Extracellular ATP can activate autonomic signal transduction pathways in cultured equine sweat gland epithelial cells

The secretion of sweat is an important thermoregulatory response in Equidae but the means by which sweating in these animals is controlled is not fully understood. Current evidence suggests that horse sweat glands are controlled via β₂-adrenoceptors and that the changes in membrane permeability underlying secretion are therefore mediated primarily by cyclic-AMP-dependent mechanisms (Snow, 1977; Johnson and Creed, 1982; Bijman and Quinton, 1984a,b). Preliminary experiments have shown, however, that external ATP can both increase intracellular free calcium concentration ([Ca²⁺]_i) and modulate membrane permeability in cultured equine sweat gland cells (Wilson et al. 1993a). This raises the possibility of a purinergic component to the sudomotor control mechanism. In the present experiments, we have therefore characterised the effects of external ATP upon equine sweat gland epithelial cells to establish the extent to which this agonist can exert regulatory control over autonomic signal-transduction pathways.

Standard techniques were used to maintain an epithelial cell line (E/92/3) derived from the equine sweat gland (Wilson et al. 1993b). The standard culture medium was serum-free William’s medium E, supplemented as detailed below. Cells were removed from culture flasks using trypsin (0.24 %)/EDTA (0.05 mmol l^-1) and resuspended (approximately 10⁵ cells ml^-1) in medium containing 5 % (v/v) foetal calf serum. Samples of this suspension were plated either on glass coverslips (approximately 0.2 ml each) or in 24-well plates (0.5 ml well^-1). Once cellular attachment had occurred (after approximately 12 h), the plating medium was replaced with standard medium.

Coverslips bearing growing cells (18–36 h after plating) were incubated (30–40 min, 37 °C) in culture medium containing the acetoxymethyl (AM) ester form of the Ca²⁺-sensitive, fluorescent dye Fura-2 (3 μmol l^-1) in order to load the cells with this dye. The coverslips were then mounted in a small (approximately 100 μl) heated chamber attached to the stage of an inverted microscope, where the cells were superfused (approximately 5 ml min^-1, 37 °C) with physiological salt solution. An increase in [Ca²⁺]_i causes a rise in the Fura-2 fluorescence ratio (350 nm and 380 nm excitation wavelengths) recorded from cells loaded with this dye and this allows changes in [Ca²⁺]_i to be monitored using standard microspectrofluorimetric techniques (Grynkiewicz et al. 1985). Absolute calibration of such signals is, however, very difficult to achieve for cells loaded with the AM form of this dye (Highsmith et al. 1986), so experimental records are presented as fluorescence ratios obtained from groups of 2–5 cells. Increases in [Ca²⁺]_i were quantified by measuring the fluorescence ratio at the peak of the response and subtracting from it the ratio measured before application of the agonist.

Cells grown (3–4 days) on 24-well plates were washed twice with physiological salt solution containing papaverine (20 μmol l^-1) and incubated (10 min, 37 °C) in this buffer (0.5 ml well^-1) either under control conditions or under the experimental conditions specified in the Results section. Papaverine is a phosphodiesterase inhibitor that was included in the incubation medium to inhibit the hydrolysis of cyclic AMP. Concentrated perchloric acid (10 μl) was then added to each well to arrest cellular metabolism and to extract cyclic AMP; these extracts were neutralised with KOH, brought to 1 ml with assay buffer (50 mmol l^-1 sodium acetate/acetic acid buffer; pH 5.8) and stored at -18 °C until their cyclic AMP contents were determined. Cellular protein was also extracted from each well with 1 ml of 300 mmol l^-1 NaOH in 1 % (w/v) sodium dodecyl sulphate and quantified (Lowry et al. 1951; Petersen, 1977) so that the experimental data could be expressed as pmoles of cyclic AMP per microgram of cellular protein (pmol μg^-1).

Cyclic AMP was determined by radioimmunoassay. This is a technique that is based upon competition between the variable amount of cyclic AMP present in each cell extract and a fixed amount of a radioactive cyclic AMP analogue (adenosine 3′,5′-cyclic phosphoric acid 2′-O-succinyl-3-[¹²⁵I]iodotyrosine methyl ester; ¹²⁵I-cyclic AMP, Amersham). A finite number of cyclic AMP binding sites was provided by a rabbit antibody against cyclic AMP, which had affinity only for acetylated cyclic AMP. Samples of the acid extracts and of cyclic AMP standards were therefore brought to 500 μl with assay buffer and excess (25 μl) acetylating reagent (1 volume of acetic anhydride:2 volumes of triethylamine) was added to each tube. Samples (100 μl) of the resulting solutions were transferred to assay tubes and 100 μl aliquots of primary antibody (Calbiochem rabbit anti-cyclic AMP diluted 1:10 000 in 50 mmol l^-1 sodium acetate/acetic acid buffer; pH 4.75, containing 1 mg ml^-1 human gamma globulin) and of ¹²⁵I-cyclic AMP (approximately 12.5 nCi) were added to each. The binding reaction was then allowed to reach equilibrium (18 h, 4 °C). A sample (500 μl) of a donkey anti-rabbit IgG antibody immobilised onto beads (Amersham Amerlex-M immunoassay separation reagent) was then added to each tube to bind the primary antibody together with any associated ¹²⁵I-cyclic AMP. After incubation (10 min, at approximately 20 °C), these beads were precipitated by centrifugation (10 min, 1500 g) and the amount of ¹²⁵I in each pellet was determined by gamma counting. The cyclic AMP content of each cellular extract was then calculated by reference to a standard curve. In some experiments, cells were stimulated with ATP at concentrations of up to 1 mmol l^-1. This exogenous ATP was thus present in the acid extracts, but appropriate experiments established that this did not interfere with the assay of cyclic AMP.

The culture medium was William’s medium E supplemented with L-glutamine (1 mmol l^-1), penicillin (100 i.u. ml^-1), streptomycin (100 μg ml^-1), bovine insulin (100 μg ml^-1), hydrocortisone (10 ng ml^-1), transferrin (10 μg ml^-1) and sodium selenite (10 ng ml^-1). The physiological salt solution contained (in mmol l^-1): NaCl, 130; KCl, 5; Hepes, 20; MgCl₂, 1; CaCl₂, 1; and D-glucose, 10. The pH was adjusted to 7.4 with NaOH. Nominally Ca²⁺-free solutions were prepared by omitting CaCl₂ from the control saline.

ATP (100 μmol l^-1) consistently increased [Ca²⁺]_i but there was some variability in the form of this response (Fig. 1). The initial rise in [Ca²⁺]_i was rapid and, in some instances, this developed into a clearly defined peak (Fig. 1A,C). The interval between switching to the ATP-containing solution and obtaining a half-maximal response was 16.4±1.3 s (N=26, range 7.1–37.4 s) and so, as the dead time in the superfusion system was 1–2 s, the onset of the response was associated with a latency of at least 4–5 s and most cells responded only after longer intervals. Withdrawal of ATP caused [Ca²⁺]_i to fall to basal levels, although there was some delay before this became apparent (Fig. 1). The decline was usually rapid but, in some instances, [Ca²⁺]_i fell very slowly (Fig. 1C). Small, spontaneous increases in [Ca²⁺]_i were occasionally seen (Fig. 1B).

The ATP-evoked increase in [Ca²⁺]_i was concentration-dependent (Fig. 2) and a half-maximal response (EC₅₀) was estimated to occur at 12±1 μmol l^-1. The pyrimidine nucleotide UTP was also an effective Ca²⁺-mobilising agonist and this compound appeared to be slightly more potent (EC₅₀=4.9±0.2 μmol l^-1) than ATP. The cells also responded to ADP, but higher concentrations of this compound were required to increase [Ca²⁺]_i (Fig. 2), indicating that it was less potent than either UTP or ATP (EC₅₀=346±60 μmol l^-1). AMP (1 and 3 mmol l^-1, N=4), adenosine (1 and 3 mmol l^-1, N=6) and α? β-methylene-ATP (30 μmol l^-1, N=6) caused no discernible change in [Ca²⁺]_i, but all cells in which these compounds were tested subsequently responded to 100 μmol l^-1 ATP.

Internal [Ca²⁺] remained above basal levels throughout 2 min applications of either ATP or UTP, although it tended to decline in the presence of agonist (Fig. 3). These compounds subsequently caused a rapid rise in [Ca²⁺]_i during superfusion with the nominally Ca²⁺-free solution but, in this situation, the response could not be sustained (Fig. 3). Subsequent elevation of external [Ca²⁺], in the continued presence of agonist, evoked a sustained and reversible rise in [Ca²⁺]_i (Fig. 3). Such an elevation of external [Ca²⁺] did not affect [Ca²⁺]_i in unstimulated cells (N=3), so the responses to ATP and to UTP consisted of an initial, transient phase independent of external Ca²⁺ and a sustained, Ca²⁺-dependent element.

Repeated applications of ATP evoked a series of increases in [Ca²⁺]_i. The peak change in the Fura-2 fluorescence ratio caused by the initial pulse was 1.50±0.10 (N=26) but the responses to subsequent stimuli were smaller, demonstrating that some loss of sensitivity developed during repeated stimulation (Fig. 4A, Table 1). In other cell types this has been attributed to receptor-mediated activation of protein kinase C (Llano and Marty, 1987; Maruyama, 1989) and so, to establish whether this was also true of the equine cells, we explored the effects of staurosporine and of the phorbol ester 12-O-tetradecanoyl phorbol 13-acetate (TPA). These are compounds that, respectively, inhibit and activate protein kinase C.

The data presented in Fig. 4A suggested strongly that the desensitisation was blocked by staurosporine (0.1 μmol l^-1) and potentiated by TPA (1.6 nmol l^-1). The desensitisation that developed under the various experimental conditions was therefore quantified (see Materials and methods) and these data are presented in Table 1; the effects of staurosporine and TPA were statistically significant. Also shown in Table 1 are data from experiments in which higher concentrations of TPA were used; there is some evidence that the effect of this compound is concentration-dependent but it did not totally abolished the response to ATP (Table 1).

The ability of staurosporine to antagonise the action of TPA was also explored. Cells were repeatedly stimulated with ATP and exposed to TPA (1.6 nmol l^-1) from immediately after the first pulse of agonist. This caused a loss of sensitivity essentially identical to that described above, but sensitivity to ATP appeared to be restored when the cells were subsequently exposed to 0.1 μmol l^-1 staurosporine (Fig. 4B). The mean fractional response measured immediately prior to addition of staurosporine (0.49±0.08) was less (P<0.002, paired t-test) than the mean response to the final two pulses of ATP (0.79±0.05); staurosporine can thus antagonise the action of ATP.

High concentrations of extracellular ATP were able to increase the cyclic AMP content of the equine cells. Although similar concentrations of ADP, AMP or adenosine were also able to evoke this response, these compounds were less potent than ATP and the pyrimidine nucleotide UTP was ineffective (Fig. 5).

We also explored the possibility that the cells may contain type 1 adenosine (A₁) receptors that characteristically allow adenosine to exert inhibitory control over adenylate cyclase (Burnstock, 1990b). In these experiments (N=6), the cyclic AMP content of the unstimulated cells was 0.250±0.064 pmol μg^-1 and this was increased by 0.489±0.118 pmol μg^-1 (N=6) on exposure to 0.1 μmol l^-1 forskolin, a diterpene that evokes cyclic AMP production by a direct effect on the catalytic subunit of adenylate cyclase. The cells responded to forskolin during exposure to the relatively selective A₁ receptor agonist N⁶-cyclopentyladenosine (1 nmol l^-1), but the increase in cyclic AMP content (0.304±0.092 pmol μg^-1) was smaller than the control (P<0.05, paired t-test).

The present study showed that ATP and certain related compounds can increase [Ca²⁺]_i in equine sweat gland epithelial cells with an apparent rank order of potency of UTP⩾ATP>ADP⪢AMP=adenosine=α?β-methylene-ATP. These results indicate that the response is mediated by a type 2 purine (P₂) receptor (Burnstock, 1990b) with properties essentially identical to those of a receptor subclass first identified in Ehlich ascites tumour cells (Dubyak and Young, 1985). These receptors are generally said to belong to the P_2U subclass and are usually coupled to phosphoinositidase C. They provide a means by which ATP can regulate the hydrolysis of a membrane phospholipid, phosphatidylinositol-4,5-bisphosphate (PtdIns4,5P₂) (Dubyak et al. 1988). The hydrophilic product of this reaction, inositol 1,4,5-trisphosphate (Ins1,4,5P₃), is the intracellular second messenger that releases Ca²⁺ from internal stores, and the resultant rise in [Ca²⁺]_i, which is characteristically sustained by Ca²⁺ influx, provides one of the central mechanisms by which hormones and neurotransmitters exert regulatory control (Berridge, 1993).

The present responses to ATP and to UTP could be attributed to the release of Ca²⁺ from a limited internal store and to receptor-regulated calcium influx. This result suggests very strongly that the P_2U receptors in these cells are functionally coupled via the production of Ins1,4,5P₃. Preliminary experiments have suggested that the resultant rise in [Ca²⁺]_i can cause changes in membrane permeability that are characteristic of a secretory response (Wilson et al. 1993a). These findings are in excellent accord with data from recent studies of epithelial cells from the amphibian kidney (Middleton et al. 1993) and the human trachea (Mason et al. 1991) and mammary gland (Flezar and Heisler, 1993). It appears that P_2U receptors provide a widespread mechanism that allows ATP to regulate transport processes in vertebrate epithelia.

ATP acts via P₂ receptors to increase [Ca²⁺]_i in rodent salivary and lacrimal acinar cells. These responses do not, however, involve either PtdIns4,5P₂ hydrolysis or the mobilisation of internal Ca²⁺ because the P₂ receptors in these cells are coupled to Ca²⁺-permeable ion channels (McMillian et al. 1988; Sasaki and Gallacher, 1990). The possibility that this mechanism contributes to the sustained Ca²⁺-dependent component of the present response cannot be excluded.

Some cell types contain adenosine receptors that are coupled to phosphoinositidase C (e.g. Ali et al. 1990). This does not, however, appear to be the case in the equine cells as adenosine evoked no discernible change in [Ca²⁺]_i.

The hydrophobic product of PtdIns4,5P₂ hydrolysis is a diacylglycerol molecule that remains in the plasma membrane and allosterically modulates the activity of protein kinase C: a Ca²⁺- and phospholipid-dependent enzyme. Protein kinase C can influence many aspects of cellular physiology by regulating the phosphorylation of specific proteins. Although the activation of the enzyme appears to be part of the normal response to agonists that evoke PtdIns4,5P₂ hydrolysis, the role of this enzyme in stimulus–secretion coupling is not altogether clear (Nishizuka, 1984). Experiments on cholinergically stimulated pancreatic and lacrimal acinar cells have suggested that the activation of protein kinase C may allow inhibitory control over receptor-regulated Ins1,4,5P₃ production and thus provide a mechanism for desensitisation (Llano and Marty, 1987; Maruyama, 1989). The present data were consistent with this view, as desensitisation to ATP did not take place if protein kinase C was inhibited and the desensitisation was augmented by an exogenous activator of the enzyme. A protein-kinase-C-dependent inhibitory pathway thus appears to become active during repeated stimulation with ATP and this supports the view that the P_2U receptors in these cells are coupled to phosphoinositidase C.

ATP and related compounds increased cellular cyclic AMP content with a rank order of potency of ATP>ADP=AMP=adenosine⪢UTP. The response thus appears to be mediated by a subclass of P₂ receptor (Burnstock, 1990b) with pharmacological properties different from those of the P_2U receptors that underlie the rise in [Ca²⁺]_i. To our knowledge, there are no previous reports that P₂ receptors can activate adenylate cyclase in epithelial cells (see, for example, Burnstock, 1990b). It is certainly clear that the P₂ receptors in epithelial cells from the amphibian kidney (Middleton et al. 1993) and the human mammary gland (Flezar and Heisler, 1993) are not coupled to adenylate cyclase in this way.

These responses to ATP can be attributed to the activation of type 2 adenosine (A₂) receptors. These receptors characteristically activate adenylate cyclase but have only a very low affinity for ATP. In many experimental situations, however, extracellular nucleotidases can dephosphorylate ATP and thus provide sufficient adenosine to evoke cyclic AMP formation (Burnstock, 1990b). The possibility that a component of the present response to ATP was due to such an effect cannot be excluded, but this does not seem likely as ATP was more potent than adenosine. There is a recent report suggesting that some of the effects of ATP upon human platelets may be due to P₂-receptor-regulated cyclic AMP production (Soslau et al. 1993). It is therefore possible that such receptors may be involved in the present responses to ATP.

Experimental manoeuvres that increase [Ca²⁺]_i can augment receptor-regulated cyclic AMP formation in isolated primate sweat glands (Sato and Sato, 1983). It is therefore possible that the present response to ATP reflects a small, A₂-receptor-mediated response that was potentiated by the simultaneous, P_2U-receptor-mediated rise in [Ca²⁺]_i. The response to adenosine would not be amplified in this way as this compound does not increase [Ca²⁺]_i. It has not, however, been unequivocally established that the potentiating interactions between [Ca²⁺]_i and cellular cyclic AMP content that occur in primate sweat glands are due to intracellular actions (Sato and Sato, 1983); it is possible that they may be secondary to stimulation of the nerve terminals that are present in isolated glands. Our earlier studies of equine sweat gland cells in culture showed that ionomycin inhibits cyclic AMP formation (Wilson et al. 1993b), and we have obtained an essentially identical result from a virally transformed, epithelial cell line derived from the human sweat gland (Wilson et al. 1994). Large increases in [Ca²⁺]_i thus appear to inhibit cyclic AMP formation, although this result does not preclude the possibility that smaller, receptor-mediated elevations of [Ca²⁺]_i may exert potentiating effects (Wilson et al. 1993b).

The present data also suggested that equine cells may contain A₁ receptors that were able to inhibit cyclic AMP synthesis. Such receptors are present in epithelial cell lines derived from the rabbit collecting tubule (Spielman et al. 1992) and from the bovine and the human ocular ciliary epithelium (Wax et al. 1993). Such receptors do not, however, appear to be present in epithelial cells of the human trachea (Lazarowski et al. 1992). It is therefore possible that adenosine exerts both excitatory and inhibitory control over adenylate cyclase in the equine sweat gland epithelium. The physiological significance of this is not clear, but the present responses to adenosine would have been due to the activation of these mutually antagonistic signalling pathways. The net effect of this compound upon cellular cyclic AMP content would, therefore, be hard to predict, especially as the A₁- and A₂-receptor-linked pathways appear to desensitise at different rates (Ramkumar et al. 1991).

The principal new finding of the present work is that purinergic receptors are present on equine sweat gland epithelial cells in culture. This raises the possibility that such receptors may also be present in the intact equine sweat gland, where they may be involved in the regulation of glandular activity. The present experiments were, however, undertaken using a permanent cell line that was derived from the secretory epithelium of the equine sweat gland by a spontaneous transformation process (Wilson et al. 1993a). The relevance of this work to the biology of the Equidae therefore depends crucially upon the extent to which these cells have retained the physiological features of the parent epithelium. We cannot unequivocally exclude the possibility that de-differentiation may have occurred in vitro and that the cells may now express physiological features that are not seen in vivo. Our initial studies of these cells did, however, suggest that they faithfully retained the adrenergic receptors and second-messenger pathways that regulate secretion from intact glands (Wilson et al. 1993a).

Although sweat glands are found in essentially all mammalian groups, it is only in primates, certain Bovidae and the Equidae, that they have become specialised as thermoregulatory organs (Jenkinson, 1973). Sweating in horses can be regulated both by adrenaline released from the adrenal medulla and by the action of postganglionic sympathetic nerves (Robertshaw and Taylor, 1969). Equine sweat glands contain β₂-adrenoceptors that characteristically have a high affinity for adrenaline. These receptors allow adrenaline to evoke the formation of cyclic AMP and this intracellular second messenger can evoke secretion in vitro; the mechanism of hormonally evoked sweating is thus reasonably clear (Robertshaw and Taylor, 1969; Snow, 1977; Bijman and Quinton, 1984a,b; Wilson et al. 1993b). The classic neurotransmitter released from the sympathetic nerve fibres is noradrenaline, and the postjunctional effects of this substance are mediated, primarily, via α- and β₁-adrenoceptors. The equine sweat gland epithelium appears to be essentially devoid of these receptors since large injections of noradrenaline do not evoke secretion in vivo and high concentrations of this compound are required to evoke either secretion or cyclic AMP formation in vitro. It is thus not obvious how the sympathetic nerves may regulate sweating (Robertshaw and Taylor, 1969; Snow, 1977; Bijman and Quinton, 1984a,b; Wilson et al. 1993b).

It is now well known that ATP, as well as noradrenaline, is contained within sympathetic neurosecretory vesicles and that these compounds are released from the nerve terminals in parallel (Burnstock, 1990a,b). The present data thus suggest that ATP released from the periglandular sympathetic nerves may have the capacity to activate autonomic signal-transduction pathways within the glandular epithelium. This work, together with our earlier experiments (Wilson et al. 1993a), raises the possibility that purinergic neurotransmission may play a central role in regulating secretory activity in the equine sweat gland and that this process may therefore be an important part of the thermoregulatory response. It is now important to establish whether functional P_2U recepors are present in intact equine sweat glands.

This work was supported by a grant from the Royal Hong Kong Jockey Club. The Authors are grateful to Godfrey Smith, Duncan MacGregor and Soma Rakhit for their helpful comments and discussions.