- Top of page
Adenosine is an important signalling molecule involved in a large number of physiological functions. In the brain these processes are as diverse as sleep, memory, locomotion and neuroprotection during episodes of ischaemia and hypoxia. Although the actions of adenosine, through cell surface G-protein-coupled receptors, are well characterized, in many cases the sources of adenosine and mechanisms of release have not been defined. Here we demonstrate the activity-dependent release of adenosine in the cerebellum using a combination of electrophysiology and biosensors. Short trains of electrical stimuli delivered to the molecular layer in vitro, release adenosine via a process that is both TTX and Ca2+ sensitive. As ATP release cannot be detected, adenosine must either be released directly or rapidly produced by highly localized and efficient extracellular ATP breakdown. Since adenosine release can be modulated by receptors that act on parallel fibre–Purkinje cell synapses, we suggest that the parallel fibres release adenosine. This activity-dependent adenosine release exerts feedback inhibition of parallel fibre–Purkinje cell transmission. Spike-mediated adenosine release from parallel fibres will thus powerfully regulate cerebellar circuit output.
Adenosine is an important neuromodulator in the central nervous system, playing a role in a plethora of physiological and pathophysiological processes. The action of adenosine on cell surface receptors is well defined with A1, A2a, A2b and A3 receptors all cloned (Fredholm et al. 2000). Although extensively studied, the cellular source and mechanisms of adenosine release remain unclear (for review see Latini & Pedata, 2001). Adenosine can in principle gain access to the extracellular space by the breakdown of ATP, by translocation from cell cytoplasm via nucleoside transport proteins or possibly by the exocytosis of adenosine itself. There has been considerable investigation of adenosine release during pathological episodes such as hypoxia, ischaemia and hypercapnia as adenosine is neuroprotective (Rudolphi et al. 1992; Fredholm, 1997; Dale et al. 2000; Dulla et al. 2005). Release under these conditions is often Ca2+ independent, relatively insensitive to TTX and is not mediated via glutamate receptor activation. In contrast, little is known about the physiological release of adenosine with few examples where a role and cellular source of adenosine have been identified (but see Dale, 1998). In many cases, adenosine release is evoked with stimuli such as high K+, prolonged electrical stimulation and glutamate receptor activation (Latini & Pedata, 2001). The physiological relevance of these experiments is unclear.
The presence of adenosine, adenosine deaminase and A1 receptors in the cerebellar cortex (Braas et al. 1986; Geiger & Nagy, 1986; Rivkees et al. 1995) strongly suggests that adenosine plays an important role in cerebellar function. The activation of A1 receptors inhibits synaptic transmission between parallel fibres and Purkinje cells (Kocsis et al. 1984). These receptors are tonically activated by endogenous adenosine, since application of A1 receptor antagonists enhances synaptic transmission (Takahashi et al. 1995; Dittman & Regehr, 1996). The source of this adenosine has not been determined but could arise from the release of adenosine or the release of ATP and its subsequent metabolism. A recent report has suggested that ATP can be released from parallel fibres (Beierlein & Regehr, 2006).
To examine this issue we have used selective and sensitive microelectrode biosensors (Llaudet et al. 2003) to measure the release of adenosine from cerebellar slices in real time. These biosensors are small enough (25–50 μm diameter) to place either in or close to defined areas in cerebellar slices. Here we report that adenosine can be released from the molecular layer by using a physiological stimulus, short bursts (1–10 s) of focal electrical stimuli at the same voltage used to elicit synaptic transmission. The adenosine release is both TTX and Ca2+ sensitive and does not appear to arise from the extracellular metabolism of ATP. Modulation of parallel fibre–Purkinje cell synaptic transmission can increase or decrease adenosine release, strongly suggesting that parallel fibres are involved in adenosine release.
- Top of page
We have demonstrated that adenosine can be released from the molecular layer of cerebellar slices by electrical stimulation. This release of adenosine was via a process that is both TTX and Ca2+ sensitive and as ATP release cannot be detected, adenosine is either released directly or rapidly produced by efficient extracellular ATP breakdown. Since adenosine release was modulated by receptors that act on parallel fibre–Purkinje cell synapses, parallel fibres are the most likely source of adenosine. No previous study has measured adenosine release from cerebellar slices, although adenosine can be released from cultured granule cells (Schousboe et al. 1989; Philibert & Dutton, 1989; Sweeney, 1996). Prolonged electrical stimulation (1–5 min) can release adenosine in other brain regions (the cortex, hippocampus and striatum, Pedata et al. 1990; Lloyd et al. 1993). However, we have characterized the properties of adenosine release evoked by a plausibly physiological stimulus: short duration trains (1–10 s), localized to a small area of the slice and at the same minimal strength that evokes transmitter release.
What is the source of adenosine?
We have provided strong evidence that parallel fibre activity is required for adenosine release. Firstly, the stimulating electrode and biosensors were arranged along a beam of parallel fibres; thus, parallel fibres are activated by the stimulus and the sensors are in the correct place to measure what is released. Secondly, activation or inhibition of the G-protein coupled receptors (GABAB, A1 and mGluR4) present on parallel fibre terminals modulates adenosine release by a similar magnitude as synaptic transmission. This proportionality in the reduction of EPSP amplitude and adenosine release suggests a common mechanism for adenosine and glutamate release. However, there are apparent differences. Our results suggest that the dependence of adenosine release on extracellular Ca2+ is ∼1 whereas previous studies have shown that the Ca2+ dependence of glutamate release at parallel fibre synapses is ∼3 (Mintz et al. 1995; Brown et al. 2004). This difference in Ca2+ dependence may be more apparent than real as different stimuli were used to elicit glutamate and adenosine release. Single stimuli were used to evoke glutamate release (Mintz et al. 1995; Brown et al. 2004) whereas in this study, trains of stimuli were required to release adenosine. The Ca2+ dynamics during a train of stimuli will be more complex than for a single stimulus and may result in intracellular Ca2+ accumulation thus reducing the apparent dependence on extracellular Ca2+. Thirdly, it is physiologically plausible that parallel fibres maintain transmitter release at frequencies up to and above the stimulation frequency used in this study (Kreitzer & Regehr, 2000). Thus, parallel fibre-dependent adenosine release could occur physiologically and act to limit parallel fibre-dependent excitation of Purkinje cells. Finally, moving the biosensor further away (along the same beam of parallel fibres) resulted in only a small reduction in current amplitude with little change in rise time.
Other possible sources of adenosine include Purkinje cells, interneurones and glia. Stimulation in the molecular layer will activate all these cell types but the TTX sensitivity of adenosine release makes Bergmann glia an unlikely source of adenosine as they do not fire action potentials (Clark & Barbour, 1997). Purkinje cells are strongly immunoreactive for adenosine (Braas et al. 1986) and express GABAB receptors in their dendrites (Lujan & Shigemoto, 2006) but there is little evidence that Purkinje cells express either A1 adenosine receptors or mGluR4 receptors. The restricted expression of mGluR4 receptors also means adenosine release from molecular layer interneurones is unlikely (Mateos et al. 1998). We can also exclude glutamate released from parallel fibres activating receptors on neurones and glia, since adenosine release was not blocked by the glutamate receptor antagonists CNQX and AP5. Furthermore Purkinje cells, interneurones and (possibly) Bergmann glia will be shunted by application of the GABAA receptor agonist muscimol, which had no effect on adenosine release. Thus, the most likely source of adenosine is the parallel fibres. However, on several occasions it has been possible to stimulate parallel fibres and record EPSPs without observing release of detectable amounts of adenosine. Thus, perhaps only a subset of parallel fibres release adenosine.
Properties of adenosine release
Some of the characteristics of adenosine release appear different from parallel fibre–Purkinje cell synaptic transmission. Firstly, a single stimulus is sufficient to evoke glutamate release from parallel fibres, yet a train of such stimuli is required to produce an adenosine signal on the biosensor. This could simply reflect that enough adenosine has to be released to diffuse through the tissue to reach the sensor on the slice surface. Clearly the distance between synaptic glutamate receptors and the release sites is much smaller than that between the adenosine biosensor and release sites. Thus, the requirement for trains of stimuli could be an artefact of the recording system. However, similar stimulation protocols were required when the sensors were pushed into the molecular layer (presumably the sensor will be closer to the release sites). At the calyx of Held, trains of stimuli (10 Hz) release adenosine, which can be detected indirectly through the activation of A1 receptors and the consequent inhibition of transmitter release. This inhibition only reaches significance after the first 20 stimuli in a train, suggesting that multiple stimuli are required to release adenosine (Kimura et al. 2003; Wong et al. 2006). In the hippocampus, 10 Hz but not 3 Hz stimulation released adenosine (measured by inhibition of EPSCs, Brager & Thompson, 2003). Secondly, once released the time course of adenosine was very slow. The decay of adenosine takes around 130 s which is vastly longer than that for classical fast transmitters like glutamate. This is similar to neuropeptides, which have a long duration of action and act in a paracrine manner affecting many neurones. This may reflect the large number of parallel fibres that are stimulated resulting in the spill over of adenosine and diffusion over a large area.
Mechanism of adenosine release?
The inhibition of ENT1 and ENT2 had no effect on adenosine release and thus adenosine was not transported across the cell membrane by these equilibrative transporters. NBTI and dipyridamole do inhibit adenosine transport leading to greater synaptic inhibition in the cerebellum (M. J. Wall, personal observation) and in the hippocampus (Frenguelli et al. 2007). However, it is possible that other transporters, which are insensitive to NBTI/dipyridamole, are involved in adenosine release. Adenosine release was blocked by TTX and by removal of Ca2+. The calcium dependence of adenosine release was lower than that reported for classical neurotransmitters, but was similar to that reported for the release of neuropeptides (for example see Peng & Zucker, 1993). The adenosine release thus has the characteristics associated with conventional synaptic release. Nevertheless there are three alternate interpretations of our data. Firstly, an intermediate transmitter could in principle cause the downstream release of adenosine. This seems unlikely as we have eliminated the obvious candidates for such an intermediate transmitter and have shown that inhibiting neurones and glia (through GABAA receptor activation) has little effect on adenosine release. Secondly, action potential activity in parallel fibres could release potassium which depolarizes other cells (including glia) resulting in adenosine release. This seems unlikely as activation of parallel fibre receptors (GABAB, A1 and mGlu4R) inhibits adenosine release but would not affect the amount of potassium released during action potentials. Furthermore any cells depolarized by the potassium efflux would probably be shunted by GABAA receptor activation suggesting that this treatment should diminish release if that were the case.
Thirdly, adenosine could arise from the breakdown of exocytotically released ATP. Beierlein & Regehr (2006) have recently described parallel fibre-mediated Ca2+ rises in Bergmann glia that are blocked by PPADs (50 μm) and may thus depend on parallel fibre-mediated ATP release. However, we have been unable to measure any ATP release (either with or without ecto-ATPase inhibition). Thus, the ATP released from parallel fibres, as reported by Beierlein & Regehr (2006), may be too small to detect by our methods. This contrasts with studies in the retina where glial cells release ATP which is rapidly converted by ectoenzymes into adenosine. In this case, the ATP can be detected using luciferin–luciferase chemiluminescence and blockers of ATP breakdown reduce the production of adenosine (reviewed by Newman, 2004). In the hippocampus, it has been reported that exogenous ATP is converted to adenosine in less than a second (Dunwiddie et al. 1997). More recent experiments demonstrated that only a small proportion (∼7%) of bath-applied ATP is converted, albeit rapidly to adenosine by hippocampal slices (Frenguelli et al. 2007). Our experiments suggest that in the cerebellum, only 10–20% of ATP is rapidly converted to adenosine. Furthermore biosensor measurements of the breakdown of bath-applied ATP show that less than 10% of the applied ATP is converted to adenosine (Wall MJ & Dale N, personal observations). For the complete conversion of parallel fibre ATP to occur, high densities of ectonucleotidases would have to be localized at parallel fibre synapses around the site of release such that no ATP escaped the confines of the synapse. Although an ecto-ATPase (CD39) is present on the soma and dendrites of Purkinje cells (Wang & Guidotti, 1998) and 5′-nucleotidase activity is present on Bergmann glia membranes (Schoen et al. 1987), the density of these enzymes is unclear and this explanation seems to us unlikely. A direct mechanism of adenosine release is supported by Wong et al. (2006) who demonstrated that trains of stimuli release adenosine at the calyx of Held (detected by A1 receptor mediated-inhibition of EPSCs) with no detectable release of ATP (measured by means of the luciferin–luciferase method). In the hippocampus, the adenosine released by trains of stimuli (detected by A1 receptor-mediated inhibition of EPSCs) could not be reduced by blocking 5′-ectonucleotidases (Mitchell et al. 1993; Brager & Thompson, 2003). Frenguelli et al. (2007) have also shown that adenosine release during ischaemia does not arise from the extracellular metabolism of ATP.
The functional significance
Irrespective of the detailed mechanism of adenosine release, its dependency on parallel fibre activity is potentially very significant. During periods of high parallel fibre activity adenosine will be released resulting in an increased extracellular concentration of adenosine and greater inhibition of parallel fibre–Purkinje cell synapses. Although the increased synaptic inhibition will reduce the amplitude of EPSPs, the effects on high-frequency synaptic transmission will probably be more important. During high-frequency parallel fibre activity the increased synaptic inhibition will reduce glutamate depletion and thus maintain transmission to Purkinje cells (Kimura et al. 2003; Wong et al. 2006). Activation of A1 receptors by endogenous adenosine at the calyx of Held causes a reduction in the relative amplitude of EPSCs at the end of the train compared with the EPSCs at the beginning. This is because there is no ambient level of adenosine at the calyx of Held and adenosine is released and accumulates during the stimulation, reaching a significant concentration late in the train. In contrast, Billups et al. (2005) have shown that activation of presynaptic group III mGluRs has no net effect on the amplitude of EPSCs. However, there is an increase in the size of the readily releasable vesicle pool which is balanced by a reduction in the probability of release. Thus, the synaptic state of the synapse has changed and the metabolic demand has been reduced. We have not measured the actions of adenosine during a train of EPSPs but have investigated recovery from depression following a train. We found that blocking A1 receptors greatly speeds the recovery of EPSP depression. Similar observations have been observed following trains of stimuli at the calyx of Held, where blocking group III mGluRs increased the speed of recovery (Billups et al. 2005) while GABAB receptor activation slowed recovery (Sakaba & Neher, 2003). The action of GABAB receptors appears to be through retardation of synaptic vesicle recruitment during sustained activity.
The dynamics of adenosine signalling are likely to be complex as release of adenosine auto-inhibits its own release. An intriguing question is whether adenosine released from active parallel fibres could laterally inhibit neighbouring but inactive parallel fibres. Interestingly adenosine is also vaso-active and there is evidence that periods of activity release adenosine resulting in the dilatation of cerebellar blood vessels (Li & Iadecola, 1994).