Skip to main content
PMC home page
The Journal of Physiology logoLink to The Journal of Physiology
. 1998 May 1;508(Pt 3):691–701. doi: 10.1111/j.1469-7793.1998.691bp.x

ATP and glutamate are released from separate neurones in the rat medial habenula nucleus: frequency dependence and adenosine-mediated inhibition of release

Susan J Robertson 1, Frances A Edwards 1
PMCID: PMC2230910  PMID: 9518726

Abstract

  1. ATP and glutamatergic synaptic currents were compared in slices of rat medial habenula nucleus using whole-cell patch-clamp techniques.

  2. In most cells low voltage stimulation resulted in glutamatergic responses and not purinergic responses. In five cells where ATP currents could be stimulated with low voltages, wash out of glutamate antagonists did not reveal evoked glutamate currents. Spontaneous glutamate currents confirmed washout of antagonist.

  3. Modulation of release probability of glutamate and ATP, assessed by changes in failure rate of synaptic currents, was compared under conditions of different stimulation frequencies and in the presence of adenosine agonists and antagonists.

  4. ATP release, but not glutamate release, was shown to be modulated by increased stimulation frequency which resulted in inhibition of ATP release via A2-like adenosine receptors. A1 receptors caused inhibition of both ATP and glutamate release.

  5. Endogenous adenosine inhibited glutamate release via A1 receptors but only inhibited ATP release via A2-like receptors.

  6. Attempts to inhibit the degradation of ATP to adenosine did not alter the frequency dependence of the failure rate.

  7. We conclude, from the direct demonstration and from the differences in pharmacology and frequency dependence of the modulation of release, that ATP and glutamate responses are due to release from separate neurones.

P2X receptors are ligand gated cation channels present on many neuronal and smooth muscle preparations (Suprenant, Buell & North, 1995). Recently several members of this family have been cloned, and shown to define a new structural class of ligand gated ion channel (Buell, Collo & Rassendren, 1996). ATP has been shown to act as a fast excitatory transmitter in the peripheral nervous system (for review see Burnstock, 1990), including at neuroneuronal synapses (Evans, Derkach & Surprenant, 1992; Galligan & Bertrand, 1994). P2X receptors also mediate central fast synaptic transmission in the rat medial habenula nucleus (Edwards, Gibb & Colquhoun, 1992) and the rat dorsal horn (Bardoni, Goldstein, Lee, Gu & MacDermott, 1997). In the present study we further characterize purinergic transmission in the rat medial habenula. The putative transmitter at these purinergic synapses is ATP, though we cannot discount the possibility that UTP or another unknown transmitter, which acts at P2X receptors, mediates this response. For the sake of simplicity the transmitter will be referred to as ATP throughout this paper. In the light of the role of ATP as a cotransmitter in peripheral synapses, it has been suggested that ATP could also act as a cotransmitter in the brain. The dominant excitatory transmitter in these cells is glutamate and thus this seemed to be the most likely candidate for a transmitter with which ATP might be released.

Recently we reported that ATP release is dependent on the frequency of stimulation and that the failure to release at high stimulation frequencies is mediated by adenosine receptors (Edwards, Robertson & Gibb, 1997). Here we investigate further the pharmacology of this effect and, amongst other methods, we use this feature of ATP receptor-mediated transmission as a tool to investigate the possibility of cotransmission with glutamate.

We have previously demonstrated that ATP synapses show many of the classical features of other fast synapses in the mammalian brain. The quantal size of the currents (as assessed by miniature currents), the amplitude distribution of miniatures, TTX block of evoked currents and the calcium dependence of the release process all seem to be similar to those shown at many glutamate or GABAergic synapses in the brain (Edwards et al. 1992, 1997; Stevens, 1993). Thus, though not yet confirmed anatomically, it is likely that ATP is stored in vesicles and released from an axon with one or more boutons, as seen elsewhere in the brain for other fast central synapses (for review see Edwards, 1995). Assuming this is the case, glutamate and ATP cotransmission could, if at all, occur in three possible forms: (1) glutamate and ATP could be co-stored in the same vesicles and thus always be released together; (2) glutamate and ATP could be stored in different vesicles but the two types of vesicles could be situated in the same boutons; and (3) glutamate and ATP could be released from different boutons but from the same axon.

We have used various techniques to investigate these possibilities. The method which directly tests whether stimulation of a particular axon will cause release of both transmitters is to identify an ATP input (in the presence of blockers of glutamate and GABA receptors) and then wash out the glutamate antagonist and see if a glutamate response can also be seen with the same stimulation. Alternatively a glutamate input can be similarly identified and investigated for cotransmission. This is only appropriate with protocols allowing very local stimulation, which is an important limitation in the case of the very sparse ATP inputs (discussed further below).

Where this method is not practical, transmission was also studied by comparison of the release processes for the two transmitters. If both transmitters are packaged together in the same vesicles then the release probability for both transmitters should be identical under a variety of conditions. To test this question we have compared the effects of varying stimulus frequency on the rate of failures seen in the two types of synapse.

Adenosine agonists are thought to cause presynaptic inhibition of transmitter release via modulation of presynaptic calcium or possibly potassium channels (White, Dickenson, Alexander & Hill, 1992; Collis & Hourani, 1993; Wu & Saggau, 1994; Umemiya & Berger, 1994; Budd & Nicholls, 1995; Dittman & Regehr, 1996; Ambrósio, Malva, Carvalho & Carvalho, 1996). We have thus used comparison of the dose-dependent effects of different adenosine receptor agonists and antagonists to investigate the possibility that ATP and glutamate might be packaged in different vesicles but released from the same bouton.

This work has been described in part in abstract form (Robertson & Edwards, 1997).

METHODS

Solutions and drugs

Slices were maintained in a standard Krebs solution of composition (mM): NaCl, 125; KCl, 2.5; NaHCO3, 26; NaH2PO4, 1.25; glucose, 25; CaCl2, 2; MgCl2, 1 (Sigma); bubbled with 95 % O2-5 % CO2. The intracellular solution contained (mM): CsCl, 133; Hepes, 10; EGTA, 10; CaCl2, 1; Mg-ATP, 2 (Sigma); pH 7.3 with CsOH. For recording purinergic EPSCs, the extracellular solution contained, in addition, bicuculline (10 μM) to block GABAA currents, CNQX (10 μM) to block AMPA currents and 7-chlorokynurenate (5 μM) to block NMDA currents (antagonists Tocris Cookson Inc.). Glutamatergic EPSCs were recorded as above but in the absence of CNQX. Suramin (30 μM; Research Biochemicals) was used to confirm that the synaptic current was mediated by P2 receptors. 8-Cyclopentyltheophylline (8CPT; 1, 5 and 10 μM; Research Biochemicals) was used to block adenosine receptors. Adenosine receptor agonists used were 2-chloro-N6-cyclopentyladenosine (CCPA; 100 nM; selective for A1 receptors) and N6-[2-(3,5-dimethoxyphenyl)-2-(2-methylphenyl)-ethyl]adenosine (DPMA; 10 and 100 nM; selective for A2 receptors; Research Biochemicals). Drugs were applied by addition to the perfusion solution.

Habenula slices

Standard methods were used to prepare parasagittal slices from 18- to 24-day-old male or female rats (Edwards, Konnerth, Sakmann & Takahashi, 1989). Animals were killed by decapitation and the brain was quickly removed and placed in cold solution. After hemisection along the mid-line, the front third of the brain was removed with a scalpel. The rear portion of the hemisphere was then trimmed so it could be glued to the tissue block of a Camden Vibra-slice (Loughborough, UK) with the mid-line surface uppermost. Parasagittal slices (200 μm thick) were cut under a dissecting microscope and the habenula region dissected free using the position of the hippocampus and the stria medularis and fasciculus retroflexus of Meynert fibre tracts as guides. After cutting, slices were incubated for around 45 min at 34°C before recordings began. All recordings were made at room temperature (21–24°C).

Recording and analysis of synaptic currents

Synaptic currents were recorded while holding the membrane potential at -70 mV with an Axopatch-1D patch-clamp amplifier (Axon Instruments) using standard methods (Edwards et al. 1989). Thick-walled borosilicate glass electrodes (1B150F-3, World Precision Instruments) were pulled to give a resistance of approximately 5 MΩ when filled with intracellular solution.

Stimulating electrodes were pulled in the same way as recording electrodes and filled with normal extracellular solution. Currents were evoked by applying rectangular stimulus voltage pulses of 100–200 μs duration (stimulus supplied by a Grass SD9 stimulator). The voltage used for stimulation was chosen by placing the stimulating electrode in the slice and slowly increasing the voltage from 1 V up to 100 V until a response was seen. In the case of glutamate currents, inputs were found with responses generally occurring at < 20 V on the first or second placement of the stimulating electrode (range, 1–52 V; mean ±s.e.m., 15.6 ± 1.6 V, n = 47). Having found a response, increasing further the stimulus voltage increased the size of the average response presumably by recruiting more fibres as the spread of the stimulus increased. The voltage chosen was the lowest voltage at which a response was clearly seen, in an attempt to stimulate only one or a few fibres. In contrast ATP inputs were much sparser so that it commonly took ten or more electrode placements and in most cases the necessity of increasing the voltage to considerably higher levels before any input was found (range, 8.5–100 V; mean ±s.e.m., 41.4 ± 3.5 V, n = 50). In the case of ATP responses, increasing the voltage further had little or no effect on the size of the mean current, confirming the sparse nature of the input. Once a minimal stimulus voltage was established this was maintained throughout the experiment.

To study the effects of changing the rate of stimulus, the frequency was increased through the range (0.5–10 or 100 Hz) and then returned to 1 Hz. At least 100 stimuli were recorded at each stimulus frequency. In some cases the order of frequencies was altered without affecting the result. During recordings, in order to verify the stability and quality of the seal, a test pulse was monitored regularly and the recording terminated if the shape or size of the pulse was significantly altered.

Currents were digitized at 44 kHz and recorded at a bandwidth of 10 kHz (4 pole Bessel) on videocassette tape using a Vetter VCR (USA). Currents were replayed through a 2 kHz filter (Frequency Devices, 8 pole Bessel) and sampled at 10 kHz with a Labmaster DMA interface (Axon Instruments) using the program WCP (kindly supplied by Dr J. Dempster, University of Strathclyde Electrophysiology Software).

All records were viewed manually and in most cases failures could be unambiguously distinguished from responses. A response was determined to be an evoked synaptic event if there was a short consistent latency (< 3 ms) from the stimulus artifact and a fast rise time, usually < 1 ms. The percentage of failures was then recorded. It is possible that small events were occasionally misclassified and so to rule out bias the average currents (including failures) were analysed and in all cases the results were consistent with the change in percentage of failures. In some cells the analysis was also performed blind by a second investigator. As discussed in detail previously (Edwards et al. 1997), the stimulus artifact has a small, prolonged tail, which adds an artifact in the automatic measurement used for figures showing scatter plots. In addition the program used, WCP, measures the minimum point as the peak and hence adds half the mean of the baseline noise to all measures. This only affects the display of the figures and not the analysis but means that failures do not cluster apparently around zero.

Statistics

Significance levels of effects were tested with two-way analysis of variance (factor 1 drug; factor 2 frequency of stimulation) or a one-way analysis of variance (to test significance of stimulation frequency on failure rate) using the statistics package ‘Prism’ (Graphpad Software Inc., San Diego, CA, USA). Results were considered significant if the probability of chance occurrence was less than 0.05. All changes referred to in the following sections were significant by this criterion. When comparing the effect of 100 Hz stimulation on the failure rate, Student's paired t test was used. Results are expressed as means ±s.e.m.

RESULTS

The direct test: are ATP and glutamate released from the same axon?

As described in Methods, the ATP inputs were sparse and hence we generally needed several electrode placements and high voltages to find them. Despite the sparsity of the inputs, in five cells we were able to position the stimulation electrode such that ATP responses were observed with very low stimulation voltages (< 16 V). Only at such low stimulus voltages would it be possible to avoid also stimulating the glutamatergic axons which densely innervate the slice (as discussed in Methods). When ATP release was achieved with such low voltage stimulation (n = 5), we then washed out the CNQX. Although spontaneous glutamate currents could be detected under these conditions, verifying that the antagonist had been effectively removed, no change in the evoked response was seen (Fig. 1), showing that this stimulation did not result in an evoked glutamate response. The evoked current was also confirmed to be purinergic as it was inhibited by suramin (30 μM; Fig. 1C). Thus in these five cells purinergic transmission occurred from one or a few axons in the absence of glutamatergic responses. Note that placement of the electrode in most positions in the slice results in a glutamate response whereas ATP responses are much more difficult to find. Thus it is clear that the slice is densely innervated with fibres which when stimulated result in a glutamate response but no ATP response. The purinergic response can be clearly distinguished from the glutamatergic response not only pharmacologically but also due to the slower decay of the evoked purinergic EPSCs (purinergic EPSCs decay, τ∼20 ms; glutamatergic EPSCs decay, τ < 5 ms). To test whether the lack of glutamatergic response at the purinergic synapses was general rather than a speciality of the few synapses where we could find a response to low voltage stimulation, we went on to compare the release properties of glutamatergic and purinergic synapses.

Figure 1. Direct test to determine if purinergic and glutamatergic transmission occurs from the same input.

Figure 1

Open in a new tab

A, fifteen superimposed traces illustrating evoked currents in the presence of CNQX (10 μM) with a low stimulus voltage (13 V). The mean evoked currents from this experiment (shown in C) exhibited a decay with a time constant, τ = 24 ms, consistent with that of ATP-mediated currents. B, fifteen superimposed traces with a low voltage stimulus (13 V), after washout of CNQX. The occurrence of spontaneous currents with a decay constant of a few milliseconds, consistent with AMPA-mediated spontaneous currents, confirmed washout of CNQX. The kinetics of mean evoked ATP receptor-mediated current did not alter on washout of CNQX (C). The ATP-mediated current was inhibited by 30 μM suramin (C). * Stimulus artifact.

The effects of stimulation frequency on glutamate and ATP release (Figs 2 and 3)

Figure 2. Effect of stimulation frequency on failure to release transmitter, for both ATP- and glutamate-mediated synaptic currents.

Figure 2

Open in a new tab

A, mean evoked ATP receptor-mediated synaptic currents, including failures, at stimulation frequencies 1, 2 and 5 Hz. Stimulation voltage 26 V for 200 μs. B, amplitudes of individual ATP receptor-mediated current evoked at stimulation frequencies of 0.5–100 Hz. Data are from the same cell as in A. The percentage failure to release was: 0.5 Hz, 24 %; 1 Hz, 43 %; 2 Hz, 54 %; 5 Hz, 96 %; 10 Hz, 97 %; 100 Hz, 96 %. C, mean evoked glutamate-mediated synaptic currents, including failures, at stimulation frequencies of 1, 5 and 10 Hz. Stimulation voltage 26 V for 200 μs. D, individual glutamate receptor-mediated current amplitudes evoked at stimulation frequencies of 1–100 Hz. Data are from the same cell as in C. The percentage failure to release was: 1 Hz, 12 %; 2 Hz, 10 %; 5 Hz, 24 %; 10 Hz, 34 %; 100 Hz, 99 %. The percentage failure rate was calculated by counting the number of times where stimulation of the presynaptic neurone failed to evoke a detectable synaptic current. At least 100 stimuli were used at each frequency.

Figure 3. Comparison of the dependence of failure rate on stimulation frequency, for evoked ATP and glutamate release.

Figure 3

Open in a new tab

Mean data for the percentage failure rate at stimulation frequencies of 0.5–100 Hz for both glutamate-mediated synaptic transmission (□), n = 22, and ATP receptor-mediated synaptic transmission (•), n = 51 (n = 10 at 10 and 100 Hz for both ATP and glutamate-mediated currents). (Note that equivalent results were obtained if the responses to all stimuli were averaged.)

Previously we have shown that ATP release is dependent on the frequency of stimulation and that this effect can be blocked by a high concentration of the adenosine antagonist 8CPT (Edwards et al. 1997). Here we further investigate the adenosine receptors mediating this inhibition and compare the effects in ATP synapses to those observed in glutamatergic synapses. If the transmitters were stored in the same vesicles, identical frequency-dependent effects would be expected.

As the frequency of stimulation was increased from 0.5 to 100 Hz the mean amplitude, including failures, of purinergic EPSCs decreased and the percentage failure rate increased to almost 100 %. This effect was highly statistically significant (0.5–10 Hz, P < 0.0001). Figure 2A shows an example of means of consecutive purinergic EPSCs in a cell where the purinergic input was stimulated at a range of frequencies. Figure 2B shows a scatter plot of the amplitudes of individual currents from the same cell. Figure 3 (filled circles) shows the relationship between stimulation frequency and mean failure rate of purinergic EPSCs (n = 51 cells). Note that the failure rate returned very quickly to the original low level on decreasing the stimulation frequency from 100 Hz back to 1 Hz (1 Hz, 20.6 ± 5.3 %; 1 Hz after 100 Hz, 27.0 ± 6.1 %, n = 10) and moreover that the relationship was not changed if the order of applied stimulation frequencies was varied (e.g. 1, 2, 0.5, 5, 10, 100, 1).

In contrast the amplitude and failure rate for glutamatergic EPSCs were relatively independent of stimulation rate up to about 5 Hz (Figs 2C and D, and 3). In fact, in all cells, as stimulation frequency was increased, some facilitation of the glutamate response was seen. There was some variability between cells as to whether the minimum failure rate occurred at 2 or 5 Hz. At 10–100 Hz the frequency of failures of glutamatergic and purinergic EPSCs was very similar. Unlike purinergic EPSCs, the failure rate of the glutamatergic EPSC did not immediately return to control levels after a brief stimulation (∼1 s) at 100 Hz (1 Hz, 45.0 ± 7.3 %; 1 Hz after 100 Hz, 58.0 ± 8.0 %; P < 0.05, n = 10). The relatively slow recovery of glutamatergic EPSCs after 100 Hz stimulation suggests that an additional mechanism influences probability of release at these high frequencies. Due to the delay in recovery of the glutamatergic EPSCs all the subsequent experiments were restricted to frequencies up to 10 Hz. Figure 3 (open squares) shows the relationship between stimulation frequency and glutamatergic EPSC failure rate (n = 22 cells) compared with the result for purinergic EPSCs. In three cells the results above have been confirmed for both purinergic and glutamatergic synapses in the same cell.

It is thus clear that at frequencies from 0.5 to 5 Hz, the relationship between stimulation frequency and release is very different for purinergic EPSCs and glutamatergic EPSCs, and thus that ATP and glutamate cannot possibly be packaged in the same vesicles.

We went on to study the role of different adenosine receptor subtypes in the modulation of failure rates in the purinergic and glutamatergic transmission.

The pharmacological profile of presynaptic adenosine receptors at ATP and glutamatergic synapses

A1 receptor agonist: 2-chloro-N6-cyclopentyladenosine (CCPA)

See Lohse, Klotz, Schwabe, Cristalli, Vittori & Grifantini, 1988; Klotz, Lohse, Schwabe, Cristalli, Vittori & Grifantini, 1989.

The failure rates of both ATP- and glutamate-mediated currents were increased by the A1 receptor agonist CCPA (100 nM) (ATP, P < 0.0001; glutamate, P < 0.01; Fig. 4). At lower stimulation frequencies (0.5–5 Hz), CCPA increased the failure rate of ATP currents by about 35 % (e.g. at 1 Hz, 18.3 ± 3.2 % up to 54.7 ± 15.6 %). At 10 Hz, in the presence of CCPA, failures increased by about 10 %. With 100 nM CCPA the effect of stimulation frequency on release of ATP was no longer statistically significant (Fig. 4A and B). This may reflect an approach to maximum adenosine-mediated inhibition at high stimulation frequencies. Glutamate release was also inhibited by 100 nM CCPA (Fig. 4C and D). The increase in percentage failures was 25 % (e.g. at 1 Hz, 47.1 ± 14.8 % up to 72.5 ± 11.9 %) throughout. For both transmitters the effect of CCPA was fully reversible on washout and was completely reversed by the antagonist 8CPT (1 μM; Fig. 4) (glutamatergic EPSCs, n = 4, P < 0.01: purinergic EPSCs, n = 3, P < 0.0001) as would be expected if the effect were mediated by A1 receptors.

Figure 4. The effect of the adenosine receptor agonist CCPA on the percentage failure rate for both ATP- and glutamate receptor-mediated synaptic transmission. Low concentrations of 8CPT can reverse the effect of CCPA.

Figure 4

Open in a new tab

A, mean ATP receptor-mediated currents stimulated at 1 Hz. I, control; II, 100 nM CCPA; III, 100 nM CCPA + 1 μM 8CPT; IV, 100 nM CCPA + 5 μM 8CPT. Calculated percentage failure: control, 27 %; 100 nM CCPA, 51 %; 100 nM CCPA + 1 μM 8CPT, 25 %; 100 nM CCPA + 5 μM 8CPT, 22 %. Stimulation voltage 20 V for 200 μs. B, mean data for all cells. ▪, control (n = 5); ▴, 100 nM CCPA (n = 5); □, 100 nM CCPA + 1 μM 8CPT (n = 3); and ○, 100 nM CCPA + 10 μM 8CPT (n = 3). C, mean glutamate-mediated currents stimulated at 1 Hz. I, control; II, 100 nM CCPA; III, 100 nM CCPA + 1 μM 8CPT. Calculated percentage failure: control, 17 %; 100 nM CCPA, 46 %; 100 nM CCPA + 1 μM 8CPT, 7 %. Stimulation voltage 20 V for 200 μs. D, mean data for all cells. ▪, control (n = 4); ▴, 100 nM CCPA (n = 4); □, 100 nM CCPA + 1 μM 8CPT (n = 3). CCPA is reported to show a high selectivity for A1 (Ki≈0.4 nM) over A2 receptors (Ki≈3.9 μM).

A2 receptor agonist:

N6-[2-(3,5-dimethoxyphenyl)-2-(2-methylphenyl)-ethyl]adenosine (DPMA). For ATP currents the application of 10 nM DPMA (n = 5), selective for A2 receptors (Bridges, Bruns, Ortwine, Priebe, Szotek & Trivedi, 1988), increased the failure rate at 1 Hz more than 2-fold (data not shown). A higher concentration (100 nM) was tested throughout the 1–10 Hz range and increased the failure rate at all frequencies (P < 0.0001, n = 14). This effect, which was not antagonized by 1 μM 8CPT (n = 8) but was completely blocked by 10 μM 8CPT (P < 0.001, n = 3; Fig. 5A and B), would be consistent with A2-like adenosine receptor activation.

Figure 5. The effect of the adenosine receptor agonist DPMA on the percentage failure rate for both ATP- and glutamate-mediated synaptic transmission. Higher concentrations of 8CPT can reverse the effect of DPMA.

Figure 5

Open in a new tab

A, amplitudes of ATP receptor-mediated currents stimulated at1 Hz. Calculated percentage failure: control, 47 %; 100 nM DPMA, 94 %; 100 nM DPMA + 1 μM 8CPT, 87 %; 100 nM DPMA + 10 μM 8CPT, 25 %. Stimulation voltage 16 V for 200 μs. B, mean data for all cells. ▪, control (n = 16); ▴, 100 nM DPMA (n = 14); □, 100 nM DPMA + 1 μM 8CPT (n = 8); and ○, 100 nM DPMA + 10 μM 8CPT (n = 3). C, amplitudes of glutamate-mediated currents stimulated at 1 Hz. Calculated percentage failure: control, 83 %; 100 nM DPMA, 80 %; 100 nM DPMA + 1 μM 8CPT, 63 %; 100 nM DPMA + 10 μM 8CPT, 34 %. Stimulation voltage 52 V for 200 μs. D, mean data for all cells. ▪, control (n = 5); ▴, 100 nM DPMA (n = 5); ○, 100 nM DPMA + 10 μM 8CPT (n = 4). DPMA is reported to have some selectivity for A2 (Ki≈4.4 nM) over A1 receptors (Ki≈142 nM).

In contrast to the effect on ATP release, the A2 receptor agonist DPMA did not inhibit glutamate release. Figure 5C and D shows the lack of effect of DPMA on glutamate release.

Thus ATP transmission is inhibited by activation of both A1 and A2 receptors while glutamate transmission is modulated only by A1 and not by A2 receptors.

The role of background adenosine in modulating release of glutamate and ATP (Fig. 6)

Figure 6. The effect of the adenosine receptor 8CPT on the percentage failure rate for both ATP- and glutamate-mediated synaptic transmission.

Figure 6

Open in a new tab

A, mean purinergic currents stimulated at 10 Hz. I, control; II, 1 μM 8CPT; III, 5 μM 8CPT; IV, 10 μM 8CPT. Calculated percentage failure: control, 61 %; 1 μM 8CPT, 70 %; and 10 μM 8CPT, 2 %. Stimulation voltage 38 V for 200 μs. B, mean data for all cells. ▪, control (n = 9); □, 1 μM 8CPT (n = 5); ○, 10 μM 8CPT (n = 3) from Edwards et al. (1997). C, mean glutamate-mediated currents stimulated at 1 Hz. I, control; II, 1 μM 8CPT; III, 10 μM 8CPT. Calculated percentage failure: control, 80 %; 1 μM 8CPT, 45 %; 5 μM 8CPT, 41 % (not shown) and 10 μM 8CPT, 48 %. Stimulation voltage 30 V for 200 μs. D, mean data for all cells. ▪, control (n = 5); □, 1 μM 8CPT (n = 3); •, 5 μM 8CPT (n = 3). 8CPT is reported to show selectivity for A1 (Ki≈10.9 nM) over A2 receptors (Ki≈1440 nM).

By blocking adenosine receptors we have studied the effect of endogenous adenosine on the release of ATP and glutamate. Concentrations of 8CPT (5 and 10 μM), which would be expected to block both A1 and A2 adenosine receptors (Bruns, Lu & Pugsley, 1986), caused a decrease in failure rate of both purinergic EPSCs (P < 0.05) and glutamatergic EPSCs (P < 0.05; Edwards et al. 1997). In fact, at low stimulation frequencies, in the presence of 10 μM 8CPT almost no failures remained in ATP receptor-mediated synaptic transmission (8.3 ± 6.0 % at 1 Hz). As previously mentioned, the failure rate of the ATP receptor-mediated currents increased significantly with increasing stimulation frequency (P < 0.0001). In contrast, in 10 μM 8CPT (where the A1 and A2 receptors are blocked), the frequency-response curve flattened out and there was no longer a significant increase in failure rate as stimulation frequency was increased from 0.5 Hz to 10 Hz (Edwards et al. 1997).

In contrast to the higher concentration of 8CPT, which would be expected to block all adenosine receptors, the lowest concentration used (1 μM 8CPT) would be expected to be relatively specific for A1 receptors, as demonstrated above in the presence of specific agonists. This concentration had no effect on the release of ATP at any stimulation frequency tested but effectively decreased the failure rate of glutamatergic EPSCs at all frequencies (P < 0.05; Fig. 6). In fact, in the case of glutamatergic transmission, a maximal effect of blocking adenosine receptors throughout the frequency range was seen at the lowest 8CPT concentration used (10 % shift at all frequencies, 1 μM) and no further effect was seen at higher concentrations of the antagonist. Thus despite the presence of A1 receptors, which can effectively inhibit both glutamatergic and purinergic transmisson (as demonstrated by the application of the exogenous agonist CCPA), endogenous adenosine only activates the A1 receptors at glutamatergic synapses, but not the A1 receptors at purinergic synapses. In contrast endogenous adenosine effectively modulates purinergic transmission through A2-like receptors, in a manner dependent on stimulation frequency. This observation suggests that the two transmitters must be released from different terminals and that receptor distribution on these terminal is different (see Discussion). It is interesting to note that, in the presence of A1 agonists, 10 μM 8CPT did not produce a significant decrease in failure rate of purinergic EPSCs compared with control, as might be expected from the actions of 8CPT in the absence of agonist. This was not investigated further but might indicate an interaction between the A1 and A2 receptor pathways.

The effect of ectonucleotidase activity on the frequency-mediated inhibition of evoked ATP currents

To determine whether the adenosine-mediated frequency dependence of ATP release was the result of breakdown of ATP in the synaptic cleft or of co-release of adenosine and ATP, we attempted to inhibit the enzymes that break down ATP to adenosine. αβ-Methylene ADP (200 μM), which inhibits the breakdown of AMP to adenosine (Cunha, Correia-De-Sá, Sebastiao & Ribeiro, 1996), did not affect the frequency dependence of the failure rate of evoked ATP receptor-mediated synaptic currents (n = 5). The putative ectonucleotidase inhibitor ARL 67156 (Crack et al. 1995) also failed to prevent the increase in failure rate (n = 3). We also tried to saturate the breakdown enzymes with a cocktail of nucleotides (background nucleotidases show no real specificity for different nucleotides; Ziganshin, Hoyle & Burnstock, 1994). Two-hundred micromolar each of GMP, UMP and UDP had no effect on the frequency dependence of the failure rate of evoked ATP receptor-mediated synaptic currents (n = 3).

Within the limits of the available pharmacology, this suggests that the source of adenosine mediating the frequency dependence of ATP release is not breakdown of the transmitter within the cleft.

DISCUSSION

The results described above show that it is highly unlikely that glutamate and ATP are co-released from the same axon. Moreover these two types of synapse appear to be modulated by different types of adenosine receptors.

Most cells stimulated by low voltage in the medial habenula release glutamate and show no ATP receptor-mediated synaptic response. In the five cells in which ATP release was stimulated at low voltages, no postsynaptic response to glutamate was seen. Thus in these five purinergic synapses and most glutamatergic synapses tested, it is clear that ATP and glutamate are not cotransmitters. It is not impossible that both ATP and glutamate are released at the two types of synapse (Fig. 7) but that the receptors for glutamate and ATP are differentially distributed, so that only one sort of postsynaptic response occurs. In either case, however, glutamatergic and purinergic responses are the result of release from different axons.

Figure 7. Schematic diagram illustrating the anatomy of glutamate and ATP synapses in the rat medial habenula.

Figure 7

Open in a new tab

Purinergic synapse: this synapse has both A1 and A2-like receptors with distinct localization. The A2 receptors are located at or near the synapse with the A1 receptor location isolated from the synapse. Endogenous adenosine acts via the lower affinity A2 receptor. An uptake mechanism may prevent activation of the higher affinity A1 receptor by endogenous adenosine. Glutamatergic synapse: there are no presynaptic A2-like receptors and background adenosine activates A1 receptors.

Due to the difficulty of stimulating ATP release at low voltages, we investigated further the question of whether co-release may have occurred, by comparing independently the properties of ATP and glutamate release using the usual protocols for minimal stimulation for the two transmitters (described in Methods). The fact that the effects of stimulation frequency were different on glutamate and ATP release discounted the possibility that the two transmitters could be co-packaged in the same vesicles.

We have demonstrated that release of ATP at synapses in the medial habenula nucleus is strongly frequency dependent and that this frequency dependence, and in fact most background adenosine effects at this synapse are prevented by blocking adenosine A2 receptors. This effect and the dose dependency of its block by the antagonist 8CPT is also mimicked by the A2 receptor agonist DPMA but not by the A1 receptor agonist CCPA. This suggests that with release of ATP at increasing frequency, adenosine builds up in the synaptic cleft where it activates A2-like receptors causing inhibition of release. Moreover even at low stimulation frequencies adenosine causes some inhibition of release also via A2-like receptors. There is no evidence for endogenous adenosine activating A1 receptors to cause effects on ATP release under the conditions used in these experiments.

In contrast, very little frequency dependence of release was observed for glutamate, and any endogenous effects of adenosine were mediated by A1 receptors. These distinct differences in the pharmacology of endogenous adenosine-activated effects strongly suggest that ATP and glutamate are not stored within the same bouton.

The difference in pharmacology of the effects of background adenosine in the two systems may suggest a differential arrangement of the two receptor subtypes. The observation that the frequency-dependent effects are mediated by A2-like receptors and not by A1 receptors (despite the higher affinity of A1 receptors for adenosine; Bruns et al. 1986) suggests that, at ATP synapses, A2-like receptors are located within the synapse while A1 receptors seem to be located extrasynaptically (Fig. 7). In fact no endogenous effects of adenosine acting on A1 receptors were observed on ATP release, suggesting that any background adenosine coming from remote sources in the slice cannot reach the vicinity of the purinergic boutons. Moreover adenosine originating from purinergic synapses cannot reach extrasynaptic regions at sufficient concentrations to activate any A1 receptors on purinergic terminals. In contrast the effects of adenosine at glutamatergic synapses seem to be mediated via A1 receptors and, moreover, endogenous adenosine reaches these receptors. In fact, the only apparent source of endogenous adenosine at these synapses is extrasynaptic (being independent of release frequency). These differential effects of endogenous adenosine clearly demonstrate that the terminals mediating the two types of transmission are different. These observations further suggest, perhaps not surprisingly, that control of adenosine uptake at purinergic synapses may be more tightly controlled than at glutamatergic synapses. Note that exogenous A1 agonists affect both glutamate and ATP release, but endogenous adenosine has a differential effect inhibiting only glutamate release. This would be more or less incompatible with co-release from the one bouton.

The synaptic adenosine that mediates the frequency-dependent effects at purinergic synapses may come from a variety of sources. The most obvious possibility is via rapid breakdown of released ATP in the synaptic cleft by ectonucleotidases either situated within the cleft or possibly released with ATP (Todorov, Mihaylova-Todorova, Craviso, Bjur & Westfall, 1996; Todorov et al. 1997). This could result in build up of the adenosine concentration in the synaptic cleft. We have tried to determine whether ATP is rapidly degraded to adenosine in our system. Although the presence of ectoATPases has been demonstrated in the rat habenula (Sperlagh, Kittel, Lajtha & Vizi, 1995), drugs which might be expected to inhibit the effects of these enzymes (αβ-methyl ADP, ARL 67156 or a cocktail of nucleotides) had no effect on the frequency dependence of the ATP response in our system. It should be noted that the pharmacological agents available for blocking nucleotidases are limited and we have no independent means in our system of determining the level to which they are inhibited. However, as far as we can determine the results suggest that the source of synaptic adenosine may be co-release of adenosine itself with ATP.

In general the presynaptic inhibitory actions of adenosine are mediated by A1 receptors (Fredholm & Dunwiddie, 1988; Mitchell, Lupica & Dunwiddie, 1993; Yawo & Chuhma, 1993; Umemiya & Berger, 1994; Dittman & Regehr, 1996; Ambrósio et al. 1996). It is interesting to note that purinergic synapses in the medial habenula are affected by an A2-like receptor which causes inhibition of transmitter release. This observation is contrary to the general understanding that has arisen from previous studies that A2-like receptors are excitatory (Sebastiao & Ribeiro, 1992; Collis & Hourani, 1993). However, several groups have also reported inhibitory actions of A2 receptors on transmitter release (O'Regan, Simpson, Perkins & Phillis, 1992; Kirk & Richardson, 1995; Mori, Shindou, Ichimura, Nonaka & Kase, 1996; Chen & Van Den Pol, 1997).

In fact, in the light of the ectonucleotidase experiments, we cannot dismiss the possibility that these A2-like effects are mediated by ATP itself rather than adenosine. This action could occur via the putative P3 receptor. These receptors have been reported to act at presynaptic terminals in the peripheral nervous system to inhibit transmitter release and to be activated both by ATP and adenosine and blocked by methylxanthines (Shinozuka, Bjur & Westfall, 1988; Forsyth, Bjur & Westfall, 1991; Ragazzi & Chinellato, 1992; Todorov, Bjur & Westfall, 1993; Kamiji, Morita & Katayama, 1994). Unfortunately the known pharmacology of these receptors is very limited and thus we cannot distinguish whether the presynaptic effects on ATP release are indeed mediated via an A2-like receptor with inhibitory actions or rather via a putative P3 receptor. However, it is clear from these experiments that one of these two receptor types is involved in the frequency-dependent effect observed at ATP synapses but not at glutamate synapses in the rat medial habenula.

Conclusions

Direct demonstration proves, in five purinergic synapses and many glutamatergic synapses, that ATP and glutamate are not co-released. Indirect evidence, obtained from the relationship between transmitter release and stimulation frequency and the pharmacology of the presynaptic adenosine receptors affected by endogenous adenosine, confirm these direct findings. We thus conclude that ATP and glutamate are not cotransmitters but are released from separate populations of neurones in the medial habenula.

Acknowledgments

The authors would like to thank Dr Hilary Lloyd for helpful advice on adenosine pharmacology, Dr Hemai Parathasarathy and Dr Richard Evans for helpful discussion and careful reading of the manuscript and Richard Hodgkinson for technical assistance. This work was supported by The Wellcome Trust.

References

  1. Ambrósio AF, Malva JO, Carvalho AP, Carvalho CM. Modulation of Ca2+ channels by activation of adenosine A1 receptors in rat striatal glutamatergic nerve terminals. Neuroscience Letters. 1996;220:163–166. doi: 10.1016/s0304-3940(96)13252-3. 10.1016/S0304-3940(96)13252-3. [DOI] [PubMed] [Google Scholar]
  2. Bardoni R, Goldstein PA, Lee CJ, Gu JG, MacDermott AB. ATP P2X receptors mediate fast synaptic transmission in the dorsal horn of the rat spinal cord. Journal of Neuroscience. 1997;17:5297–5304. doi: 10.1523/JNEUROSCI.17-14-05297.1997. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bridges AJ, Bruns RF, Ortwine DF, Priebe SR, Szotek DL, Trivedi BK. N6-[2-(3,5-dimethoxyphenyl)-2-(2-methylphenyl)-ethyl]adenosine and its uronamide derivatives. Novel adenosine agonists with both high affinity and high selectivity for the adenosine A2 receptor. Journal of Medical Chemistry. 1988;31:1282–1285. doi: 10.1021/jm00402a004. [DOI] [PubMed] [Google Scholar]
  4. Bruns RF, Lu GH, Pugsley TA. Characterization of the A2 adenosine receptor labeled by [3H]NECA in rat striatal membranes. Molecular Pharmacology. 1986;29:331–346. [PubMed] [Google Scholar]
  5. Budd DC, Nicholls DG. Protein kinase C-mediated suppression of the presynaptic adenosine A1 receptor by a facilitatory metabotropic glutamate receptor. Journal of Neurochemistry. 1995;65:615–621. doi: 10.1046/j.1471-4159.1995.65020615.x. [DOI] [PubMed] [Google Scholar]
  6. Buell G, Collo G, Rassendren F. P2X receptors: An emerging channel family. European Journal of Neuroscience. 1996;8:2221–2228. doi: 10.1111/j.1460-9568.1996.tb00745.x. [DOI] [PubMed] [Google Scholar]
  7. Burnstock G. Purinergic mechanisms. Annals of the New York Academy of Sciences. 1990;603:1–18. doi: 10.1111/j.1749-6632.1990.tb37657.x. [DOI] [PubMed] [Google Scholar]
  8. Chen G, Van Den Pol AN. Adenosine modulation of calcium currents and presynaptic inhibition of GABA release in suprachiasmatic and arcuate nucleus neurons. Journal of Neurophysiology. 1997;77:3035–3047. doi: 10.1152/jn.1997.77.6.3035. [DOI] [PubMed] [Google Scholar]
  9. Collis MG, Hourani SMO. Adenosine receptor subtypes. Trends in Pharmacological Sciences. 1993;14:360–366. doi: 10.1016/0165-6147(93)90094-z. [DOI] [PubMed] [Google Scholar]
  10. Crack BE, Pollard CE, Beukers MW, Roberts SM, Hunt SF, Ingall AH, Mckechnie KCW, Ijzerman AP, Leff P. Pharmacological and biochemical analysis of FPL 67156, a novel, selective inhibitor of ecto-ATPase. British Journal of Pharmacology. 1995;114:475–481. doi: 10.1111/j.1476-5381.1995.tb13251.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Cunha RA, Correia-De-Sá P, Sebastiao AM, Ribeiro JA. Preferential activation of excitatory adenosine receptors at rat hippocampal and neuromuscular synapses by adenosine formed from released adenine nucleotides. British Journal of Pharmacology. 1996;119:253–260. doi: 10.1111/j.1476-5381.1996.tb15979.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Dittman JS, Regehr WG. Contributions of calcium-dependent and calcium-independent mechanisms to presynaptic inhibition at a cerebellar synapse. Journal of Neuroscience. 1996;16:1623–1633. doi: 10.1523/JNEUROSCI.16-05-01623.1996. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Edwards FA. Anatomy and electrophysiology of fast central synapses lead to a structural model for long-term potentiation. Physiological Reviews. 1995;75:759–787. doi: 10.1152/physrev.1995.75.4.759. [DOI] [PubMed] [Google Scholar]
  14. Edwards FA, Gibb AJ, Colquhoun D. ATP receptor-mediated synaptic currents in the central nervous system. Nature. 1992;359:144–147. doi: 10.1038/359144a0. 10.1038/359144a0. [DOI] [PubMed] [Google Scholar]
  15. Edwards FA, Konnerth A, Sakmann B, Takahashi T. A thin slice preparation for patch clamp recordings from neurones of the mammalian central nervous system. Pflügers Archiv. 1989;414:600–612. doi: 10.1007/BF00580998. [DOI] [PubMed] [Google Scholar]
  16. Edwards FA, Robertson SJ, Gibb AJ. Properties of ATP receptor-mediated synaptic transmission in the rat medial habenula. Neuropharmacology. 1997;36:1253–1268. doi: 10.1016/s0028-3908(97)00127-5. [DOI] [PubMed] [Google Scholar]
  17. Evans RJ, Derkach V, Surprenant A. ATP mediates fast synaptic transmission in mammalian neurons. Nature. 1992;357:503–505. doi: 10.1038/357503a0. [DOI] [PubMed] [Google Scholar]
  18. Forsyth KM, Bjur RA, Westfall DP. Nucleotide modulation of norepinephrine release from sympathetic nerves in the rat vas deferens. Journal of Pharmacology and Experimental Therapeutics. 1991;256:821–826. [PubMed] [Google Scholar]
  19. Fredholm BB, Dunwiddie TV. How does adenosine inhibit transmitter release? Trends in Pharmacological Sciences. 1988;9:130–134. doi: 10.1016/0165-6147(88)90194-0. [DOI] [PubMed] [Google Scholar]
  20. Galligan JJ, Bertrand PP. ATP mediates fast synaptic potentials in enteric neurons. Journal of Neuroscience. 1994;14:7563–7571. doi: 10.1523/JNEUROSCI.14-12-07563.1994. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Kamiji T, Morita K, Katayama Y. ATP regulates synaptic transmission by pre- and postsynaptic mechanisms in guinea-pig myenteric neurons. Neuroscience. 1994;59:165–174. doi: 10.1016/0306-4522(94)90107-4. [DOI] [PubMed] [Google Scholar]
  22. Kirk IP, Richardson PJ. Inhibition of striatal GABA release by the adenosine A2a receptor is not mediated by increases in cyclic AMP. Journal of Neurochemistry. 1995;64:2801–2809. doi: 10.1046/j.1471-4159.1995.64062801.x. [DOI] [PubMed] [Google Scholar]
  23. Klotz KN, Lohse MJ, Schwabe U, Cristalli G, Vittori S, Grifantini M. 2-Chloro-N6-[3H]cyclopentyladenosine ([3H]CCPA)—a high affinity agonist radioligand for A1 adenosine receptors. Naunyn-Schmiedeberg's Archives of Pharmacology. 1989;340:679–683. doi: 10.1007/BF00717744. [DOI] [PubMed] [Google Scholar]
  24. Lohse MJ, Klotz KN, Schwabe U, Cristalli G, Vittori S, Grifantini M. 2-Chloro-N6-cyclopentyladenosine: a highly selective agonist at A1 adenosine receptors. Naunyn- Schmiedeberg's Archives of Pharmacology. 1988;337:687–689. doi: 10.1007/BF00175797. [DOI] [PubMed] [Google Scholar]
  25. Mitchell JB, Lupica CR, Dunwiddie TV. Activity-dependent release of endogenous adenosine modulates synaptic responses in the rat hippocampus. Journal of Neuroscience. 1993;13:3439–3447. doi: 10.1523/JNEUROSCI.13-08-03439.1993. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Mori A, Shindou T, Ichimura M, Nonaka H, Kase H. The role of adenosine A2a receptors in regulating GABAergic synaptic transmission in striatal medium spiny neurons. Journal of Neuroscience. 1996;16:605–611. doi: 10.1523/JNEUROSCI.16-02-00605.1996. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. O'Regan MH, Simpson RE, Perkins LM, Phillis JW. Adenosine receptor agonists inhibit the release of γ-aminobutyric acid (GABA) from the ischemic rat cerebral cortex. Brain Research. 1992;582:22–26. doi: 10.1016/0006-8993(92)90312-w. [DOI] [PubMed] [Google Scholar]
  28. Ragazzi E, Chinellato A. Purinergic receptors, prostacyclin and atherosclerosis. Pharmacological Research. 1992;26:123–129. doi: 10.1016/s1043-6618(05)80125-2. [DOI] [PubMed] [Google Scholar]
  29. Robertson SJ, Edwards FA. ATP is not a co-transmitter with glutamate at synapses of the rat medial habenula. The Journal of Physiology. 1997;501.P:11. P. [Google Scholar]
  30. Sebastiao AM, Ribeiro JA. Evidence for the presence of excitatory A2 adenosine receptors in the rat hippocampus. Neuroscience Letters. 1992;138:41–44. doi: 10.1016/0304-3940(92)90467-l. [DOI] [PubMed] [Google Scholar]
  31. Shinozuka K, Bjur RA, Westfall DP. Characterization of prejunctional purinoceptors on adrenergic nerves of the rat caudal artery. Naunyn-Schmiedeberg's Archives of Pharmacology. 1988;338:221–227. doi: 10.1007/BF00173391. [DOI] [PubMed] [Google Scholar]
  32. Sperlagh B, Kittel A, Lajtha A, Vizi ES. ATP acts as fast neurotransmitter in rat habenula: neurochemical and enzymecytochemical evidence. Neuroscience. 1995;66:915–920. doi: 10.1016/0306-4522(94)00588-v. [DOI] [PubMed] [Google Scholar]
  33. Stevens CF. Quantal release of neurotransmitter and long-term potentiation. Neuron. 1993;10:55–63. doi: 10.1016/s0092-8674(05)80028-5. suppl. [DOI] [PubMed] [Google Scholar]
  34. Suprenant A, Buell G, North RA. P2X receptors bring new structure to ligand-gated ion channels. Trends in Neurosciences. 1995;18:224–229. doi: 10.1016/0166-2236(95)93907-f. [DOI] [PubMed] [Google Scholar]
  35. Todorov LD, Bjur RA, Westfall DP. Inhibitory and facilitatory effects of purines on transmitter release from sympathetic nerves. Journal of Pharmacology and Experimental Therapeutics. 1993;268:985–989. [PubMed] [Google Scholar]
  36. Todorov LD, Mihaylova-Todorova S, Craviso GL, Bjur RA, Westfall DP. Evidence for the differential release of the cotransmitters ATP and noradrenaline from sympathetic nerves of the guinea-pig vas deferens. The Journal of Physiology. 1996;496:731–748. doi: 10.1113/jphysiol.1996.sp021723. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Todorov LD, Mihaylova-Todorova S, Westfall TD, Sneddon P, Kennedy C, Bjur RA, Westfall DP. Neuronal release of soluble nucleotidases and their role in neurotransmitter inactivation. Nature. 1997;387:76–79. doi: 10.1038/387076a0. [DOI] [PubMed] [Google Scholar]
  38. Umemiya M, Berger AJ. Activation of adenosine A1 and A2 receptors differentially modulates calcium channels and glycinergic synaptic transmission in rat brainstem. Neuron. 1994;13:1439–1446. doi: 10.1016/0896-6273(94)90429-4. [DOI] [PubMed] [Google Scholar]
  39. White TE, Dickenson JM, Alexander SPH, Hill SJ. Adenosine A1-receptor stimulation of inositol phospholipid hydrolysis and calcium mobilisation in DDT1 MF-2 cells. British Journal of Pharmacology. 1992;106:215–221. doi: 10.1111/j.1476-5381.1992.tb14317.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Wu L-G, Saggau P. Adenosine inhibits evoked synaptic transmission primarily by reducing presynaptic calcium influx in area CA1 of hippocampus. Neuron. 1994;12:1139–1148. doi: 10.1016/0896-6273(94)90321-2. [DOI] [PubMed] [Google Scholar]
  41. Yawo H, Chuhma N. Preferential inhibition of omega-conotoxin-sensitive presynaptic Ca2+ channels by adenosine autoreceptors. Nature. 1993;365:256–258. doi: 10.1038/365256a0. [DOI] [PubMed] [Google Scholar]
  42. Ziganshin AU, Hoyle CHV, Burnstock G. Ecto-enzymes and metabolism of extracellular ATP. Drug Development Research. 1994;32:134–146. [Google Scholar]

Articles from The Journal of Physiology are provided here courtesy of The Physiological Society

RESOURCES