce constriction of IPAs in our hands, although we cannot rule out a small rise in [Ca²⁺]_i (see Results), but it did cause a near-maximal rise in NAD(P)H fluorescence (Fig. 3). This does not necessarily obviate a coexistent increase in ROS, which could in any case be compartmentalised or locally targeted. There is increasing evidence, for example, that mitochondria and the peripheral sarcoplasmic reticulum are closely approximated in VSM (Drummond & Tuft, 1999; Gordienko et al. 2001). Moreover, we have shown not only that succinate can reverse the inhibitory effect of rotenone on HPV, but that it does so without altering NAD(P)H fluorescence. The actions of succinate are difficult to explain if HPV is caused by a reduction in ROS and/or a simple increase in NAD(P)H, and if the reason that rotenone inhibits HPV is because it has already maximally stimulated the system, as implied by Archer et al. (2000). Although it is entirely possible that an increase in NAD(P)H during hypoxia is vital to some mechanisms, we observed no relationship between the presence or degree of HPV and the increase in NAD(P)H fluorescence, other than that the latter was always present during hypoxia whether or not constriction occurred.

The question arises as to how the signal from the mitochondria, presumably either superoxide or its product H₂O₂, causes the rise in [Ca²⁺]_i. Although it has been widely proposed that hypoxia causes inhibition of K⁺ channels and Ca²⁺ entry through voltage-gated L-type channels (e.g. Weir & Archer, 1995; Archer et al. 2000), we have recently demonstrated that IPAs still exhibit HPV under conditions where depolarisation and L-type channels can play no part (Robertson et al. 2000b). Instead, we provided evidence for other mechanisms, including capacitative Ca²⁺ entry subsequent to store emptying, and activation of a voltage-independent Ca²⁺ entry pathway (Robertson et al. 2000b). Oxidant stress in the form of superoxide or H₂O₂ can activate certain channels associated with these processes, and also induce Ca²⁺ release from intracellular stores (Graier et al. 1998; Herson et al. 1999; Balzer et al. 1999). It has been suggested that intracellular Ca²⁺ release is an important mechanism in HPV (Gelband & Gelband, 1997; Jabr et al. 1997). A recent and intriguing report has shown that cADPR-induced Ca²⁺ release may be a key event in HPV, specifically for the increase in [Ca²⁺]_i during phase 2 (Wilson et al. 2001). It is therefore interesting to note that superoxide has been shown to directly stimulate cADPR synthesis, and that this mechanism has been proposed to explain the Ca²⁺-releasing effect of superoxide (Kumasaka et al. 1999).

The role of glucose and glycolysis

Although our results show that both phases of HPV rely on the ETC, they also clearly demonstrate that the sustained phase 2 constriction is critically dependent on both the concentration of extracellular glucose and the presence of an endothelium. In contrast, the transient phase 1 constriction is essentially independent of extracellular glucose, and only partially reduced following removal of the endothelium (Figs 1 and 2). In the absence of glucose, intact IPAs responded to hypoxia with a phase 1 constriction followed by relaxation below initial tension, behaving in a similar fashion to endothelium-denuded IPAs in normal glucose. High glucose, however, potentiated phase 2 in intact IPAs, but had no effect on denuded IPAs or mesenteric arteries. High glucose has also been shown to potentiate HPV in perfused ferret lungs (Wiener & Sylvester, 1991). An important finding is that the effect of glucose on tension development during phase 2 was not due to changes in [Ca²⁺]_i (Fig. 4), and that in the absence of glucose tension was uncoupled from [Ca²⁺]_i for phase 2, but not phase 1. The possible implications of this are discussed below.

As neither pyruvate nor succinate were able to substitute for glucose, it is likely that phase 2 is dependent on glycolysis per se, rather than production of substrate for the mitochondria. Consistent with this, the rise in NAD(P)H fluorescence during hypoxia was only slightly reduced in zero glucose, although its rate of decline was increased (Fig. 3). As might therefore be expected, inhibition of glycolysis also abolished phase 2. These results are generally consistent with a previous study which proposed that phase 2, but not phase 1, requires glycolytic production of ATP (Zhao et al. 1996).

Glycolytic ATP is not essential for tension development in normoxia, as KPSS-induced tension was not affected by removal of glucose or inhibition of glycolysis. Inhibition of oxidative respiration with rotenone also had no effect on KPSS-induced constriction during normoxia, suggesting that either source of ATP is sufficient on its own to support contraction under these conditions. During hypoxia, depolarisation-induced tension was only partially depressed when glucose was absent or in the presence of rotenone, implying that adequate ATP is normally generated during hypoxia to support substantial force generation. This reflects our previous report that energy state and ATP are maintained in porcine pulmonary arteries during hypoxia, except in the absence of glucose (Leach et al. 1998, 2000).

Phase 2 of HPV was more highly correlated to glucose concentration than either phase 1 or high K⁺-induced constriction. This is illustrated in Fig. 7, where data derived from Fig. 1 and the depolarisation-induced tension experiments have been replotted in terms of the tension developed in the presence of control (5 mm) glucose. The effect of glucose on phase 2 therefore clearly does not involve an energy requirement for tension. We have previously suggested that phase 2 is associated with Ca²⁺ sensitisation of the VSM, mediated by an as yet unidentified endothelium-derived factor (Robertson et al. 1995; Robertson et al. 2000a), and we have recently isolated a pulmonary-specific Ca²⁺-sensitising agent from effluent of hypoxic perfused rat lungs that is not endothelin-1 (Robertson et al. 2001). As high glucose had no effect on the response to hypoxia in the absence of an endothelium, it seems reasonable to suggest that its potentiating action on phase 2 might be related to this endothelium-dependent Ca²⁺ sensitisation process. This hypothesis is strongly supported by the data in Fig. 4, which show that although tension during phase 2 was very significantly affected by glucose concentration, the latter had no effect on the changes in [Ca²⁺]_i during HPV.

The strong dependence of phase 2 tension on glucose could therefore reflect a requirement for glycolytic ATP within the signalling pathways associated with Ca²⁺ sensitisation in the VSM. There is significant compartmentalisation of ATP production in both VSM and other tissues, and several membrane-delimited processes are preferentially coupled to ATP derived from membrane-associated glycolysis (Paul, 1989; Zhang & Paul, 1994). It is possible to speculate therefore that phase 2 involves an energy- or phosphorylation-dependent step that occurs at the cell membrane. One plausible membrane-delimited process is activation of Rho-associated kinase by the monomeric G-protein RhoA. This is known to play an important role in Ca²⁺ sensitisation, and we have recently demonstrated that it is implicated in phase 2 of HPV in both IPAs and perfused lungs (Robertson et al. 2000a).

The above suggests that phase 2 would be potentiated by increased glucose uptake. In skeletal muscle hypoxia enhances glucose uptake by > 300 % (Cartee et al. 1991), but this has not been examined in VSM. We show here that glucose uptake was increased by hypoxia in IPAs to ∼250 % in an endothelium-independent manner. Although uptake was also increased in mesenteric arteries, it remained only ∼50 % of that in IPAs. Transport of glucose into rat VSM is primarily via GLUT 4 (Banz et al. 1996). In skeletal muscle, hypoxia stimulates translocation of GLUT 4 to the cell membrane over a period of ∼1 h (Cartee et al. 1991). This is similar to the time taken for phase 2 of HPV to reach a plateau (see Fig. 1), and in view of the sensitivity of phase 2 to glucose it is tempting to speculate that the slow rise in phase 2 may in part reflect a gradual translocation of GLUT 4 and a concomitant gradual increase in glucose influx and glycolysis. Insulin also enhances glucose uptake in VSM (Banz et al. 1996) and skeletal muscle (Cartee et al. 1991) by translocation of GLUT 4. We therefore examined whether preincubation with insulin would enhance phase 2 and/or allow it to reach a plateau more rapidly. As shown in Fig. 9, insulin potentiated at least the early part of phase 2, essentially abolishing its time dependency.

The effect of insulin shows some similarities to that of cyanide, in that both potentiate the early but not late stages of phase 2, and both require preincubation. The action of cyanide does not appear to be directly related to the putative ETC-associated mechanism driving the increase in [Ca²⁺]_i during hypoxia, as although cyanide increased tension it did not increase [Ca²⁺]_i; in essence cyanide appeared to potentiate the proposed glucose-dependent Ca²⁺ sensitisation (see above, and Fig. 5). It is therefore interesting to note that like hypoxia and insulin, cyanide also causes an increase in glucose uptake by translocation of GLUT transporters to the cell membrane (Zhang et al. 1999). Further time course studies on GLUT 4 translocation during hypoxia and in the presence of insulin and cyanide are required to confirm these speculations.

In conclusion, our results strongly suggest that complex III of the mitochondrial ETC plays a key role as a hypoxic sensor for HPV, and that during hypoxia a signal from complex III, probably superoxide anions, leads to the rise in VSM [Ca²⁺]_i that underlies both phase 1 and phase 2 of HPV. However, glucose and glycolysis are also critical for the sustained phase 2 constriction, and coupled with our previous studies these results suggest that they support hypoxia-induced, endothelium-dependent Ca²⁺ sensitisation. We have previously suggested that both a rise in VSM [Ca²⁺]_i and Ca²⁺ sensitisation are required for the development of sustained HPV (Ward & Aaronson, 1999). Figure 10 illustrates how we believe the various manipulations used in this study modulate HPV in rat small IPAs. Many questions remain to be elucidated, in particular concerning the means by which glucose may support Ca²⁺ sensitisation during hypoxia, and whether endothelium also relies on the ETC or a different hypoxic sensor to stimulate production of the putative Ca²⁺-sensitising agent. The identity of this agent is still unknown, although there is good evidence for both its existence and its end-point action via Rho-kinase (Kovitz et al. 1993; Robertson et al. 1995, 2000a, 2001; Gaine et al. 1998). Our results also highlight a potentially serious problem with the use of short wavelength fluorescent dyes when used to quantify intracellular ions during hypoxia, due to concomitant changes in NAD(P)H autofluorescence; re-evaluation of some previous studies may therefore be required.

Note added in proof

Since this work was submitted for publication, a relevant and important paper was published by G. B. Waypa, N. Chandel & P. T. Schumacker (Circulation Research88, 1259–1266 (2001)). They showed, using perfused lungs and cultured pulmonary arterial myocytes of rat, that mitochondria function as O₂ sensors during hypoxia, and that ROS generated in the ETC act as second messengers in HPV. Their results are entirely consistent with those presented here.