ATP release from non-excitable cells

The luciferin–luciferase technique

The standard technique to measure extracellular ATP in bulk solutions uses the bioluminescence of the ATP-dependent luciferase-mediated oxidation of luciferin [22]. This is a straightforward and robust method using commercially available luminometers to prove the presence of ATP in a given solution, with or without cells. Other relevant nucleotide signalling molecules like uridine triphosphate (UTP), uridine diphosphate (UDP) or adenosine diphosphate (ADP) are not detected. It is exquisitely sensitive, shows fast response times (milliseconds) and a broad linear detection range and with careful optimisation allows the detection of ATP concentrations down to the femtomolar range. There are, however, many factors that interfere with the luciferase activity. The enzymatic action has its optimum at room temperature (20°C), pH 7.8 and requires Mg²⁺ [23]. It is significantly inhibited by various anions (e.g. Br⁻, I⁻, NO⁻₃, etc.) [24] and high salt concentrations in general [25]. The anion transport inhibitor DIDS [26], the cation channel blocker gadolinium [27] or P2 receptor antagonist suramin [28] inhibits the luciferase reaction. All these characteristics have proven to be significant obstacles in the characterisation of the ATP release pathway(s). In cells triggered to release nucleotides, bulk extracellular ATP concentrations are commonly found to lie in the nanomolar range. Table 1 summarises some examples of extracellular ATP measurements in bulk solutions. Evidently, this method may not be very suitable to measure the ‘true’ ATP concentrations that a cell may encounter near the plasma membrane and does not allow the disclosure of kinetic information on ATP release because the released ATP is diluted into the relatively large surrounding fluid compartment. It needs to be appreciated that the extracellular space surrounding a cell in the native setting is very small. This contrasts with most commonly used experimental settings of cultured cells or freshly isolated tissues in which an enormous, non-physiological extracellular space is necessary to maintain homeostatic control of the experiment. Attempts to reduce the size of the extracellular fluid compartment have proven successful and show, e.g. in airway cell surface liquid layer that the hypotonicity-stimulated ATP concentration increase can be as high as 614 nM [29]. Despite this essential shortcoming, this method supplies a necessary fundamental observation because it proves the molecular identity of ATP.

Imaging of ATP release

The luciferin/luciferase technique has been used successfully to visualise ATP release [23, 25, 26, 35, 36]. In this approach, the cells (or tissue) are bathed with the ATP-specific luminescence indicator and the ATP is imaged directly as it is being released after a given stimulus. The first successful ATP imaging was conducted by Wang and co-workers in 2000, demonstrating non-destructive cell poking-induced ATP release from astrocytes and quantifying the ATP travelling wave velocity [23]. Beautifully, this study also succeeded in semi-simultaneous detection of cellular [Ca²⁺]_i with fluo-3 [23]. Also, poking of retina glia cells showed a luminescence ATP wave propagating at similar speed to that observed with [Ca²⁺]_i imaging [35]. The luciferase-generated light intensity is very low and requires highly sensitive imaging equipment (e.g. a nitrogen-cooled charge-coupled device (CCD) camera), together with long temporal integration, to achieve meaningful data. The achievable images are diffuse, with low spatial resolution. In the initial study, a camera integration time of 0.5 s was sufficient to detect ATP concentrations as low as 10 nM [23]. The second study also used temporal integrations of 0.5 s, together with the 2 × 2 binning function of a high-quality CCD camera, which was sufficient to monitor the kinetics of a 30-s period of touch-induced ATP release from glial cells [35]. A third study used cultured astrocytes, integration time of 10 s and a liquid nitrogen CCD camera to show spontaneous point source bursts of released ATP when the extracellular Ca²⁺ concentration was lowered below normal physiological values (0.5 mM) [36]. Undoubtedly, this novel technical extension carries the strength of a specific signal which reports extracellular ATP concentrations directly. The low temporo-spatial resolution of the luminescence imaging technique is a significant limiting factor and may preclude the possibility of ‘zooming’ closer into the mechanism of ATP release. Resting or spontaneous ATP release has not so far been imaged by the luciferin–luciferase method.

Two other studies using alternative ATP-dependent enzymatic reactions were also able to detect and image extracellular ATP. One exploited the disappearance of light absorption of consumed luciferin (as substrate in the ATP-dependent luciferin–luciferase reaction) to detect muscarinic receptor-stimulated release of ATP from pancreatic acinar cells [37]. The other study imaged ATP at the leading edge of a migrating neutrophil with the use of a two-enzyme assay system which catalyses the conversion of NADP⁺ to NADPH in the presence of ATP. The real-time generation of NADPH was measured as the appearance of NADPH fluorescence [13, 38].

Biosensor cells and ATP detection via an increase of cytosolic Ca²⁺

The use of a biosensor cell placed in the direct vicinity of an ATP-releasing cell was first introduced in 1989 by Cheek et al. who used NIH-3T3 fibroblasts co-cultured around single bovine adrenal chromaffin cells. After stimulation with nicotine, the chromaffin cells released ATP, which was sensed via a P2 receptor-dependent [Ca²⁺]_i increase by the neighbouring fibroblasts [39]. Extracellular ATP and other nucleotides commonly produce elevations of cytosolic Ca²⁺ via activation of either P2Y or P2X receptors [40]. Thus, the increase of [Ca²⁺]_i is used as a read-out to measure extracellular ATP. Also, the pioneering study demonstrating the ATP dependency of travelling [Ca²⁺]_i waves in rat basophilic leukaemia cells applied this biosensor technique to substantiate ATP as a paracrine factor [41]. Later, this approach was refined by Okada and colleagues and was applied to ATP-secreting B cells from the pancreas [42]. A detailed technical note describing the essentials and features was recently published [43]. A critical issue for the biosensor cell approach is the biosensor’s own susceptibility to stimuli. Thus, the experimental protocol must ascertain that an observed increase of [Ca²⁺]_i is caused by the extracellular nucleotide and not by other unidentified factors. An indicator cell can also be moved freely within the bath chamber using a micromanipulator and thereby positioned in the vicinity of the cell of interest to detect whether it releases ATP upon a given stimulus. This technique enables the researcher to answer in a convincing way the temporo-spatial characteristics of ATP release in different cells. Cultured mouse mesangial smooth muscle cells have been used as biosensors to detect ATP release from the basolateral side of rabbit renal macula densa cells [11, 17, 44]. ATP release was stimulated via an elevation of the luminal NaCl concentration without changing the environment of the biosensor cell. These findings were confirmed by PC12 cells [44]. Thus, little doubt is left that the release of ATP from the macula densa cells after exposure to high NaCl in the luminal perfusate is a key component of the tubulo-glomerular feedback mechanism [11]. Yet another example is the use of microglia cells as detectors for mechanical-induced astrocytic ATP release [45].

P2X₂ receptor-mediated ion currents as biosensors for extracellular ATP

A similar principle of ATP biosensing uses P2X receptor-expressing cells and patch clamp recordings to measure ATP-induced inward currents either in the whole-cell configuration or in outside-out excised patches [43, 46]. It was coined the ATP ‘patch sniffing’ method [47]. P2X₂ receptor-expressing neuro-endocrine (PC12) cells, derived from rat phaeochromocytoma, are often used in this approach. In the PC12 cell biosensor technique, a single non-attached cell is patched and therewith hooked up to a glass pipette. After breaking the seal, ATP-induced whole-cell currents can be recorded when holding the cell at −50 mV. This technique has been successfully applied to measure stimulated ATP release from a variety of different cultured cells, including various epithelial cells (C127 mammary epithelial cancer cells, 407 intestinal epithelial cells, airway epithelial cells, renal macula densa cells), primary cultured myocardial cells and pancreatic B cells [43] and Xenopus oocytes [48]. The outside-out patch approach adds a higher degree of specificity to the system because the membrane patch, in contrast to an entire cell, is less likely to respond to other external signals and is unable to release ATP itself.

The P2X₂ receptor tagged with enhanced green fluorescence protein can also be transfected heterologously into other cells of interest, which therewith become equipped with an ATP biosensor. This has been done in insulin-secreting INS1 cells and used to show that ATP is released via exocytosis in these cells. ATP release was detected as spontaneous spiking of suramin-sensitive inward currents. This signal is also described as ‘autaptic’ activation of the expressed P2X₂ receptor [17]. In a similar fashion, clusters of PC12 cells display such ‘autaptic’ behaviour, as detected by spontaneous transient inward currents (STICs) through P2X₂ receptors. There is firm evidence that this phenomenon reflects quantal exocytotic release of ATP from PC12 cells [2]. In addition, the discovery of pannexin 1 as an ATP-releasing channel in taste bud epithelia convincingly used the P2X2/3 biosensor approach [10]. Precise quantification of extracellular ATP concentration with the cell-based biosensor approaches can be difficult because the used ATP receptors often desensitise.

The amperometric ATP biosensor microelectrode

This novel method provides hope of increasing our knowledge of ATP as a secreted extracellular molecule. It is based on an ATP biosensor built by coating a platinum microelectrode with an ultra thin layer of various enzymes [49, 50]. Different enzymes and coating protocols have successfully been used to produce functional, robust and fast-acting ATP electrodes [49, 50]. The first ATP microelectrode used immobilised glucose oxidase and hexokinase and measured ATP concentrations between 10 and 200 nM [49]. A second study used glycerol kinase (GK) and glycerol-3-phosphatase (G3P) in the enzyme layer. In the presence of glycerol, ATP leads to GK-dependent formation of glycerol-3-phosphate, which together with O₂ is converted to H₂O₂ and glycerone phosphate by G3P. Subsequently, H₂O₂ is oxidised at the platinum electrode (+500 mV) which provides a current signal directly proportional to the amount of consumed ATP [50]. ATP-induced current signals are linear over the physiologically relevant ATP concentration range (200 nM to 50 µM) and the microelectrode appears to be very sensitive: ~250 mA M⁻¹ cm⁻². It works robustly in physiological extracellular solutions, has the dimensions of 50 µm in diameter and 2 mm in length, and can be positioned with a micromanipulator next to the tissue. It has been used to detect ATP release during locomotor activity of the spinal cord of Xenopus embryos. It does not detect other nucleotides but may perceive disturbances from other electro-active compounds like 5HT or ascorbate [50]. Using appropriate controls with the uncoated non-sensor electrode or without glycerol, it is possible to verify that the current signals are ATP-specific. Possibly, this technique can be optimised to record only from a few or even single cells. This ATP electrode was instrumental in demonstrating recently that ATP is an important central sensory transmitter in the medulla oblongata, stimulating breathing after elevation of peripheral pCO₂ [4]. Interestingly, in a very similar fashion, the same group has also produced an adenosine electrode [51]. This has the potential of recording the ATP breakdown product adenosine and detecting the temporo-spatial formation of ATP and adenosine in native tissues [51]. Adaptation of the ATP electrode to micromanipulator scanning devices or even an atomic force microscope cantilever is under technical development and may evolve into an electrode-based ATP imaging method applicable to physiological preparations [52, 53].

Other ATP biosensors

An interesting methodological approach used the atomic force microscope (AFM) to scan the surface of cystic fibrosis transmembrane conductance regulator (CFTR)-expressing respiratory epithelial cells [54]. In this technique, the luminal cellular surface was scanned with the AFM cantilever. The cantilever was ‘myosin-functionalised’, which caused this detecting device to move slightly during myosin-induced ATP hydrolysis. Why ATPase activity produces cantilever movements is not fully explained. The ATP-induced cantilever movements were taken to indicate ATP release, and results suggest several ATP release point sources on each cell. The ATP release could be blocked by glibenclamide and was not seen in non-CFTR-expressing epithelial control cells [54]. However, no follow-up studies with this technique have been published so far, probably reflecting the significant technical difficulty and expense of this method.

Another way to exploit an ATP-dependent enzymatic reaction to measure extracellular ATP was recently reported by Yegutkin and colleagues [55]. They made use of the fact that lymphocytes express surface adenylate kinase, which generates ATP/ADP from its precursor adenosine monophosphate (AMP) [55]. Using [³H]AMP and radio thin layer chromatography, they found that phosphotransfers to form ADP/ATP occurred spontaneously without added exogeneous ATP. This reaction must therefore have been driven by extracellular ATP provided by the lymphocytes themselves and was defined as the ‘intrinsic’ AMP phosphorylation property [56]. Quantification of this reaction can be used to calculate [³H]AMP phosphorylation capacities and close estimation of the ATP concentrations necessary for this reaction. They found ATP values around 2–5 µM, which must be present to drive this reaction, and suggested that a tonically released ATP ‘aura’ surrounds the lymphocytes. This method provides independent evidence that extracellular ATP concentrations can reach micromolar concentrations. However, the method is probably not applicable to all cells as it relies on the surface expression of extracellular ATP-regenerating enzymes like adenylate kinases. It is not suitable for kinetic evaluations of the ATP release process.

A nucleotide-conductive pore

ATP is a weak acid (pK_a 6.9) and therefore mainly present in anionic form in the cytosol [58]. The large electrochemical, outwardly directed ATP gradient has led researchers to suggest that cells may be equipped with an ATP-conductive pore [59, 60]. This could potentially allow ATP to permeate in a regulated fashion through a plasma membrane anion/chloride channel. Molecular evidence for functionally relevant plasma membrane anion channels includes the members of the ClC chloride channel family, CFTR and the different receptor-operated chloride channels (e.g. glycine receptor) [61]. The molecular nature of the cell swelling-induced anion conductances and the Ca²⁺-activated anion conductances is still unresolved [61]. Crystal structure and electrophysiological data from ClC channels clearly exclude permeation of ATP. The large organic anion ATP is much too big to pass through the ClC pore [62, 63]. Likewise, high-resolution structural data from water and K⁺ channels confirm that the narrow conductive pore is a general feature of ion channels [64, 65]. Thus, the available structural information on molecularly and structurally defined ion channels in general, and ClC channels in particular, do not support the proposition that the large bulky organic ion ATP can be transported through these pores.

Vesicular release of nucleotides

Vesicular release of ATP from neuronal or neuro-endocrine cells is an established phenomenon and many studies have determined ATP as a co-transmitter in peripheral and central neurons as well as from various neuro-endocrine cells. The interested reader is directed to several reviews of this area [15, 100, 101]. It is thus established that, e.g. chromaffin cells and the closely related PC12 cells store large amounts of ATP (>100 mM) together with catecholamines [2, 102], that insulin secretion from B cells occurs together with ATP [17] and that α-dense granules in thrombocytes contain very high concentrations of ADP, a known key factor in thrombus formation [14]. There is strong evidence that astrocytic release of nucleotides under physiological conditions involves an exocytotic mechanism [45, 103, 104], in contrast to the above-mentioned controversial issue of connexon hemichannel-mediated ATP release from astrocytes under non-physiological situations [36]. It is noteworthy that the inhibition of vesicular release with tools like brefeldin A or bafilomycin effectively inhibits travelling [Ca²⁺]_i wave propagation in astrocytes [104]. Both brefeldin A and bafilomycin have been shown to inhibit ATP release in numerous systems, supporting the notion that many non-excitable cells may be able to perform nucleotide release by this mechanism [28, 45, 82, 105]. Additional evidence supporting vesicular ATP release comes from experiments that interfere with the molecular players of constitutive and regulated exocytosis. Thus, knockdown of members of the soluble NSF attachment receptor family (SNAREs), or neurotoxins like tetanus toxin in astrocytes, inhibits nucleotide release [45, 103]. Of note is a recent observation of spontaneous autocrine quantal release of protons from HEK 293 cells, a phenomenon that may be suggested to coincide with ATP release from these vesicles [106]. Taken together, the above findings provide strong and unequivocal evidence that vesicular ATP release is a ubiquitous phenomenon of most, if not all, cells except erythrocytes [107]. The key challenge will be to unravel the physiological context in which the different organ systems ‘use’ vesicular ATP release to regulate a specific function. One example of this is the way in which tonic vesicular astrocytic purinergic signalling dampens neuronal synaptic networks in the CNS [103].

Spontaneous or constitutive nucleotide release

Very low concentrations (nanomolar range) of extracellular ATP surround most, if not all, cultured cells. As externally applied ATP is rapidly broken down by ecto-ATPases on the cell plasma membranes, the stable extracellular ATP levels must reflect a steady-state situation of tonic ATP release balanced by ATP degradation at the same rate [19, 32, 56, 74]. Tonic ATP release can be unmasked by non-specific inhibition of ecto-ATPases. Substances like ARL67156 and α,β-methylene-ATP, β,γ-methylene-ATP, ebselen or levamisole result in a slow and sustained increase of extracellular ATP concentrations [30, 32, 108–110]. Kinetic measurements of this increase allow calculation of actual cellular ATP release rates, which lie in the range of 10–500 fmol min⁻¹ per 10⁶ cells in different cell types [19]. Several findings suggest that spontaneous nucleotide release is of physiological significance. Scavenging of extracellular ATP with apyrase reduces the amount of tonically released arachidonic acid and lowers cAMP values in MDCK cells [111]. Slow spontaneous [Ca²⁺]_i oscillations in intact and cultured epithelial cells are greatly reduced in cells without relevant P2 receptors and are abrogated by ATP scavenging or pharmacological P2 receptor blockade [109]. Thus, constitutive nucleotide release appears to provide a tonic set-point regulation of a given organ function. In this context, the renal phenotype of the P2Y₂ receptor knockout mouse appears very interesting. These mice have hypertension, low plasma renin/aldosterone levels and renal hyperabsorption of Na⁺ [112]. It is established that P2Y₂ receptor stimulation inhibits ENaC-mediated Na⁺ and aquaporin2-mediated water transport in the renal collecting duct [113–115]. The absence of this receptor in the collecting duct should therefore result in an increased Na⁺ and H₂O reabsorption, leading to volume expansion and reduction of the ‘tonus’ of the renin–angiotensin–aldosterone system. Exactly, this was found in a recent publication [112]. This, however, would imply that the P2Y₂ receptors receive ‘tonic’ signalling input, a phenomenon, which should be preceded by tonic nucleotide release. This tonic nucleotide release can indeed be observed in cultured and intact renal epithelia [109, 111]. In summary, it is likely that control of organ function is exerted by tonic stimulation of P2 receptors. Most certainly, adenosine receptors are equally important in this scenario.

Stimulated nucleotide release

Nucleotide release can be stimulated to produce rapid and transient elevation of, e.g. ATP around the cells. Two main stimulation modes are described regarding this triggered ATP release: (a) mechanical perturbation of cells, and (b) through agonists, including nucleotides themselves. These two topics will be elaborated below.

Agonist-stimulated nucleotide release

An increasing number of studies has shown that membrane receptor stimulation triggers an increase of [Ca²⁺]_i and release of ATP. Table 2 lists various examples.

Most of these receptors are GPCRs which couple to an increase of [Ca²⁺]_i via the generation of IP₃. Direct stimulation of G proteins with GTP-γ-S or the G protein activator compound 48/80 strongly suggests that G protein activation is a pre-requisite for agonist-stimulated ATP release [26, 41]. In addition, most of these studies clearly indicated that the intracellular Ca²⁺ chelator BAPTA strongly or completely inhibited the agonist-induced ATP release [26, 28, 30, 82]. Receptor-independent [Ca²⁺]_i elevations with, e.g. ionomycin or A23187 have been shown to trigger ATP release, but these are apparently poor agonists [26, 29, 67]. Inhibition of signalling events in the IP₃ pathway, such as phospholipase C using U73122, can inhibit agonist-triggered ATP release [26, 28]. Taking these findings together, it is likely that agonist signals can trigger ATP release from most, if not all, cells. The released nucleotides could then, in a positive feedback mode, amplify a cellular response. These data indicate that agonist-induced ATP release involves a vesicular release process or activates a [Ca²⁺]_i-sensitive conductive nucleotide pore (e.g. pannexin 1). Of note is the finding that agonist- and swelling-induced ATP releases were found to be additive and may therefore involve different mechanisms [30]. Finally, it is worth mentioning that aldosterone could trigger the release of basolateral ATP from frog-derived renal epithelium (A6). This phenomenon was suggested by the authors to be important for the activation of ENaC-mediated Na⁺ absorption in the distal renal tubule [139].