Cellular mechanisms by which adenosine induces vasodilatation in rat skeletal muscle: significance for systemic hypoxia

Abstract

1. In anaesthetized rats, we recorded arterial blood pressure (ABP), heart rate (HR), femoral blood flow (FBF) and femoral vascular conductance (FVC). We tested the effects of the nitric oxide (NO) synthesis inhibitor l-NAME (nitro-l-arginine methyl ester), or the ATP-sensitive K⁺ (K_ATP) channel inhibitor glibenclamide, on responses evoked by systemic hypoxia (breathing 8% O₂ for 5 min) or i.a. infusion for 5 min of adenosine, the NO donor sodium nitroprusside (SNP), the adenosine A₁ receptor agonist CCPA (2-chloro-N⁶-cyclopentyladenosine) or the adenosine A_2A receptor agonist CGS 21680 (2-p-(2-carboxyethyl)-phenethylamino-5′-N-ethylcarboxamidoadenosine hydrochloride).

2. l-NAME (10 mg kg⁻¹ i.v.) greatly reduced the increase in FVC induced by hypoxia or adenosine, as we have shown before, but had no effect on the increase in FVC evoked by SNP. In addition, l-NAME abolished the increase in FVC evoked by CCPA and greatly reduced that evoked by CGS 21680. These results substantiate the view that muscle vasodilatation induced by systemic hypoxia and infused adenosine are largely NO dependent. They also indicate that muscle dilatation induced by A₁ receptor stimulation is entirely NO dependent while that induced by A_2A receptors is largely NO dependent; dilatation may also be induced by direct stimulation of A_2A receptors on the vascular smooth muscle.

3. Glibenclamide (10 or 20 mg kg⁻¹ i.v.) reduced the increase in FVC induced by hypoxia, preferentially affecting the early part (< 1 min). In addition, glibenclamide greatly reduced the increase in FVC induced by adenosine, but it had no effect on that evoked by SNP. Further, glibenclamide abolished the increase in FVC evoked by CCPA and greatly reduced that evoked by CGS 21680. These results substantiate the view that hypoxia-induced muscle vasodilatation is initiated by K_ATP channel opening. They also indicate that NO does not induce muscle vasodilatation by opening K_ATP channels on the vascular smooth muscle, but indicate that the dilatation induced by adenosine and by A_2A receptor stimulation is largely dependent on K_ATP channel opening, while that induced by A₁ receptor stimulation is wholly dependent on K_ATP channel opening.

4. These results, together with previous evidence that hypoxia-induced vasodilatation in skeletal muscle is largely mediated by adenosine acting on A₁ receptors, lead us to propose that adenosine is released from endothelium during systemic hypoxia and acts on endothelial A₁ receptors to open K_ATP channels on the endothelial cells and cause synthesis of NO, which then acts on the vascular smooth muscle to cause dilatation. During severe systemic hypoxia we propose that adenosine may also act on A_2A receptors on the endothelium to cause dilatation by a similar process and may act on A_2A receptors on the vascular smooth muscle to cause dilatation by opening K_ATP channels.

In previous studies we have provided evidence, by using adenosine receptor antagonists, that a major part of the vasodilatation induced in skeletal muscle of the rat during systemic hypoxia is mediated by adenosine (Neylon & Marshall, 1991). From the effects of glibenclamide, which inhibits ATP-sensitive K⁺ (K_ATP) channels, we argued that a part of the hypoxia-induced muscle vasodilatation, particularly the early part, is dependent on the opening of K_ATP channels (Marshall et al. 1993). On the other hand, since L-NAME (nitro-L-arginine methyl ester), which blocks the synthesis of nitric oxide (NO), greatly reduced both the hypoxia-induced muscle vasodilatation and that evoked by infusion of adenosine, we concluded that the adenosine that is released in skeletal muscle during systemic hypoxia evokes dilatation in a largely NO-dependent manner (Skinner & Marshall, 1996). Most recently, we have shown by using selective agonists and antagonists of the A₁ and A_2A adenosine receptor subtypes, that both receptor subtypes are present in skeletal muscle and can evoke dilatation when stimulated, but that only the A₁ receptors are involved in the muscle vasodilatation of systemic hypoxia (Bryan & Marshall, 1999).

Vasodilatation might be produced in skeletal muscle by the opening of K_ATP channels on the skeletal muscle fibres, vascular smooth muscle or endothelial cells. These might be coupled to A₁ or A_2A receptors and either of these receptor subtypes might increase NO synthesis by the endothelium. In fact, from the effects of systemic hypoxia and adenosine infusion on the K⁺ concentration in the venous efflux of skeletal muscle ([K⁺]_v), and the effects of glibenclamide and the adenosine receptor antagonist 8-phenyltheophylline (8-PT) on these changes, we argued that adenosine released during systemic hypoxia opens K_ATP channels on skeletal muscle fibres that are coupled to adenosine receptors, so releasing K⁺ and that this K⁺ contributes to the hypoxia-induced dilatation (Marshall et al. 1993). However, in our recent experiments, A_2A receptor blockade had no effect on hypoxia-induced muscle vasodilatation, but reduced K⁺ efflux from muscle, while A₁ and A_2A receptor agonists both caused muscle vasodilatation without affecting muscle K⁺ balance (Bryan & Marshall, 1999). Thus, it seems likely that the K_ATP channels that are most important in the hypoxia- and adenosine-induced muscle vasodilatation may be on the vascular smooth muscle or endothelium, rather than on the skeletal muscle fibres. In fact, in pig coronary artery smooth muscle in vitro, adenosine can open K_ATP channels and induce vasorelaxation via A₁ receptors (Merkel et al. 1992; Dart & Standen, 1993). Further, the dilatation evoked by exogenous adenosine in arterioles of diaphragm muscle in vivo was largely mediated by A₁ receptors and was apparently dependent on the opening of K_ATP channels (Danialou et al. 1997), although these experiments did not discriminate between the opening of K_ATP channels on the vascular smooth muscle and endothelium. By contrast, in mesenteric artery myocytes, adenosine activated K_ATP channels via A_2A receptors, but not via A₁ receptors (Kleppisch & Nelson, 1995).

As far as NO is concerned, adenosine has been shown to increase the synthesis and release of NO by endothelial cells in vitro (Li et al. 1995; Sobrevia et al. 1997), via A_2A receptors (Sobrevia et al. 1997). Similarly, dilatation induced by stimulation of A₂ receptors in pig and guinea-pig coronary arteries in vitro was shown to be endothelium and NO dependent (Vials & Burnstock, 1993; Abebe et al. 1995). However, in diaphragm arterioles in vivo, the dilatation evoked by an A₁ receptor agonist was found to be NO dependent (Danialou et al. 1997).

Since it is known that endothelial cells possess K_ATP channels (Janigro et al. 1993; Langheinrich & Daut, 1997) and since adenosine and other agonists that increase NO synthesis cause hyperpolarization of endothelial cells (Mehrke & Daut, 1990; Sobrevia et al. 1997), it is a reasonable hypothesis that the hyperpolarization that is associated with NO synthesis is at least partly dependent on the opening of endothelial K_ATP channels. Accordingly, the endothelium- and NO-dependent component of the dilatation induced in pig coronary arterioles in vitro by adenosine was blocked by glibenclamide (Kuo & Chancellor, 1995). On the other hand, evidence has also been presented for some arteries in vitro, that NO can cause hyperpolarization of vascular smooth muscle cells by opening K_ATP channels (e.g. Murphy & Brayden, 1995) so raising the possibility that this process contributes to vasodilatation in vivo.

In view of these varied and, to some extent, disparate findings that were largely made in vitro, the aim of the present study was to attempt to establish the mechanisms by which adenosine and its receptor subtypes may induce vasodilatation in skeletal muscle vasculature in vivo and by implication, to elucidate the mechanisms by which adenosine produces muscle vasodilatation during systemic hypoxia. To this end, we tested the effects of glibenclamide and L-NAME upon muscle vasodilatation induced by specific A₁ and A_2A receptor agonists and by the NO donor sodium nitroprusside (SNP), in the same experiments in which we tested the effects of these antagonists on the responses induced by systemic hypoxia and by exogenous adenosine. Since the A₁ and A_2A agonists have long-lasting effects on the cardiovascular system and since we could not be sure, in anticipation, that either L-NAME or glibenclamide would completely reverse their effects we tested the responses evoked by the A₁ or A_2A agonists after administration of L-NAME or glibenclamide in the present study and compared them with those evoked in our previous experiments before any antagonist (Bryan & Marshall, 1999).

Some of the results of the present paper have already been reported in brief (Bryan & Marshall, 1996, 1997).

METHODS

Experiments were performed on male Wistar rats anaesthetized and prepared for recording as described previously (Neylon & Marshall, 1991; Bryan & Marshall, 1999). Briefly, anaesthesia was induced with 3.5 % halothane in O₂ and maintained with a continuous infusion of Saffan (7–12 mg kg⁻¹ h⁻¹i.v. during surgery and 5–8 mg kg⁻¹ h⁻¹i.v. during the experimental period; Pitman-Moore Ltd, Uxbridge, Middlesex, UK). During the experimental period, the level of anaesthesia was adjusted so that pinching the paw evoked a sluggish withdrawal reflex and a transient rise in arterial pressure of 10–15 mmHg. Cannulae were placed in the brachial artery for recording arterial blood pressure (ABP) and heart rate (HR), in the right femoral artery for removing blood samples for analysis of the partial pressures of O₂ and CO₂ (P_a,O2 and P_a,CO2) and pH and in the ventral tail artery for infusion of drugs (see below). Blood flow was recorded from the left femoral artery (FBF) with an electromagnetic flow probe after the arterial and venous supply of that limb had been isolated; femoral vascular conductance (FVC) was computed on-line as FBF/ABP. In addition, the trachea was cannulated with a T-piece which was connected to tubing through which atmospheric air, or 8 % O₂, was directed via an air pump or rotameter system. All variables were collected to computer and displayed on a monitor. All animals were allowed to stabilize at the experimental level of anaesthesia (see above) for at least 45 min before the experimental protocol began.

Experimental protocols

Group 1

In seven rats (body weight, 310 ± 5 g; mean ±s.e.m.), the responses evoked by a 5 min period of systemic hypoxia (breathing 8 % O₂), a 5 min infusion of adenosine (1.2 mg kg⁻¹ min⁻¹i.a.) and a 5 min infusion of SNP (0.016 mg kg⁻¹ min⁻¹i.a.) were recorded before and after administration of L-NAME (10 mg kg⁻¹i.v.). Periods of at least 10 min were allowed between these stimuli for stabilization of all variables. The infusion rates of adenosine and SNP were chosen so that they induced an increase in FVC comparable to that induced by hypoxia (see also Bryan & Marshall, 1999). Arterial samples for analysis of blood gases were taken during air breathing and in the 5th minute of hypoxia. Following the infusion of SNP after L-NAME, when cardiovascular variables had stabilized, a 5 min infusion of the selective A₁ receptor agonist CCPA (2-chloro-N⁶-cyclopentyladenosine; 0.35 μg kg⁻¹ min⁻¹i.a.) was given as in our previous study (Bryan & Marshall, 1999). CCPA was dissolved in 3 % dimethylsulphoxide (DMSO) in saline: this vehicle does not affect cardiovascular variables (see Bryan & Marshall, 1999).

Group 2

In six rats (316 ± 7 g), a similar protocol to that for Group 1 was used except that the selective A_2A receptor agonist CGS 21680 (2-p-(2-carboxyethyl)-phenethylamino-5′-N-ethylcarboxamidoadenosine hydrochloride; 1.2 μg kg⁻¹ min⁻¹i.a.; see Bryan & Marshall, 1999) was given instead of CCPA, the vehicle for CGS 21680 being the same as for CCPA (see above).

Group 3

In nine rats (308 ± 8 g), the same protocol as for Group 1 was used except that glibenclamide (10 mg kg⁻¹i.v.) was given instead of L-NAME. The vehicle for glibenclamide was 100 % polyethylene glycol 400 (dissolved at 10 mg ml⁻¹)-0.1 M NaOH (50/50 v/v). To avoid undesirable effects of the vehicle, the drug and vehicle were infused gradually over 4 min and then flushed in with saline. Following the final infusion of SNP, CGS 21680 was infused as in Group 2.

Group 4

In nine rats (321 ± 7 g), the responses evoked by a 5 min period of breathing 8 % O₂ were recorded as in Group 1. Levcromakalim (1 mg kg⁻¹ min⁻¹) was then infused into the brachial vein and when the cardiovascular responses had reached a plateau, after approximately 45 min, and whilst the levcromakalim infusion was continued, the vehicle for glibenclamide was infused over 4 min. Following this, glibenclamide at 10 mg kg⁻¹i.v. was infused over 4 min and when the variables had stabilized again a further 10 mg kg⁻¹ was infused over 4 min. At this point the levcromakalim infusion was stopped and the responses evoked by a 5 min period of breathing 8 % O₂ were re-tested. Finally, CCPA was infused as in Group 1 at 0.35 μg kg⁻¹ min⁻¹i.a.

Group 5

In view of the results obtained in Groups 3 and 4 (see below) a further set of experiments was performed on eight rats (312 ± 6 g) in which CCPA was infused at 0.35 μg kg⁻¹ min⁻¹i.a. (as in Groups 1 and 4) after glibenclamide had been given at 10 mg kg⁻¹i.v. When all the variables had stabilized a further dose of glibenclamide was given (10 mg kg⁻¹i.v.) resulting in a cumulative dose of 20 mg kg⁻¹ and when the variables had again stabilized, CGS 21680 was infused at 1.2 μg kg⁻¹ min⁻¹ (as in Groups 2 and 3).

CCPA and CGS 21680 were supplied by Research Biochemicals Inc., adenosine, glibenclamide and L-NAME by Sigma, while SNP was supplied by BDH Chemicals Limited. Levcromakalim was a gift from Smithkline Beecham Pharmaceuticals.

At the end of the experiment, the animals were killed with an overdose of Saffan.

Statistical analyses

All results are expressed as means ±s.e.m. For the purposes of analysing the effects of L-NAME on responses induced by 8 % O₂, adenosine and SNP, the results obtained in Groups 1 and 2 were considered together. Responses evoked by 8 % O₂ before and after L-NAME, or glibenclamide, were compared at 45 s, and 1, 2 and 5 min of the period of hypoxia as changes from baseline by two-way ANOVA, followed by Scheffé‘s test for comparisons at different time points (Bryan & Marshall, 1999). The responses evoked by adenosine and SNP before and after the antagonists were analysed by comparing the differences between the values recorded before and at the 5th minute of infusion both before and after administration of the antagonist, using Student's paired t test. As in our previous study (Bryan & Marshall, 1999), since both the concentration and infusion rate of the agonists were adjusted in individual animals to achieve the same final dose per body weight, no attempt was made to compare the changes at different time points during the infusion. Similarly, the responses evoked by CCPA and CGS 21680 before and after the antagonists were analysed by comparing the differences between the values recorded before and at the 5th minute of infusion in the absence of the antagonist in our previous study (Bryan & Marshall, 1999) with those recorded before and at the 5th minute of infusion after the antagonist in the present study using Student's unpaired t test. Blood gas values during air breathing and 8 % O₂ were compared using Student's paired t test. In all cases, P < 0.05 was considered statistically significant.

RESULTS

Effects of L-NAME

In Groups 1 and 2, administration of L-NAME caused a substantial increase in the baseline level of ABP (P < 0.001), a decrease in HR (P < 0.01) and a decrease in the baseline level of FVC (P < 0.001) indicating peripheral vasoconstriction (Fig. 1A and B). Under these conditions, the fall in ABP induced by 8 % O₂ was not significantly different from that induced before L-NAME, but the increase in FVC was significantly smaller than that after L-NAME (Fig. 1A). L-NAME had no effect on the blood gas values during air breathing or during 8 % O₂ (Table 1). Similarly, L-NAME reduced the increase in FVC induced by adenosine infusion, but had no significant effect on the simultaneous fall in ABP (Fig. 1B). These results are comparable to those of our previous study (Skinner & Marshall, 1996).

Figure 1. Effects of L-NAME on cardiovascular responses evoked by systemic hypoxia (A; breathing 8 % O₂ for 5 min) and adenosine (B; 1.2 mg kg⁻¹ min⁻¹; i.a. infusion for 5 min).

A, means ±s.e.m. recorded at the times indicated; hypoxia began at time 0. ▾, control values; •, values after L-NAME (10 mg kg⁻¹i.v.). B, means ±s.e.m. recorded before (0), and at the 5th minute of infusion as indicated. □, control; , after L-NAME. In B, ***P < 0.001, *P < 0.05, significant difference between values recorded at 0 and 5 min. ²²²P < 0.001, ²²P < 0.01, significant difference between baseline values (at 0 min) before and after L-NAME. For hypoxia, the corresponding values are indicated in the text. In A and B, †††P < 0.001, significant difference between change recorded during hypoxia or adenosine infusion before and after L-NAME. Abbreviations: ABP, arterial blood pressure; HR, heart rate; FBF, femoral blood flow; FVC, femoral vascular conductance.

Table 1.

Blood gas values and arterial pH (pH_a) recorded during air-breathing and at the 5th minute of hypoxia (8% O₂) before and after administration of L-NAME or glibenclamide at 10 or 20 mg kg⁻¹ I.V.

	Pa,O2(mmHg)	Pa,CO2(mmHg)	pHa
Before L-NAME
Air	82.4 ± 2	45.1 ± 1	7.40 ± 0.01
8% O2	39.1 ± 2	35.7 ± 1	7.46 ± 0.01
After L-NAME (10 mg kg−1)
Air	84.7 ± 2	44.6 ± 1	7.38 ± 0.01**
8% O2	40.2 ± 2	35.6 ± 1	7.43 ± 0.01***
Before glibenclamide
Air	84.9 ± 1	48.1 ± 1	7.34 ± 0.01
8% O2	33.5 ± 2	35.0 ± 2	7.41 ± 0.01
After glibenclamide (10 mg kg−1)
Air	90.1 ± 3	50.8 ± 2	7.32 ± 0.02
8%O2	35.1 ± 1	37.7 ± 2	7.38 ± 0.02*
Before glibenclamide
Air	87.6 ± 5	49.7 ± 2	7.39 ± 0.02
8% O2	34.6 ± 1	34.6 ± 1	7.49 ± 0.02
After glibenclamide (20 mg kg−1)
Air	90.6 ± 3	51.8 ± 3	7.38 ± 0.01
8% O2	32.2 ± 1	41.1 ± 3	7.45 ± 0.02*

P_a,O2, arterial O₂ pressure; P_a,CO2, arterial CO₂ pressure. Values are means ± S.E.M.

^***P < 0.001

^**P < 0.01

^*P < 0.05

significant difference between values recorded before and after drug.

By contrast, administration of L-NAME did not reduce either the fall in ABP, or increase in FVC induced by SNP; in fact the fall in ABP induced by SNP was greater after L-NAME (Fig. 2). SNP also induced a fall in HR before L-NAME, but not after.

Figure 2. Effects of L-NAME on cardiovascular responses evoked by SNP (0.016 mg kg⁻¹ min⁻¹; i.a. infusion for 5 min).

Means ±s.e.m. at 0 and 5 min as indicated. □, control; , after L-NAME (10 mg kg⁻¹i.v.). ***P < 0.001, **P < 0.01, significant difference between values recorded at 0 and 5 min. ‡‡‡P < 0.001, ‡‡P < 0.01, significant difference between baseline values (at 0 min) before and after L-NAME. ††P < 0.01, significant difference between change recorded during SNP infusion before and after L-NAME. Abbreviations as in Fig. 1.

In Group 1, the A₁ agonist CCPA, infused after L-NAME, had no effect on either ABP or FVC (Fig. 3A). This contrasts with the fall in ABP and increase in FVC that was evoked by CCPA in our previous study (Fig. 3A; Bryan & Marshall, 1999). Nevertheless, the fall in HR induced by CCPA before and after L-NAME was comparable (Fig. 3A).

Figure 3. Effect of L-NAME on cardiovascular responses evoked by the A₁ receptor agonist CCPA (A; 0.35 μg kg⁻¹ min⁻¹) and the A_2A receptor agonist CGS 21680 (B; 1.2 μg kg⁻¹ min⁻¹) (i.a. infusion for 5 min in each case).

Means ±s.e.m. recorded before (0) and at the 5th min of infusion as indicated. In A and B:▪, control data recorded in experiments of Bryan & Marshall (1999); , values recorded after L-NAME (10 mg kg⁻¹i.v.). ***P < 0.001, **P < 0.01, significant difference between values recorded at 0 and 5 min; †††P < 0.001, ††P < 0.01, †P < 0.05, significant difference between change evoked by agonist before and after L-NAME. Abbreviations as in Fig. 1.

In Group 2, the A_2A agonist CGS 21680, infused after L-NAME, evoked a fall in ABP and tended to increase FVC (P= 0.08; FVC increased in 5 of the 6 rats). However, the change in ABP and the increase in FVC were smaller than that evoked by CGS 21680 in the absence of L-NAME in our previous study (see Fig. 3B; Bryan & Marshall, 1999).

Effects of glibenclamide at 10 mg kg⁻¹

In Group 3, glibenclamide at 10 mg kg⁻¹i.v. had no effect on the baseline level of ABP but reduced the baseline levels of HR, FBF and FVC (P < 0.001, P < 0.05 and P < 0.05, respectively; Fig. 4A). Glibenclamide at 10 mg kg⁻¹ significantly reduced the fall in ABP and increase in FVC evoked by 8 % O₂ over the 5 min of hypoxia, as tested by ANOVA. The effects on the early part of these responses were most obvious (see Fig. 4A), as we have reported previously (Marshall et al. 1993), although in the present study these effects just failed to reach statistical significance (P= 0.06 and P= 0.09 for the difference between the ABP and FVC values recorded at 45 s and 1 min, respectively, before and after glibenclamide). Glibenclamide had no effect on the blood gas values recorded during air breathing or during 8 % O₂ (Table 1).

Figure 4. Effect of the K_ATP channel antagonist glibenclamide at 10 mg kg⁻¹ (A) or 20 mg kg⁻¹ (B) i.v. on responses evoked by systemic hypoxia (8 % O₂ for 5 min).

Means ±s.e.m. recorded at the times indicated; hypoxia began at time 0. ▾, control values; •, values after glibenclamide. †††P < 0.001, ††P < 0.01, †P < 0.05, significant difference between change recorded during hypoxia before and after glibenclamide. Abbreviations as in Fig. 1.

On the other hand, glibenclamide (10 mg kg⁻¹) had no effect on the fall in ABP induced by adenosine infusion, but considerably reduced the increase in FVC (Fig. 5A), such that any change in FVC after glibenclamide did not reach statistical significance. By contrast, glibenclamide (10 mg kg⁻¹) had no effect on either the fall in ABP, or the increase in FVC induced by infusion of SNP (Fig. 5B).

Figure 5. Effect of the K_ATP channel antagonist glibenclamide (10 mg kg⁻¹i.v.) on responses evoked by adenosine (A; 1.2 mg kg⁻¹ min⁻¹) and SNP (B; 0.016 mg kg⁻¹ min⁻¹) (i.a. infusion for 5 min in each case).

Means ±s.e.m. recorded before (0) and at the 5th minute of infusion as indicated. □, control; , after glibenclamide. ***P < 0.001, **P < 0.01, *P < 0.05, significant difference between values recorded at 0 and 5 min. †††P < 0.001, ††P < 0.01, significant difference between change recorded during adenosine infusion before and after glibenclamide. Abbreviations as in Fig. 1.

Considering the effects of the selective A₁ and A_2A agonists, in Group 5, the A₁ agonist CCPA, infused after glibenclamide, still evoked a fall in ABP and a decrease in HR that were comparable to those produced in our previous study in the absence of glibenclamide (Fig. 6A; Bryan & Marshall, 1999), but it had no significant effect on FVC. Thus, the effect of CCPA on FVC was significantly different from that seen in the absence of glibenclamide (Fig. 6A). In Group 3, the A_2A agonist CGS 21680, infused after glibenclamide, still induced a fall in ABP and an increase in FVC, but these responses were considerably smaller than those evoked by CGS 21680 in the absence of glibenclamide in our previous study (Fig. 6B; Bryan & Marshall, 1999).

Figure 6. Effect of the K_ATP channel antagonist glibenclamide (10 mg kg⁻¹i.v.) on responses evoked by the A₁ receptor agonist CCPA (A; 0.35 μg kg⁻¹ min⁻¹) and the A_2A receptor agonist CGS 21680 (B; 1.2 μg kg⁻¹ min⁻¹) (i.a. infusion for 5 min in each case).

Means ±s.e.m. recorded before (0) and at the 5th minute of infusion as indicated. ▪, control data recorded in experiments of Bryan & Marshall (1999); , values recorded after glibenclamide. ***P < 0.001, **P < 0.01, *P < 0.05, significant difference between values recorded at 0 and 5 min. ††P < 0.01, †P < 0.05, significant difference between change recorded during CCPA or GCS 21680 infusion before and after glibenclamide. Abbreviations as in Fig. 1.

Effects of glibenclamide on responses evoked by levcromakalim

In Group 4, infusion of the K_ATP channel opener levcromakalim induced a fall in ABP, an increase in HR and an increase in FVC, that were comparable in magnitude to changes induced by hypoxia (see Fig. 7). Infusion of the vehicle for glibenclamide had a small effect on ABP (P < 0.01), but no effect on FVC (Fig. 7). Glibenclamide at 10 mg kg⁻¹i.v. reversed the fall in ABP induced by levcromakalim, but any effect it had on FVC did not reach statistical significance. By contrast, subsequent administration of glibenclamide to give a total dose of 20 mg kg⁻¹ fully reversed the increase in HR and the increase in FVC (Fig. 7).

Figure 7. Effects of the K_ATP channel antagonist glibenclamide on cardiovascular changes evoked by the K_ATP channel opener levcromakalim.

Means ±s.e.m.□, control values; ▪, values recorded at plateau of response to levcromakalim infusion (1 mg kg⁻¹ min⁻¹i.v.); , values recorded after vehicle for glibenclamide; , values recorded after glibenclamide at 10 mg kg⁻¹i.v.;, values recorded after a further dose of glibenclamide at 10 mg kg⁻¹i.v., to give a final dose of 20 mg kg⁻¹. ***P < 0.001, **P < 0.01, significant effect of levcromakalim. †††P < 0.001, ††P < 0.01, significant change from value recorded during levcromakalim infusion. Abbreviations as in Fig. 1.

Effects of glibenclamide at 20 mg kg⁻¹

In Group 4, administration of glibenclamide at 20 mg kg⁻¹ had a similar effect on the cardiovascular responses evoked by 8 % O₂ to that seen after administration of glibenclamide at 10 mg kg⁻¹i.v. in Group 3 (see Fig. 4B). Notably, the fall in ABP and increase in FVC were reduced, the values recorded at 45 s and 1 min of hypoxia being preferentially affected (Fig. 4B); P < 0.05 and P < 0.001 for ABP and FVC, respectively, recorded at same time points before and after drug. Glibenclamide at 20 mg kg⁻¹ also had similar effects on the responses evoked by infusion of the A₁ agonist CCPA as did glibenclamide at 10 mg kg⁻¹ (see Fig. 6A); the fall in ABP evoked by CCPA was significantly smaller after glibenclamide at 20 mg kg⁻¹ than that seen in the absence of glibenclamide (P < 0.05). ABP fell from 125 ± 5 to 119 ± 4 mmHg, while FVC did not change significantly (0.0064 ± 0.0009 before and 0.0069 ± 0.0011 ml min⁻¹ mmHg⁻¹ at the 5th minute of CCPA infusion; cf. Fig. 6A). Further, in Group 5, infusion of the A_2A agonist CGS 21680, after glibenclamide at 20 mg kg⁻¹, produced similar changes to those seen after glibenclamide at 10 mg kg⁻¹ in Group 3 (see Fig. 6B). In fact, after glibenclamide at 20 mg kg⁻¹, CGS 21680 produced a larger increase in FVC (from 0.0091 ± 0.0006 to 0.0154 ± 0.0012 ml min⁻¹ mmHg⁻¹) than that obtained after administration of glibenclamide at 10 mg kg⁻¹ (cf. Fig. 6B).

DISCUSSION

In the present study on the rat, the NO synthesis inhibitor L-NAME not only increased baseline ABP and decreased baseline FVC, but also greatly reduced the fall in ABP and increase in FVC induced both by systemic hypoxia (breathing 8 % O₂) and by infused adenosine, as we have shown previously (Skinner & Marshall, 1996). Our new findings are that L-NAME (i) did not reduce the fall in ABP and increase in FVC that were evoked by the NO donor SNP, but that (ii) it abolished the fall in ABP and increase in FVC evoked by the adenosine A₁ receptor agonist CCPA, and (iii) greatly reduced the fall in ABP and concomitant increase in FVC evoked by the adenosine A_2A receptor agonist CGS 21680. In addition, the present study confirmed our previous finding that the K_ATP channel inhibitor glibenclamide reduced the increase in FVC evoked by systemic hypoxia with a preferential effect on the early part of the response (Marshall et al. 1993). Our new findings are that glibenclamide (i) greatly reduced the increase in FVC evoked by infused adenosine, but (ii) had no effect on that induced by SNP. It also (iii) abolished the increase in FVC evoked by CCPA, and (iv) greatly reduced that evoked by CGS 21680.

The effects of L-NAME

Since the fall in ABP and increase in FVC evoked by SNP persisted after L-NAME, the former even being enhanced (see Moncada et al. 1991), this reinforces our view that the ability of L-NAME to reduce the hypoxia- and adenosine-induced responses was due to blockade of NO synthesis rather than to non-specific effects on baseline (see Skinner & Marshall, 1996). We have already argued that adenosine released during systemic hypoxia induces dilatation predominantly by acting on adenosine receptors on the endothelium rather than on the vascular smooth muscle (Mian & Marshall, 1995; Skinner & Marshall, 1996; Bryan & Marshall, 1999). Thus, it is likely that the effects of L-NAME on the hypoxia- and adenosine-induced dilatation were due to blockade of NO synthesis by the endothelium rather than by neurones or other cells. Similarly, the fact that the A₁ receptor agonist CCPA had no effect on ABP or FVC after L-NAME, strongly suggests that its vasodilator effect in skeletal muscle is NO dependent, as was found for dilatation evoked by A₁ receptor stimulation in arterioles of the diaphragm muscle (Danialou et al. 1997). Taken together, these findings are consistent with our proposal that, at least in the rat, the component of hypoxia-induced muscle vasodilatation that is due to adenosine is mediated by A₁ receptors and that these receptors are on the vascular endothelium (Bryan & Marshall, 1999). It might be argued that L-NAME achieved its effects by reducing tonic levels of NO and thereby cGMP, so attenuating the known synergism between cGMP and cAMP in mediating vasodilatation (Dewitt et al. 1994). Whilst this may have contributed, it is unlikely to be the full explanation given the evidence that there are adenosine receptors on endothelial cells that can increase the synthesis of NO (Li et al. 1995; Sobrevia et al. 1997).

Since the muscle vasodilator response to the A_2A receptor agonist CGS 21680 was reduced by L-NAME, this indicates that there are also A_2A receptors on the endothelium of muscle vasculature that induce dilatation in an NO-dependent manner. This contrasts with the observation that CGS 21680 had no effect on the diameter of diaphragm arterioles (Danialou et al. 1997), but is consistent with evidence from coronary arterioles in vitro, that dilatation mediated by an A₂ receptor agonist was NO dependent (Vials & Burnstock, 1993; Abebe et al. 1995), and with evidence that adenosine can stimulate NO synthesis by endothelial cells via A_2A receptors (Sobrevia et al. 1997).

Taken together, the effects of L-NAME on the muscle vasodilator responses evoked by the A₁ and A_2A receptor agonists and by infused adenosine are compatible with our proposal that ∼50 % of the muscle vasodilatation induced by infusion of adenosine was mediated by A₁ receptors and about 50 % by A_2A receptors (Bryan & Marshall, 1999), and with the idea that the A₁ and A_2A receptors that mediate this response are largely on the endothelium. Nevertheless, given that both adenosine itself and the A_2A agonist CGS 21680 could still evoke a small increase in FVC after L-NAME, it seems that infused adenosine can stimulate A_2A receptors to induce muscle vasodilatation in an NO-independent manner. Baker & Sutton (1993) similarly concluded that, when adenosine reaches muscle arterioles in the blood stream, that is from the adluminal surface, it acts largely in an NO-dependent manner via endothelial receptors, but to a small extent in an NO-independent manner.

Adenosine might stimulate A_2A receptors on the skeletal muscle fibres to induce vasodilatation in an indirect way, by opening K_ATP channels and releasing K⁺ which then relaxes vascular smooth muscle (Marshall et al. 1993; Bryan & Marshall, 1999). However, since our evidence suggests this is a minor effect (Bryan & Marshall, 1999), the obvious conclusion is that the NO-independent component of the vasodilatation is mainly mediated by A_2A receptors on the vascular smooth muscle. This accords with much previous evidence that stimulation of A₂ receptors on vascular smooth muscle induces vasodilatation by increasing cAMP levels (Olsson & Pearson, 1990; for further discussion see below). It is also consistent with the proposal that the A_2A receptors that mediate the vasodilator response to skeletal muscle contraction (Poucher, 1996) are on the vascular smooth muscle (see Bryan & Marshall, 1999).

The effects of glibenclamide

In our previous study, glibenclamide at 10 and 20 mg kg⁻¹ had similar effects on the muscle vasodilatation and increase in [K⁺] in the venous blood of muscles ([K⁺]_v) to those induced by systemic hypoxia (Marshall et al. 1993). Therefore, to avoid non-specific effects of glibenclamide, we initially decided to perform the present study with a dose of 10 mg kg⁻¹. It should be noted that, in the rat, glibenclamide at 0.1–0.2 mg kg⁻¹ is sufficient to block pancreatic K_ATP channels and stimulate the release of insulin (Boosboon et al. 1973) and that insulin can induce muscle vasodilatation (McKay & Hester, 1996). Further, high concentrations of glibenclamide in vitro can block Ca²⁺-activated K⁺ channels (Gelband & McCullough, 1993). However, we were aware from another study (Smits et al. 1997) that even though 10 mg kg⁻¹ glibenclamide was sufficient to abolish a fall in ABP evoked by infusion of the K_ATP channel opener levcromakalim, which was equivalent in magnitude to that induced by systemic hypoxia in the present study, the concurrent increase in hindlimb vascular conductance, which was also equivalent in magnitude to the hypoxia-induced increase in FVC, was only partly reduced; a dose of 20 mg kg⁻¹ glibenclamide was required to abolish this response. Thus, we performed some of the present experiments with glibenclamide at a dose of 20 mg kg⁻¹.

In fact, glibenclamide at 10 mg kg⁻¹ did not affect the increase in FVC induced by our test infusion of levcromakalim but, in agreement with Smits et al. (1997), glibenclamide at 20 mg kg⁻¹ fully reversed this response. However, we confirmed that glibenclamide at 10 and 20 mg kg⁻¹ had similar effects on the hypoxia-induced increase in FVC, as in our previous study (Marshall et al. 1993). Moreover, we found that glibenclamide at 10 mg kg⁻¹ almost abolished the increase in FVC evoked by our test infusion of adenosine, while glibenclamide at 20 mg kg⁻¹ had no greater effect on the increase in FVC evoked by the selective A₁ and A_2A agonists than that observed when given at 10 mg kg⁻¹. Thus, the proposals we can make are the same whether we use the results we obtained with glibenclamide at a dose of 10 or 20 mg kg⁻¹.

Our finding that glibenclamide almost abolished the muscle vasodilatation evoked by infused adenosine, suggests it was mainly dependent on the opening of K_ATP channels. This is consistent with observations on rat diaphragm arterioles in vivo and porcine coronary arterioles in vitro (Kuo & Chancellor, 1995; Danialou et al. 1997). Since we have already argued that this response was predominantly NO dependent (see above and Skinner & Marshall, 1996), it seems reasonable to propose that the K_ATP channels that are particularly important are on the endothelium and that stimulation of adenosine receptors on the endothelial cells opens K_ATP channels so leading to hyperpolarization and stimulation of NO synthesis.

We are aware that dilatation induced by the K_ATP channel opener levcromakalim, in isolated mesenteric arterial preparations with intact endothelium, was enhanced by blockade of NO synthesis with L-NAME (McCulloch & Randall, 1996; White & Hiley, 1997) and that this seems to argue against our proposal. However, the interpretation of this finding is not straightforward as K_ATP channel openers may also cause the release of endothelium-derived hyperpolarizing factor (EDHF) and the actions of EDHF and NO may interfere with one another (White & Hiley, 1997). Moreover, in in vivo studies on rats, L-NAME was able to significantly reduce the vasodilatation evoked in mesenteric circulation by levcromakalim (Gardiner et al. 1991), which is consistent with our proposal. Further, as indicated in the Introduction, there is direct evidence that endothelial cells have K_ATP channels (Janigro et al. 1993; Langheinrich & Daut, 1997) and that adenosine acts on endothelial cells to increase NO synthesis by causing hyperpolarization (Sobrevia et al. 1997). Indeed, in an independent series of pharmacological experiments on pig coronary arterioles, Kuo & Chancellor (1995) were led to the same proposal, namely that dilatation caused by adenosine is partly mediated by synthesis of NO that is triggered by the opening of endothelial K_ATP channels. Since, in the present study, glibenclamide had no effect on the muscle vasodilatation induced by the NO donor SNP this indicates that, in skeletal muscle vasculature, NO does not induce dilatation by opening K_ATP channels on the vascular smooth muscle, as was shown for some arteries in vitro (see Murphy & Brayden, 1995).

Since the dilatation evoked by infused adenosine was not totally dependent on NO (see above), the blocking effect of glibenclamide also allows the possibility that part of this response was mediated by the opening of K_ATP channels on the vascular smooth muscle, as proposed for pig coronary arteries in vitro (Merkel et al. 1992; Dart & Standen, 1993; Kuo & Chancellor, 1995). However, any muscle vasodilatation that remained as a response to adenosine after glibenclamide did not reach statistical significance. Thus, it seems that the ability of adenosine to cause dilatation by acting on receptors on the vascular smooth muscle and increasing cAMP levels (see above) was also mainly dependent on K_ATP channel opening. This is compatible with the evidence of Kleppisch & Nelson (1995) on mesenteric artery myocytes, that adenosine can, by acting on A_2A receptors, increase intracellular cAMP, stimulate protein kinase A and thence open K_ATP channels, most probably by a phosphorylation step.

Seen in this context, our finding that the muscle vasodilatation induced by the A_2A receptor agonist CGS 21680, which was largely NO dependent (see above), was greatly reduced by glibenclamide indicates that A_2A receptors on the endothelium of muscle vasculature are coupled to K_ATP channels and stimulate NO synthesis by inducing hyperpolarization. Evidence for this process has been found in that endothelial cells have K_ATP channels (Janigro et al. 1993; Langheinrich & Daut, 1997), while stimulation of A_2A receptors on human umbilical vein endothelial cells increases NO synthesis by hyperpolarizing the cells (Sobrevia et al. 1997). The A_2A receptor-mediated dilatation that is NO independent (see above) is consistent with an action on vascular smooth muscle which increases cAMP and opens the K_ATP channels, as shown by Kleppisch & Nelson (1995).

Since the muscle vasodilatation induced by the A₁ receptor agonist CCPA was not only abolished by L-NAME, but also by glibenclamide, this response can be attributed to the opening of K_ATP channels on the endothelium and consequent stimulation of NO synthesis; there is no need to implicate A₁ receptors on the vascular smooth muscle. This proposal is consistent with the observations of Danialou et al. (1997) that dilatation induced by an A₁ agonist in diaphragm arterioles in vivo was attenuated by glibenclamide as well as by an inhibitor of NO synthesis.

Given our evidence that the component of the hypoxia-induced muscle vasodilatation that is dependent on adenosine is mediated by A₁ receptors (Bryan & Marshall, 1999) and is largely NO dependent, the finding that glibenclamide preferentially reduced the early part of the hypoxia-induced dilatation (present study and Marshall et al. 1993) now allows us to propose that the hypoxia-induced dilatation is initiated by stimulation of A₁ receptors on the endothelium, which opens K_ATP channels and triggers the synthesis of NO. It is somewhat surprising that glibenclamide preferentially affected the early part of the hypoxia-induced dilatation, whereas the selective A₁ receptor antagonist DPCPX (8-cyclopentyl-1,3-dipropylxanthine), the non-selective A₁/A₂ receptor antagonist 8-sulphophenyltheophylline (8-SPT), and L-NAME each affected the early and later parts (Marshall et al. 1993; Skinner & Marshall, 1996; Bryan & Marshall, 1999). Furthermore, although we did not analyse the results in a quantitative manner (see Methods), glibenclamide seemed to be equally effective in reducing the early and late parts of the dilatation induced by infused adenosine (P. T. Bryan & J. M. Marshall, unpublished observations) as did 8-SPT and L-NAME (Skinner & Marshall, 1996). However, it may be that when the K_ATP channels are blocked, adenosine that is released under hypoxic conditions is able to stimulate A₁ receptors on the endothelium and stimulate NO synthesis by an alternative mechanism, for example by opening Ca²⁺-activated K⁺ channels (see Mehrke & Daut, 1990).

Summarizing the evidence we have obtained so far on the role of adenosine in the muscle vasodilatation of systemic hypoxia in the rat (Fig. 8), we propose that, during hypoxia, adenosine is released from the endothelial cells and stimulates A₁ receptors on the endothelial cells so opening K_ATP channels and stimulating the synthesis of NO, which then acts on the vascular smooth muscle to cause a large part of the dilatation. We further propose that adenosine produced at the boundaries of the skeletal muscle fibres stimulates adenosine receptors on the skeletal muscle fibres, which may be of A₁ and/or A_2A subtype, so opening K_ATP channels and causing K⁺ efflux, but suggest that this K⁺ makes only a minor contribution to the dilatation (Bryan & Marshall, 1999). Our results also indicate there are A_2A receptors on the vascular smooth muscle whose stimulation can induce dilatation. However, these apparently make no contribution to the hypoxia-induced dilatation, presumably because the local concentration of adenosine in the interstitium of resting hypoxic muscle is not sufficiently high; adenosine is ∼80-fold more potent at A₁ than A_2A receptors (Ueeda et al. 1991). Similarly, we propose that there are A_2A receptors on the endothelial cells that may open K_ATP channels and stimulate NO synthesis, but that these are not important in the hypoxia-induced dilatation we have studied; they may contribute in more severe hypoxia if the local concentration of adenosine produced in the vicinity of the endothelial receptors is sufficiently high.

Figure 8. Schematic diagram showing proposed sites, and mechanisms of action of adenosine in rat skeletal muscle vasculature.

Systemic hypoxia causes vascular endothelium to release adenosine which then acts on A₁ receptors on the endothelial cells that are coupled to K_ATP channels, thereby increasing the synthesis of NO which relaxes the vascular smooth muscle. In addition, systemic hypoxia causes adenosine to be released from skeletal muscle fibres and this may then act on A₁ and A_2A receptors on those fibres to open K_ATP channels and release K⁺ which is a vasodilator. Evidence suggests that A_2A receptors are also present on vascular smooth muscle and endothelium and can induce vasodilatation, but these are not stimulated during moderate systemic hypoxia.