Propionate-induced relaxation in rat mesenteric arteries: a role for endothelium-derived hyperpolarising factor

Abstract

Short chain fatty acids, including propionate, are generated in the caecum and large intestine, and when absorbed may elicit localised increases in intestinal blood flow. We sought to assess the mechanism by which propionate caused vasorelaxation. Propionate-mediated relaxation of noradrenaline-preconstricted rat mesenteric small arteries (RMSAs, i.d. 200–300 μm) was studied using small vessel myography. Propionate (1–30 mm) produced a concentration-dependent relaxation. Relaxation induced by 10 mm propionate (the approximate EC₅₀) was almost abolished by endothelial denudation, although a marked relaxation to a very high concentration of propionate (50 mm) persisted in the absence of the endothelium. In endothelium-intact RMSAs, relaxation to 10 mm propionate was almost abolished by elevating [K⁺]_o to 25 mm, but was unaffected by 100 μm N^ω-nitro-l-arginine methyl ester (l-NAME) (68 ± 4 vs. 66 ± 3 % in controls, n = 35), or by 1 μm indomethacin (60 ± 4 vs. 61 ± 7 % in controls, n = 15). In the presence of l-NAME, relaxation to 10 mm propionate was significantly and markedly (i.e. > 50 %) inhibited by 50 μm Ba²⁺ and by the combination of 100 nm charybdotoxin and 100 nm apamin. A similar effect on propionate-mediated relaxation was also exerted by 100 μm ouabain, and by the combination of 50 μm barium with ouabain. Relaxation was also significantly and markedly inhibited by pre-treatment of RMSAs with 100 nm thapsigargin or 10 μm cyclopiazonic acid (CPA). The results demonstrate that 10 mm propionate relaxes RMSAs via endothelium-derived hyperpolarising factor (EDHF). The observation that relaxation by propionate is inhibited by thapsigargin and CPA suggests that this action of propionate involves the release of endothelial cell Ca²⁺ stores.

Short chain fatty acids (SCFAs) are produced in large amounts in the caecum and large intestine, mainly as a result of the fermentation of carbohydrates. The main SCFAs produced are acetate, propionate, and butyrate, in a molar ratio of approximately 60:20:18, respectively (Cummings & MacFarlane, 1997). These SCFAs are absorbed into the portal circulation, where they act as vasodilators (Mortensen et al. 1994). The mean total SCFA acid concentration in the portal blood of sudden death victims has been measured to be 375 μm, of which ∼90 μm was propionate (Cummings et al. 1987). The local concentrations of SCFAs surrounding the microvasculature in the walls of these organs, which contribute only a fraction of the total portal vein blood flow, are, however, likely to be much higher (Mortensen et al. 1990).

Many types of vascular smooth muscle have been shown to be relaxed by SCFAs (Aalkjaer & Poston, 1996). We have previously demonstrated that lactate and butyrate relaxed noradrenaline-preconstricted rat mesenteric small arteries (RMSAs), an effect which occurred at high concentrations (EC₅₀ ∼22 mm), and was independent of the endothelium and of changes in the intracellular pH. We suggested that relaxation was mediated by activation of the cAMP second messenger system, since it was inhibited by Rp-cAMPs, an inhibitor of the binding of cAMP to protein kinase A (Aaronson et al. 1996; McKinnon et al. 1996). Alternative mechanisms for vasorelaxation by SCFAs have been proposed by other workers, including activation of the cGMP system (Omar et al. 1993) and intracellular acidification (Austin & Wray, 1994), which may act through multiple pathways (Austin & Wray, 2000).

In preliminary experiments with propionate, we observed that this SCFA, in contrast to lactate and butyrate, caused a relaxation of RMSAs which was abolished by removal of the endothelium. We therefore assessed in more detail the mechanism by which propionate relaxed RMSAs.

METHODS

Preparation of arteries

Male Wistar rats (200–300 g) were killed by stunning and cervical dislocation, in accordance with UK Home Office Schedule 1 protocols. Small resistance mesenteric arteries (i.d. approximately 300 μm, length 3–4 mm) were dissected free of surrounding fat and connective tissue and mounted as isometric preparations on a Mulvany-Halpern wire myograph (Danish Myo Technology, Aarhus, Denmark). Arteries were maintained in Krebs buffer containing (mm): NaCl, 119; KCl, 4.7; CaCl₂, 2.5; MgSO₄, 1.17; NaHCO₃, 25; KH₂PO₄, 1.18; EDTA, 0.026; glucose, 5.5; at 37 °C, gassed with 5 % CO₂−95 % O₂. Sodium propionate was prepared in Krebs buffer (by equimolar substitution for NaCl). Vessels were stretched to a circumference 90 % of that obtained when subjected to a transmural pressure of 13.4 kPa (Mulvany & Halpern, 1977) prior to a routine ‘run-up’ procedure consisting of four alternate contractions to 5 μm noradrenaline (NA) in Krebs buffer containing 70 mm KCl (equimolar substitution of KCl for NaCl), or 5 μm NA in normal Krebs solution. Endothelial viability was assessed by application of 1 μm acetylcholine following preconstriction with 2.5 μm NA (intact endothelium resulting in 75–100 % relaxation).

Experimental protocol

Arteries were pre-constricted with a concentration of NA approximating 75–80 % of maximum (2.5 μm in most experiments, occasionally 3.5 or 4.5 μm). Each experiment consisted of pairs of 20 min contractions to NA in normal Krebs buffer in which the bath solution was replaced after the initial 5 min, the first being a time control contraction wherein the solution was replaced with another containing NA, and the second being the test contraction wherein the solution was replaced with another containing both NA and propionate. The effects of K⁺ channel blockers and enzyme inhibitors, alteration of the extracellular K⁺ concentration, or removal of the endothelium on both time control contractions and contractions in the presence of propionate were examined. Unless otherwise stated, K⁺ channel blockers were applied when the solution was changed (i.e. immediately after the first 5 min of each contraction), whereas enzyme inhibitors (l-NAME, indomethacin, thapsigargin, and cyclopiazonic acid) were applied to the bath prior to the NA for at least 5 min pre-incubation. A 5 min rest period was allowed between contractions. The effects of inhibitors on time controls and propionate-induced relaxations were compared against their appropriate controls in the same arteries. Mechanical endothelium removal was achieved by rubbing a human hair through the lumen of the artery while mounted on the myograph, and confirmed by subsequent complete lack of relaxation to 1 μm acetylcholine in preconstricted arteries.

Statistics and data analysis

Data were expressed as means ± standard error of the mean (s.e.m.) and (unless otherwise stated) compared by Student's paired t test, whereby a P value of less than 0.05 denoted significant differences between means. Calculations were performed in Microsoft Excel and data were presented graphically using SigmaPlot 2000 (SPSS, Chicago, IL, USA).

Drugs and chemicals

BaCl₂ and the salts used to make Krebs buffer were obtained from either Sigma (Poole, UK) or BDH-Merck (Lutterworth, UK). Charybdotoxin, apamin, and iberiotoxin were from Bachem (St Helens, UK), and cyclopiazonic acid, thapsigargin, ouabain, indomethacin, acetylcholine, and N^ω-nitro-l-arginine methyl ester (l-NAME) were from Sigma. Noradrenaline was from Abbott Laboratories (Queenborough, UK).

RESULTS

As seen in Fig. 1A, 10 mm propionate caused an initial rapid, transient contraction above that already existing before the solution change, followed by a slower, more sustained relaxation lasting up to 10 min, after which the noradrenaline progressively reasserted its contractile effect. It is also apparent from time controls that replacing the NA-containing bath solution per se also caused an initial transient contraction followed by a small degree of gradual spontaneous relaxation. Tension recordings were processed by averaging all points in the traces at 1 min intervals, and then normalising the resultant values to express them as a fraction of the values recorded just before the solution change at 5 min (normalisation shown in Fig. 1B). Since all propionate responses were paired with their own time controls, it was then possible to correct responses to propionate for spontaneous time-related changes in contraction amplitude, as illustrated in Fig. 1C. As a measure of the amplitude of propionate-induced relaxation under various conditions (Table 1), the effects at the three time points around which maximum relaxation occurred were averaged. In most cases these were the points at 12, 13 and 14 min.

Figure 1. Responses to 10 mm sodium propionate compared to time control contractions.

A, examples of tension measurements in a single artery showing time control and effect of 10 mm propionate. The arrow indicates where the bath solution was changed and the dotted line represents basal tone. B, mean (± s.e.m.) contraction amplitude measured at 1 min intervals and normalised to amplitude at 5 min (n = 41). For each time point following the solution change, values ‘a’ and ‘b’ were calculated and, using the formula a/(a + b), propionate-induced relaxation was corrected for spontaneous time-related changes in contraction amplitude, as shown in C.

Table 1.

Modulation of propionate-induced relaxation

	% Relaxation
Inhibitor	Control	With inhibitor	n	P
l-NAME	67.6 ± 4.4	66.2 ± 2.8	35	N.S.
Indomethacin	59.7 ± 3.8	60.5 ± 6.6	15	N.S.
25 mm K+	67.7 ± 10.6	2.8 ± 4.4	5	< 0.01
Charybdotoxin and apamin	59.7 ± 5.5	27.5 ± 2.3	9	< 0.001
Barium	57.0 ± 7.0	27.4 ± 9.4	12	< 0.01
100 μm ouabain	66.6 ± 8.2	10.8 ± 1.7	6	< 0.001
Barium and 1 μm ouabain	71.9 ± 6.3	26.5 ± 7.0	8	< 0.0001
Barium and 100 μm ouabain	64.8 ± 5.6	25.7 ± 3.7	10	< 0.0001
Charybdotoxin and apamin and barium	63.9 ± 6.4	14.9 ± 2.4	10	< 0.0001
Iberiotoxin	55.2 ± 6.2	53.9 ± 6.8	13	N.S.
Thapsigargin	51.6 ± 7.6	14.0 ± 2.8	8	< 0.001
CPA	81.8 ± 2.7	24.8 ± 3.1	8	< 0.0001

Data are means ±s.e.m.n, number of arteries. l-NAME at 100 μm was always present when the effect of other agents was tested and propionate relaxation in the presence of l-NAME alone was used as the ‘control’. Values are derived from combining the effects at 7, 8 and 9 min (corrected using the time control as shown for Fig. 1) after propionate was applied (except for indomethacin, where 6, 7 and 8 min were used).

In all but a few experiments (see below), we examined the mechanisms of the vasorelaxing response using 10 mm propionate, which typically caused ∼65 % relaxation. Figure 2A illustrates the relaxation to this concentration of propionate in sets of endothelium-intact and endothelium-denuded RMSAs. Removal of the endothelium inhibited the relaxation induced by 10 mm propionate by approximately 75 %. As depicted in Fig. 2B and Table 1, the propionate-induced relaxation was not significantly inhibited by the endothelial nitric oxide synthase (eNOS) blocker l-NAME (100 μm) (68 ± 4 vs. 66 ± 3 % in controls, n = 35). This concentration of l-NAME reduced relaxation to 1 μm acetylcholine, measured 1 min after its application, from 75 ± 6 % to 18 ± 8 % (P < 0.001, n = 15). The cyclo-oxygenase inhibitor indomethacin (1 μm) also did not affect the magnitude of relaxation (61 ± 7 vs. 60 ± 4 % in controls, n = 10). Although these observations indicated that NO and prostanoids were not involved in the propionate-induced relaxation, the further investigations into the effects of propionate described hereafter were routinely performed in the presence of 100 μm l-NAME, since this agent tended to retard the intrinsic decay of the contraction which typically occurred over the 20 min period of stimulation. Indomethacin, on the other hand, tended to suppress contractions and make them more unstable, and was not used further.

Figure 2. Endothelium and concentration dependence of propionate-induced relaxation.

A-C, mean (± s.e.m.) normalised and corrected contraction amplitude plotted at 1 min intervals, showing the effects of endothelium removal (A, n = 6), 100 μm l-NAME (B, n = 35) or 25 mm K⁺ (C, n = 5) in the bath solution on responses to 10 mm propionate. D, concentration-dependent relaxation induced by propionate. Points show the mean of the percentage relaxation measured at 12, 13 and 14 min for 0.1 (n = 6), 0.3 (n = 7), 1 (n = 13), 3 (n = 13), 10 (n = 41) and 30 mm propionate (n = 10). * Significant relaxation (P < 0.05).

In the presence of l-NAME, the relaxation to 10 mm propionate was virtually abolished when the K⁺ concentration in the bathing medium was raised from 4.7 to 25 mm (Table 1, Fig. 2C). The concentration dependency of the propionate relaxation of the noradrenaline-evoked contracture is shown in Fig. 2D. Relaxation was significant at 1 mm, and was almost complete at 30 mm.

The inhibition of propionate-induced relaxation achieved by raising extracellular [K⁺] to 25 mm suggested that propionate was acting via hyperpolarisation of the smooth muscle. It has previously been shown that relaxation of RMSAs to a variety of vasodilating agonists is, in large part, due to endothelium-derived hyperpolarising factor (EDHF) (Zygmunt et al. 1995; Edwards et al. 1998), and we therefore went on to determine whether this was also true for propionate. In this case, it would be predicted that relaxation to propionate should also be suppressed by agents that block several types of K⁺ channels (Edwards et al. 1998).

Figure 3 and Table 1 show that the response to propionate was substantially inhibited by charybdotoxin and apamin (each 100 nm), toxins that block medium conductance and small conductance calcium-activated K⁺ channels, respectively (van Renterghem et al. 1995; Marchenko & Sage, 1996). Significant inhibition of the propionate-induced relaxation also occurred following treatment with either Ba²⁺ (50 μm, Fig. 4A and B, and Table 1), or ouabain (100 μm, Fig. 4B and C, and Table 1), which block inwardly rectifying K⁺ currents (Edwards & Hirst, 1988) and the Na⁺,K⁺-ATPase, respectively.

Figure 3. Combined effect of charybdotoxin and apamin on propionate-induced relaxation.

A, mean (± s.e.m.) normalised contraction amplitude plotted at 1 min intervals, showing the effects of 100 nm charybdotoxin (ChTx) and 100 nm apamin (Ap) on responses to 10 mm propionate and paired time controls (n = 9). B, propionate-induced relaxations after correction for time controls, before and after treatment with ChTx and Ap.

Figure 4. Effect of Ba²⁺ or ouabain on propionate-induced relaxation.

A and C, mean (± s.e.m.) normalised contraction amplitude plotted at 1 min intervals, showing the effects of 50 μm Ba²⁺ (A, n = 12) or 100 μm ouabain (C, n = 6) on responses to 10 mm propionate and paired time controls. B and D, propionate-induced relaxations after correction for time controls, showing the effects of Ba²⁺ or ouabain, respectively.

A similar inhibition of the propionate response occurred when 50 μm Ba²⁺ was combined with either 1 μm or 100 μm ouabain, under which conditions both the inward rectifier and the Na⁺,K⁺-ATPase (Fig. 5 and Table 1) should be blocked. In contrast, the combination of charybdotoxin and apamin with Ba²⁺ produced a slightly, but significantly, greater inhibition of the propionate relaxation than did charybdotoxin/apamin alone (Fig. 6 and Table 1).

Figure 5. Combined effects of Ba²⁺ and ouabain on propionate-induced relaxation.

A and C, mean (± s.e.m.) normalised contraction amplitude plotted at 1 min intervals, showing the effects of 50 μm Ba²⁺ and 1 μm ouabain (A, n = 8) or 50 μm Ba²⁺ and 100 μm ouabain (C, n = 10) on responses to 10 mm propionate and paired time controls. B and D, propionate-induced relaxations after correction for time controls, showing the effects of Ba²⁺ together with 1 or 100 μm ouabain, respectively.

Figure 6. Effect of combining Ba²⁺ with charybdotoxin and apamin on propionate-induced relaxation.

A, mean (± s.e.m.) normalised contraction amplitude plotted at 1 min intervals, showing the effects of combining 50 μm Ba²⁺ with 100 nm charybdotoxin (ChTx) and 100 nm apamin (Ap) on responses to 10 mm propionate and paired time controls (n = 10). B, propionate-induced relaxations after correction for time controls, before and after treatment with Ba²⁺ plus ChTx and Ap.

The effect of iberiotoxin, which blocks large conductance but not medium conductance Ca²⁺-activated K⁺ channels, is shown in Fig. 7 and Table 1. Unlike the other drugs tested, iberiotoxin had a pronounced effect upon the NA contraction itself (Fig. 7A), reversing its typical decay. It did not, however, reduce the amplitude of the propionate-mediated relaxation.

Figure 7. Effect of iberiotoxin on propionate-induced relaxation.

A, mean (± s.e.m.) normalised contraction amplitude plotted at 1 min intervals, showing the effects of 100 nm iberiotoxin (IbTx) on responses to 10 mm propionate and paired time controls (n = 13). B, propionate-induced relaxations after correction for time controls, before and after treatment with IbTx.

We did not evaluate the role of delayed rectifier K⁺ (K_V) channels in the response to propionate because 4-aminopyridine, which is widely used as a non-selective K_V blocker, almost abolished the NA contraction even at a fairly low concentration (1 mm).

Previous evidence indicates that the release of EDHF by acetylcholine can be attenuated by interventions which empty intracellular Ca²⁺ stores in the endothelial cells (Fukao et al. 1997). There are also suggestions that acidification of endothelial cells can cause the release of Ca²⁺ stores (Ziegelstein et al. 1993; Ladilov et al. 2000). We therefore examined whether thapsigargin, which should cause depletion of these stores in both the endothelium and the smooth muscle, had any effect on the propionate-mediated relaxation. Figure 8 illustrates one such experiment. Acetylcholine applied during the first contraction (a) caused a large relaxation. This was then greatly inhibited following pre-treatment with l-NAME (b). The response to propionate in the continuing presence of l-NAME was then examined (d) following a time control (c). Thapsigargin was then applied, and caused a complete inhibition of the residual acetylcholine-mediated relaxation (e). After another time control in the presence of both l-NAME and thapsigargin (f), the response to propionate was re-assessed (g), and found to be diminished. Figure 9A shows the mean results from eight similar experiments, in which the response to propionate was reduced from 52 ± 8 to 14 ± 3 %, P < 0.001, n = 8, see also Table 1). Thapsigargin itself caused a small but consistently observed inhibition of contraction (compare Fig. 8 c and f). Very similar results were obtained when the same protocol was followed using 10 μm cyclopiazonic acid (CPA) rather than thapsigargin (Fig. 9C and D).

Figure 8. Example of effect of thapsigargin on propionate-induced relaxation.

Trace showing acetylcholine (Ach, a, b, e), time control (c, f) and propionate responses (d, g) in the absence and presence of 100 μm l-NAME, and the combination of l-NAME with 100 nm thapsigargin. The arrows show the points at which solutions were exchanged (and addition of 10 mm propionate). The dotted lines represent basal tone.

Figure 9. Effect of thapsigargin and cyclopiazonic acid.

Mean (± s.e.m.) normalised contraction amplitude plotted at 1 min intervals, showing the effects of 100 nm thapsigargin (thaps, A) or cyclopiazonic acid (CPA, C) on responses to 10 mm propionate and paired time controls (n = 8), and propionate-induced relaxations after correction for time controls, before and after treatment with thapsigargin (B) or CPA (D).

We previously showed that relaxation to both lactate and butyrate at a high concentration (50 mm) was endothelium independent (McKinnon et al. 1996; Aaronson et al. 1996). When, however, we repeated these experiments using 10 mm butyrate in order to allow a comparison with propionate, it emerged that relaxation to this lower concentration of butyrate was significantly inhibited by removal of the endothelium (control relaxation: 71 ± 3 % vs. relaxation in absence of endothelium, 43 ± 3 %; n = 12; P < 0.005). Moreover, 50 mm propionate was able to cause a very marked relaxation in the absence of the endothelium (82 ± 7 %, n = 10), as was 50 mm butyrate (76 ± 5 %, n = 12).

DISCUSSION

The results show that propionate relaxes RMSAs by both endothelium-dependent and -independent mechanisms. At a concentration of 10 mm (and presumably below), the effect of propionate is almost entirely endothelium dependent, while at a very high concentration (50 mm), propionate is able to cause relaxation in the absence of the endothelium. We did not investigate the nature of the endothelium-independent vasorelaxation by propionate in this study, and the discussion below focuses on the mechanism of the endothelium-dependent effect, since this has not previously been observed for any SCFAs.

The data suggest strongly that at a concentration of 10 mm propionate acts on the RMSAs to cause a relaxation which resembles the NO/prostanoid-independent component of the endothelium-dependent vasodilatation elicited by agonists such as acetylcholine, bradykinin and substance P (Garland et al. 1995). This relaxation has been thought of as resulting from the release of an endothelium-derived hyperpolarising factor (EDHF), although the involvement of a factor per se is contentious. In any case, the EDHF response can be functionally defined by its sensitivity to moderate elevation of [K⁺]_o and to K⁺ channel blockers but not to inhibitors of eNOS and cyclo-oxygenase (Campbell & Harder, 2001). The EDHF response is generally more sensitive to combined blockade of small and intermediate conductance K⁺ channels with apamin and charybdotoxin than to inhibition of large conductance K⁺ channels with iberiotoxin (Zygmunt & Hogestatt, 1996), and is also typically attenuated by low concentrations of Ba²⁺, which are thought to act by blocking the inward rectifier K⁺ current, and by inhibition of the Na⁺ pump with ouabain (Edwards et al. 1998).

There are at present three main proposals as to the mechanism by which to endothelium-dependent (NO and prostanoid independent) hyperpolarisation and relaxation (hereafter termed the EDHF response for convenience) occurs. According to the first of these, hyperpolarisation is due to the release of epoxyeicosatrienoic acids (EETs). EETs are cytochrome P-450 metabolites of arachidonic acid, which are produced by the endothelium, and have been shown to hyperpolarise smooth muscle by activating Ca²⁺-activated K⁺ channels (Campbell et al. 1996). The second model suggests that vasodilating agonists open sufficient Ca²⁺-activated K⁺ channels on the endothelium to raise [K⁺]_o in the subendothelial space. This then stimulates inwardly rectifying K⁺ channels and also the Na⁺,K⁺-ATPase on the neighbouring smooth muscle cells, causing hyperpolarisation (Edwards et al. 1998). The third model also envisions an initiating role for the opening of endothelial Ca²⁺-activated K⁺ channels. This results in hyperpolarisation of the endothelium, which is transmitted to the smooth muscle via myoendothelial gap junctions (Chaytor et al. 1998).

The contribution of these various mechanisms to the EDHF response is currently the subject of intense controversy. The relative importance of each pathway seems to vary between different arteries, such that each preparation needs to be considered individually. In RMSAs, attention has recently focused on the role of K⁺ and myoendothelial gap junctions in RMSAs, as EDHF is unlikely to be a cytochrome P-450 metabolite (Okazaki et al. 1998; Tanaka et al. 1999). Edwards et al. (1998) demonstrated that the EDHF-mediated relaxation to acetylcholine in RMSAs was partially antagonised by Ba²⁺ plus ouabain. An elevation of [K⁺]_o also caused relaxation, and this was attenuated by removal of the endothelium. Based on these data, the authors suggested a role both for myoendothelial gap junctions and a direct hyperpolarising effect of high K⁺ on the smooth muscle in this response. Subsequent studies have favoured the importance of the former hypothesis, since the gap junction inhibitors Gap-27 (Dora et al. 1999; Doughty et al. 2000) and 18-α-glycyrrhetinic acid (Doughty et al. 2000) greatly suppressed the EDHF response to acetylcholine in RMSAs. Moreover, work by Lacey et al. (2000) and Doughty et al. (2000) demonstrated that the EDHF response in RMSAs could not be mimicked simply by raising [K⁺]_o from its normal concentration of ∼5 mm, in part because this manoeuvre only caused relaxation in a minority of experiments. Recently however, Dora & Garland (2001) showed that elevated [K⁺]_o can consistently evoke vasodilatation as long as preconstriction was submaximal. They also found that the blocking effect of Ba²⁺ on the relaxation to high K⁺ was abolished by removal of the endothelium, whereas the effect of ouabain was greatly enhanced by this manoeuvre. These results suggest that elevation of K⁺ causes RMSA hyperpolarisation and relaxation through multiple mechanisms, including stimulation of the smooth muscle cell Na⁺ pump, and activation of endothelial cell K⁺ channels (inwardly rectifying as well as Ca²⁺ activated), leading to a hyperpolarisation which is conducted to the smooth muscle via gap junctions.

As shown in the present report, propionate caused a relaxation of the NA contracture which was substantially blocked by ouabain and/or Ba²⁺, and by the combination of apamin and charybdotoxin but not iberiotoxin. l-NAME and indomethacin, however, had no effect on relaxation to propionate. It would therefore seem that propionate is causing relaxation primarily via an EDHF response.

Dora & Garland (2001) found that in RMSAs ouabain is effective in preventing the vasodilatory effect of elevated K⁺ in the absence of an endothelium, implying it must be acting through the smooth muscle Na⁺ pump. It is notable that in our experiments ouabain was as effective at preventing relaxation as was the combination of apamin and charybdotoxin, which are thought to work on endothelial rather than smooth muscle K⁺ channels in this artery (Edwards et al. 1998). This suggests that the bulk of the propionate-mediated relaxation can be explained by a stimulation of the smooth muscle Na⁺ pump resulting from the efflux of K⁺ from the endothelium and an increase in the extracellular K⁺. In this case, there is no need to posit an involvement of myoendothelial gap junctions in the effect of propionate, which in any case is difficult to envision considering the theoretical unlikelihood that the monolayer of endothelial cells could provide enough hyperpolarising current though myoendothelial gap junctions to drive the much larger smooth muscle mass to a more negative potential (Campbell & Harder, 2001). Perhaps the role of these junctions is to supply K⁺ to the endothelium from the smooth muscle, thus allowing a larger rise in [K⁺]_o.

Further evidence that EDHF is involved in the response to propionate, as well as a clue as to how propionate might be acting, emerges from the finding that both thapsigargin and CPA, which deplete intracellular Ca²⁺ stores in both smooth muscle and endothelial cells, greatly inhibited this effect of propionate. Fukao et al. (1997) have previously shown that both drugs abolished the hyperpolarising response to acetylcholine in RMSAs. They concluded that EDHF release occurs in response to the release of endothelial Ca²⁺ stores, and also a subsequent activation of capacitative Ca²⁺ entry. This is of course consistent with a role for endothelial Ca²⁺-activated K⁺ channels in the EDHF response, as is the observation that thapsigargin itself caused hyperpolarisation (Fukao et al. 1997) and, in our hands, a slight inhibition of contraction (Fig. 8 and Fig. 9).

The inhibition by both thapsigargin and CPA of the propionate-mediated relaxation suggests that propionate could be acting by releasing endothelial Ca²⁺ stores. Although Hsu et al. (1995) found that [Ca²⁺]_i was not affected by intracellular acidification in cultured piglet cerebral microvascular endothelial cells, Ziegelstein et al. (1993) reported that intracellular acidification of cultured rat aortic endothelial cells caused a release of intracellular [Ca²⁺]. Although this group found that the response was not prevented by thapsigargin or bradykinin, Ladilov et al. (2000) have more recently demonstrated that acidification caused a large and thapsigargin-inhibitable augmentation of a rise in intracellular [Ca²⁺] in cultured rat coronary endothelial cells exposed to glucose-free medium. The mechanism by which a fall in cellular pH could lead to Ca²⁺ release is unclear; however, Wolosker et al. (1997) have reported that acidification from pH 7.0 to 6.5 caused a doubling of the rate of passive Ca²⁺ efflux from sarco/endoplasmic vesicles prepared from cardiac muscle, cerebellum, and platelets.

Propionate has been shown to cause a fall in pH_i in both piglet cerebral microvascular endothelial cells (Hsu et al. 1995), and in ECV304 cells, which are derived from human umbilical vein endothelial cells (Wakabayashi & Groschner, 1996). We speculate that a similar propionate-induced fall in pH_i in the RMSAs endothelium triggers the EDHF response via a release of endothelial Ca²⁺ stores. The prominent decay of the vasodilating response to propionate which is evident in our results might be related either to the exhaustion of these Ca²⁺ stores, or to an apparent transience of the propionate-mediated reduction in endothelial cell pH_i (Hsu et al. 1995).

The possibility that a reduction in endothelial cell pH_i is responsible for the release of EDHF is consistent with studies by Harder (1982) and Peng et al. (1998) demonstrating that hypercapnic acidosis caused a hyperpolarisation of rat cerebral arteries which was accompanied by an apparent increase in the membrane K⁺ conductance (Harder, 1982). Moreover, Okazaki et al. (1998) found that imposition of hypercapnia in the presence of elevated NaHCO₃, which caused a transient intracellular acidification, also produced a marked relaxation of rat mesentery which was blocked by raising [K⁺]_o to 12.5 mm and also by 0.3 mm Ba²⁺, suggesting that it was also due to the EDHF response. Hypercapnia, however, has also been shown to cause endothelium-dependent relaxation by both NO- and prostanoid-dependent mechanisms (e.g. Carr et al. 1993; Hsu et al. 1995), and also exerts direct vasodilating effects on vascular smooth muscle (e.g. Peng et al. 1998).

The vasodilating effect of 10 mm butyrate was only partially endothelium dependent, and a small relaxation to propionate also persisted after removal of the endothelium. Moreover, a very marked endothelium-independent relaxation was elicited by both propionate and butyrate at a high concentration (50 mm). Clearly, therefore, SCFAs have direct vasodilating effects on RMSA vascular smooth muscle, particularly at high concentrations. A variety of mechanisms have been proposed to contribute to this direct effect (see reviews by Aalkjaer & Poston, 1996; Smith et al. 1998, Austin & Wray, 2000), the nature of which we did not pursue in the present study.

Although vasodilatation by propionate itself may be of direct physiological relevance only for the arterioles of the intestinal wall, it is should be borne in mind that shear stress has been shown to cause both intracellular acidification of cultured rat endothelial cells (Ziegelstein et al. 1992), and a vasodilatation of RMSAs which is also inhibited by pre-treatment with charybdotoxin and apamin (Takamura et al. 1999). It is possible, therefore, that the vasodilating effect of propionate may represent only one example of a more general vascular mechanism by which acidification of the endothelium could activate the EDHF response.