High-resolution real-time recording with microelectrode biosensors reveals novel aspects of adenosine release during hypoxia in rat hippocampal slices

Authors


Address correspondence and reprint requests to Professor Nicholas Dale, School of Biological Sciences, University of Warwick, Coventry, CV4 7AL, UK. E-mail: N.E.Dale@warwick.ac.uk

Abstract

We have used improved miniaturized adenosine biosensors to measure adenosine release during hypoxia from within the CA1 region of rat hippocampal slices. These microelectrode biosensors record from the extracellular space in the vicinity of active synapses as they detect the synaptic field potentials evoked in area CA1 by stimulation of the afferent Schaffer collateral-commissural fibre pathway. Our new measurements demonstrate the rapid production of adenosine during hypoxia that precedes and accompanies depression of excitatory transmission within area CA1. Simultaneous measurement of adenosine release and synaptic transmission gives an estimated IC50 for adenosine on transmission in the low micromolar range. However, on reoxygenation, synaptic transmission recovers in the face of elevated extracellular adenosine and despite a post-hypoxic surge of adenosine release. This may indicate the occurrence of apparent adenosine A1 receptor desensitization during metabolic stress. In addition, adenosine release is unaffected by pharmacological blockade of glutamate receptors and shows depletion on repeated exposure to hypoxia. Our results thus suggest that adenosine release is not a consequence of excitotoxic glutamate release. The potential for adenosine A1 receptor desensitization during metabolic stress implies that its prevention may be beneficial in extending adenosine-mediated neuroprotection in a variety of clinically relevant conditions.

Abbreviations used
aCSF

artificial cerebrospinal fluid

fEPSC

field excitatory postsynaptic current

fEPSP

field excitatory postsynaptic potential

PPE

post-hypoxic purine efflux

The release of adenosine during metabolic or traumatic insults to the mammalian central nervous system is an important neuroprotective mechanism by virtue of adenosine's ability to inhibit glutamate release via inhibitory presynaptic adenosine A1 receptors (de Mendonça et al. 2000). Pharmacological manipulations that either increase extracellular adenosine or the activation of A1 receptors generally reduce neuropathology, whereas antagonism or knockout of A1 receptors or reducing extracellular adenosine by enzymic degradation exacerbates the effects of traumatic or metabolic stress (de Mendonça et al. 2000; Johansson et al. 2001).

However, despite the protective role of A1 receptors, two additional subtypes of the four G-protein coupled adenosine receptors, the A2A and A3 receptors, promote glutamate release, most likely via desensitization of A1 receptor-mediated inhibition (Dunwiddie et al. 1997; Lopes et al. 2002). Indeed, adult A2A knockout mice (Chen et al. 1999), but not neonatal mice (Aden et al. 2003), show reduced brain damage after focal cerebral ischaemia and A3 antagonists have variously been shown to promote neuroprotection when given acutely (Von Lubitz et al. 2001). These observations suggest a complex interplay between different adenosine receptor subtypes, the outcome from which may depend upon developmental age, the concentration of extracellular adenosine and the relative abundance of each of these receptors.

We have pioneered the use of biosensors to study release of adenosine during hypoxia (Dale et al. 2000; Pearson et al. 2001). Our previous results have shown the production of adenosine coincident with the depression of excitatory glutamatergic synaptic transmission in stratum radiatum of area CA1 of the rat hippocampus (Dale et al. 2000) and demonstrated the unexpected and important observation that adenosine release is reduced in response to repeated periods of hypoxia (Pearson et al. 2001). This earlier work relied upon the ‘Mk-1’ adenosine sensor, which although a major advance in terms of spatial and temporal localization of adenosine production, was too large to insert into tissue, and was thus used to measure from the surface of the slice. Furthermore, being on the surface of the slice, this sensor was rather slow in responding and with a diameter of 500 µm was capable of only limited spatial resolution. This gave rise to the criticism raised by Latini and Pedata (2001), but addressed extensively in the original description of the use of the sensor in hippocampus (Dale et al. 2000), that the bulky sensor may not adequately report the changes of adenosine levels close to their site of action at synapses within the slice. We now report the use of a ‘Mk-2’ sensor (Llaudet et al. 2003) that circumvents many of these drawbacks. The new sensor is far smaller than the Mk-1, much quicker responding and can be inserted within the hippocampal slice to measure adenosine levels from the interior of the slice.

Our refined methods give unprecedented real time measurement of adenosine release during hypoxia, confirm our previous conclusions, give a more accurate representation of the dynamics and concentration dependence of the actions of adenosine and for the first time provide evidence consistent with apparent desensitization of A1 receptor-mediated inhibition of glutamate release during metabolic stress.

Methods

Slice preparation

Sprague–Dawley rats of either sex, aged 10–27 days, were killed by cervical dislocation in accordance with Schedule 1 of the UK Government Animals (Scientific Procedures) Act 1986. The brain was rapidly removed and placed in ice-cold artificial cerebrospinal fluid (aCSF) containing 11 mm Mg2+ wherein 400 µm transverse hippocampal slices were cut with a Vibratome or Leica VT1000S as previously described (Dale et al. 2000). Slices were placed in an incubation chamber comprising a nylon mesh within a beaker of continuously circulating, oxygenated (95% O2/5% CO2) standard aCSF (1 mm Mg2+) and kept at room temperature for at least 1 h before use. The composition of the aCSF solution was (in mm): NaCl, 124; KCl, 3; CaCl2, 2; NaHCO3, 26; NaH2PO4, 1.25; d-glucose, 10; MgSO4, 1; pH 7.4 with 95% O2/5% CO2.

Extracellular recording

A single slice was transferred to a recording chamber, fully submerged in aCSF and perfused at 6 mL/min (33–34°C). A twisted Teflon-coated tungsten (Advent Research Materials, Eynsham, Oxon, UK) bipolar stimulating electrode (∼100 µm in diameter) positioned in the stratum radiatum was used to stimulate the Schaffer collateral commissural pathway at 15 s intervals. The stimulus intensity (approximately 30–50% of maximum) was subthreshold for population spike activation. Extracellular recordings of the evoked field excitatory postsynaptic potentials (fEPSPs) were made from stratum radiatum with an aCSF-filled glass microelectrode. Electrical signals were acquired at 10 kHz, filtered at 1 Hz to 3 kHz and recorded to a Pentium computer using ‘LTP’ data acquisition and analysis software (courtesy of Dr Bill Anderson and Professor Graham Collingridge, Bristol University, UK (http://www.ltp-program.com; Anderson and Collingridge 2001).

Induction of hypoxia

In all experiments, hypoxia was induced by the substitution of standard aCSF with identical aCSF pre-equilibrated with 95% N2/5% CO2 as previously described (Dale et al. 2000). Hypoxic episodes lasted for 5–10 min and up to two such episodes were given to any one slice.

Adenosine sensors

Sensors were fabricated following the methods of Llaudet et al. (2003) or obtained from Sarissa Biomedical Ltd (Coventry, UK). In brief the Mk-2 adenosine sensor relied on entrapping a three enzyme cascade [adenosine deaminase, AD (EC 3.5.4.4); nucleoside phosphorylase, NP (EC 2.4.2.1) and xanthine oxidase, XO (EC 1.1.3.22)] within a matrix that was deposited around the working electrode which consisted of either pure Pt or Pt/Ir (90/10) wire etched to a final diameter that ranged from 25 to 50 µm. The sensor had an exposed length of around 400 µm to 2 mm that was coated with enzymes and thus capable of detecting purines. Three types of sensor were used in the present study to identify the nature of the released substance. First, null sensors, possessing only the polymeric coating but no enzymes, were used as a control to check whether any non-specific electroactive interferents were released from the slice during hypoxia and could confound the measurements. Secondly, sensors containing just NP and XO and sensors with all three enzymes (AD, NP and XO) were used. The difference signal between these two types of sensors gave the specific adenosine signal. The sensors were either laid on the surface of the slice (indenting it slightly) or inserted along their longitudinal axis at an angle of ∼70° through the entire depth of stratum radiatum in area CA1 of the hippocampal slice. For the sensors that were inserted, the polymer/enzyme coating was restricted to around 400 µm in length such that the entire coating was within the slice. Careful matching of the sizes and sensitivities of the two sensor types as well as their positioning and insertion into the slice was the key to obtaining large signals and reliable differential recordings.

Sensors were calibrated with known concentrations (usually 10 µm) of adenosine and inosine applied for 2–3 min. To avoid complications associated with the uptake and metabolism of exogenous adenosine and inosine in the hippocampal slice, calibrations were performed with the sensors placed a few mm above the surface of the slice at the beginning and end of experiments, following their removal from the slice. Adenosine calibration allowed quantification of adenosine signals obtained from the slice and determination of any rundown in sensor sensitivity over the lifetime of an experiment. Only sensors containing the full complement of enzymes responded to adenosine (see Fig. 2d). The application of inosine, allowed any difference in sensitivity of sensors to be compensated for, especially important when differential recordings were made between sensors. The response of the sensor to exogenous adenosine and inosine was rapid and symmetrical [10–90% response time of a few seconds; (Fig. 2d; Dale et al. 2002; Llaudet et al. 2003)]. Thus, the polymer matrix does not pose a diffusional barrier to purines or their reaction products and any persistence of signals evoked by physiological or pathological conditions from intact tissue will not reflect trapping of substrates or metabolites.

Figure 2.

Measurement of purine production during hypoxia from within stratum radiatum of area CA1 in hippocampal slices. (a) Representative synaptic waveforms as measured with the sensor (top) and conventional glass microelectrode (bottom). Both the sensor and microelectrode were inserted into stratum radiatum. Inset shows sensor fEPSC (grey) scaled, inverted and superimposed on microelectrode fEPSP. The short delay between the two peaks probably reflects differences in recording location. (b) Detection of extracellular purine release during hypoxia (black bar) via a sensor placed within stratum radiatum of a hippocampal slice (top) and (bottom) simultaneous recordings of fEPSPs. Note rise in signal from sensor coincident with depression of synaptic transmission (periodic deflections on both traces). Shortly after reoxygenation and a delay consistent with the dead space in the system, the purine signal displays a sharp inflection (arrowhead), which we have termed the post-hypoxic purine efflux (PPE). Notice that during this time the fEPSP/C (periodic deflection on traces) begins to recover. (c) The hypoxic purine signal depends upon the presence of the metabolizing enzymes within the sensor. Top trace shows combined adenosine and inosine signal in response to hypoxia (black bar). Note PPE (arrowhead) on reoxygenation. Bottom trace, simultaneously recorded signal from a ‘null’ sensor lacking enzymes also inserted within stratum radiatum. Note lack of appreciable signal and the PPE on the null sensor indicating the absence of potentially confounding non-specific electroactive species. Periodic deflections on traces reflect simultaneously recorded fEPSPs showing reversible depression of synaptic transmission during hypoxia. The slow oscillations on the sensor traces are due to periodic changes in bath level resulting from slight imbalance in the inflow and outflow. (d) Responses of an adenosine sensor placed in the bath a few mm above a slice to adenosine (Ado) and inosine (Ino; both 10 µm). Note the complete lack of any response to hypoxia (black bar). Note also the symmetrical calibration response to exogenous adenosine and inosine indicating no trapping of adenosine or metabolites within the polymer matrix.

Chemicals

Salts used in the aCSF were obtained from Sigma-Aldrich or Fischer. Adenosine, inosine, kynurenic acid, adenosine deaminase (AD; EC 3.5.4.4), nucleoside phosphorylase (NP; EC 2.4.2.1) and xanthine oxidase (XO; EC 1.1.3.22) were obtained from Sigma-Aldrich. Coformycin was purchased from Calbiochem.

Statistics

Data are expressed as means ± SEM with n = number of slices. Statistical comparisons of data sets were performed either using the paired t-test or the Wilcoxon matched pairs signed ranks test.

Results

Single sensor measurements

The new smaller sensor gives two possible methods for measuring purine release: either by laying it along the surface, or inserting the sensor into the slice itself. The former has the advantage that a larger surface area of sensor can be placed in contact with the tissue, potentially giving a larger signal. This comes at the expense of spatial localization and also measurement of adenosine only at the slice surface. The latter necessarily gives a restricted amount of contact (effectively the thickness of the slice) but has the advantage that the sensor is in more intimate contact with the tissue and is probably closer to the sites of adenosine release.

Purine measurements from the surface of the slice

Surface measurements, in which 1–2 mm of the sensor was in intimate contact with the slice, gave records very similar to those we have previously reported (Fig. 1). Following removal of O2, the sensor current firstly exhibited a negative shift due to its inherent O2-sensitivity (cf. Dale et al. 2000), and then started to rise. The increase in sensor current accompanied the depression of synaptic transmission as indicated by the reduction in the size of the fEPSP. The sensor current reached a value of 824 ± 153 pA (n = 10) after 5 min of hypoxia. Following the hypoxic period both the sensor current and synaptic transmission recovered back to baseline levels.

Figure 1.

Measurement of purine production during hypoxia from the surface of the hippocampal slice at stratum radiatum of area CA1. (a) Representative fEPSP/Cs as measured with the sensor (top) and conventional glass microelectrode (bottom). The sensor and microelectrode were placed on and inserted into stratum radiatum, respectively. Inset shows sensor fEPSC (grey) scaled, inverted and superimposed on microelectrode fEPSP. The short delay between the two peaks likely reflects differences in recording location. (b) Detection of extracellular purine release during hypoxia (black bar) via a sensor placed on the surface of the hippocampal slice (top) and (bottom) simultaneous recordings of fEPSPs. Note the initial negative shift on the sensor record indicating the time of O2 removal, followed by a rise in signal from sensor coincident with depression of synaptic transmission (periodic deflections on both traces). On reoxygenation, the sensor signal reverses and fEPSP/C recovers rapidly.

Interestingly the sensor on the surface was able to pick up a small field potential that was an inverted and scaled version of the fEPSP recorded by the microelectrode (inset Fig. 1a). The reason that an inverted version of the fEPSP was recorded by the sensor is probably that the sensors measure current at a constant imposed voltage (in our case +500 mV), whereas conventional extracellular field recordings measure changes in voltage in the vicinity of the electrode. The sensor, on the surface of the slice, must therefore be sufficiently close to active synapses to record the current flow that underlies the voltage changes accompanying the fEPSP. We shall thus refer to the sensor field recordings as field excitatory postsynaptic currents (fEPSCs).

Our observations of fEPSCs on the sensor records demonstrates that firstly, even on the surface of the slice the sensor can record synaptic transmission and secondly, that the neuronal tissue adjacent to the sensor was viable and capable of supporting synaptic transmission.

Purine measurements from within the interior of the slice

When sensors (25–50 µm in diameter) were inserted through the thickness of the slice (400 µm), two main differences were seen compared with surface measurements. First, fEPSCs recorded with sensors inserted into the slice were very much larger than those recorded from the surface of the slice (Fig. 2). This indicates that neither the physical damage caused by insertion of the sensors into the slice, nor the proximity of neuronal tissue to a local potential (+ 500 mV) was sufficient to adversely affect neuronal function. Secondly a distinct increase in the rate of purine release was seen on return to normoxia, which we term the post-hypoxic purine efflux (PPE). The PPE was always seen when the sensor was inserted into the slice. However, the PPE never seen when the sensor was merely laid on the surface (Fig. 1), when null sensors were inserted into the slice (Fig. 2c), or when an adenosine sensor a few mm above the slice was subjected to an episode of hypoxia (Fig. 2d). The longer diffusional paths from the release sites to the sensor inherent in surface recordings may have prevented us from seeing the PPE with this recording method. As this signal was not seen with null sensors, we conclude that it is due to release of purines. Indeed it was evident on the differential records (Fig. 4) suggesting that at least some of the PPE is due to release of adenosine.

Figure 4.

Differential recording of adenosine release from within the hippocampal slice during hypoxia. (a) The top two superimposed traces represent the signals obtained from the sensor containing the full complement of metabolic enzymes (adenosine + inosine) and from the sensor lacking adenosine deaminase (inosine) during a 5-min period of hypoxia (black bar). Note the post-hypoxic purine efflux (PPE; arrowhead and vertical dotted line), which occurs on reoxygenation and which is more prominent on the full sensor (top). The lower trace shows the difference signal (adenosine only) obtained by subtracting the upper two traces and yielding the profile of adenosine production during hypoxia, as measured from the interior of the hippocampal slice. (b) The adenosine deaminase inhibitor coformycin (5 µm) effectively abolishes the full sensor response to adenosine released during hypoxia (black bar). Arrowhead indicates the occurrence of the PPE on the adenosine trace, indicating that a large part of the PPE is due to a surge of adenosine release.

Relationship between purine release and the inhibition of synaptic transmission during hypoxia

Casual inspection of the sensor records shows a clear correlation between the rise of the sensor signal and inhibition of the fEPSP. However although the fEPSP recovers after hypoxia, it appears to do so faster than the sensor current. This is especially true when the PPE is evident. We therefore plotted inhibition of the fEPSP during the onset of and recovery from hypoxia against sensor current (Fig. 3). This relationship describes a loop (inset Fig. 3) typical of hysteresis between the onset and offset of hypoxia. In other words the concentration-response relation between purine release and inhibition of synaptic transmission is shifted to the right between the onset and offset of hypoxia. In single electrode measurements this could be at least partially explained by the late persistence of inosine, the breakdown product of adenosine (which will also be detected by the sensor). However differential recordings, which are selective for adenosine (see below), also demonstrate hysteresis suggesting an apparent shift in the affinity of the receptor for adenosine.

Figure 3.

Hysteresis in the relationship between purine release, measured with a single sensor within the slice, and the post-hypoxic recovery of synaptic transmission. ▪, the inhibition of the fEPSP during hypoxia (grey box). Each data point represents the amplitude of the evoked fEPSP at 15-s intervals. ○, the sensor current during the same period of hypoxia. Note the rapid depression of the fEPSP (inset fEPSP, b) at very low sensor currents, reflecting potent inhibition of transmission by the released purines, yet the rapid recovery of the fEPSP in the presence of high levels of extracellular purines (inset fEPSP, c). The arrowhead indicates the occurrence of the post-hypoxic purine efflux shortly after reoxygenation. The top inset plots inhibition of fEPSP versus sensor current. Time flows in the direction of the arrows. Note the graph describes a loop, indicative of hysteresis between sensor current, and hence released purines, and the depression and recovery of excitatory synaptic transmission. The lower inset shows the average of four consecutive fEPSPs taken at the three times indicated (a, b, c), at different levels of sensor current, during and after the period of hypoxia.

Differential measurements

A single sensor inserted into the slice will measure adenosine and all other down-stream metabolites plus any non-specific electroactive interferents. Our null sensor measurements suggest that during hypoxia there is little production of electroactive species, which can confound our measurements. The mean current recorded by null sensors after 5 min of hypoxia was 46 ± 32 pA (n = 13; Fig. 2c). Similarly, sensors containing the full complement of enzymes, placed in the perfusate, but not within the slice, recorded no appreciable signal during hypoxia (Figs 2d, n = 3).

Adenosine will be broken down to inosine by endogenous adenosine deaminase. We have therefore performed differential measurements (see Methods) to characterize the specific adenosine component of the signal. Sensors sensitive to adenosine + inosine were calibrated to both purines (e.g. Fig. 2d), whereas sensors sensitive to only inosine were calibrated with inosine. These calibrations were then used to take into account the different sensitivities of the two sensors to inosine. The differential signal was obtained by subtracting a weighted value of the inosine sensor from the adenosine + inosine sensor.

Figure 4(a) shows an example of such a differential recording. The signal from the two sensors is similar. However that of the adenosine + inosine sensor is larger by virtue of its additional sensitivity to adenosine. The difference between the two signals thus gives specific measurement of adenosine. In 16 slices the mean level of adenosine produced after 5 min of hypoxia was 3.0 ± 0.7 µm. To test further the specificity of the differential signal for adenosine, we used the adenosine deaminase inhibitor, coformycin. Coformycin (5 µm) essentially abolished the differential signal, including the PPE, recorded by the sensors (Fig. 4b, mean reduction 100 ± 0.4%, n = 4). Thus the vast majority of the signal detected by the sensor in differential mode is adenosine.

Detailed examination of the sensor signals (Figs 4 and 5) revealed firstly that the PPE was evident on the adenosine-specific differential record and secondly that the adenosine levels rose and reached an earlier peak value than those for inosine. The adenosine levels also declined more quickly to baseline than the inosine signals. This difference in relative time course is consistent with the notion of adenosine release followed by conversion to inosine. Furthermore by comparing the weighted signals from the two sensors, we found that 50 ± 13% (n = 5) of the signal recorded by an adenosine + inosine sensor was due to the presence of adenosine. We have also examined the effect of coformycin on the absolute signal recorded by the sensors. As might be expected, coformycin greatly reduced the current on both the adenosine + inosine and inosine sensors by 81 ± 6% and 73 ± 3% (n = 4), respectively. The very high proportion of the inosine signal that is abolished by coformycin suggests that production through the breakdown of adenosine rather than direct release of inosine is the predominant pathway for the generation of extracellular inosine during hypoxia. Furthermore the approximate parity in the contribution of inosine and adenosine to the mixed (non-subtracted) signal suggests that on average every adenosine molecule is converted to inosine before being taken up.

Figure 5.

Differences in time course of hypoxic adenosine release with respect to that of inosine. The sensor signals have been individually normalized to go from 0 to 100%. (a) Average (black line; n = 4) and SEM (grey lines) of sensor signals from full sensor (adenosine + inosine) during hypoxia (black bar). (b) Average (black line; n = 4) and SEM (grey lines) of sensor signals from inosine sensor (lacks adenosine deaminase). (c) Superimposition of averaged full sensor signal (adenosine + inosine; grey line) and inosine sensor signal (black line). Note how the adenosine + inosine signal (i) rises earlier; (ii) shows pronounced post-hypoxic purine efflux (arrowhead); (iii) peaks earlier (vertical dotted line); and (iv) decays faster than the inosine signal.

As with the single sensor experiments, the differential recording of adenosine release within the slice demonstrates that synaptic transmission recovers more quickly from the adenosine-mediated inhibition than might be expected from the return of adenosine levels back to baseline (Fig. 6). When the inhibition of transmission was plotted against the change in adenosine concentration, in every case the relationship described a loop, indicating hysteresis of the response between the onset and offset of hypoxia. To quantify this we estimated the IC50 for adenosine versus synaptic inhibition during the onset and recovery of hypoxia from graphs like the inset in Fig. 6. During the onset of hypoxia the IC50 was 1.4 ± 0.7 µm, while during recovery this shifted to 7.6 ± 3.9 µm (n = 9, p < 0.02, Wilcoxon matched pairs signed ranks test).

Figure 6.

Hysteresis between the concentration of extracellular adenosine and the depression and recovery of excitatory synaptic transmission. Similar experiment to that described in Fig. 3 in which inhibition of synaptic transmission (▪) during hypoxia (grey box) and post-hypoxic recovery is compared with adenosine production measured from within the slice (○). Note full depression of fEPSP (inset fEPSP, b) at low concentrations of extracellular adenosine (e.g. 10 µm), whereas full recovery (inset fEPSP, c) is observed in the face of higher adenosine concentrations (e.g. 20–30 µm). The top inset plots inhibition of fEPSP versus adenosine concentration. Arrows depict direction of time during the experiment. Note shift in IC50 (dotted lines) from 7 µm during the depression phase to 38 µm during recovery. The lower inset shows the average of four consecutive fEPSPs taken at the three times indicated (a, b, c), at different levels of extracellular adenosine before, during and after the period of hypoxia, respectively. Arrowhead indicates the occurrence of the post-hypoxic purine efflux during which time the fEPSP paradoxically recovers.

Depletion of adenosine release

In our previous studies (Pearson et al. 2001), we reached a major and novel conclusion that adenosine release during hypoxia is depletable by prior exposure to hypoxia. Given that our new recordings demonstrate superior temporal and spatial resolution from within the slice, we decided to repeat this finding. Figure 7(a) illustrates an example of differential recordings of adenosine release from two consecutive hypoxic episodes. Note how less adenosine emerges during the second episode. On average we found that a 5-min episode of hypoxia gave a reduction in the amount of adenosine released in the second episode of 46 ± 9.4% (n = 9). This is almost identical to the value reported in a study using the Mk-1 sensor (Pearson et al. 2001).

Figure 7.

Sequential exposure to hypoxia results in reduced adenosine release but hypoxic purine release is not dependent on the activation of ionotropic glutamate receptors. (a) Adenosine release (within slice, differential measurements) during the first and second of two sequential 5-min periods of hypoxia (black bar) separated by 28 min. The arrowhead indicates the occurrence of the PPE. Note reduction in extracellular adenosine during second hypoxic episode compared to the first. (b) Adenosine and inosine release (within slice, single-ended measurements) during two sequential periods of hypoxia separated by 48 min, in which the second period of hypoxia was administered in the presence of 5 mm kynurenic acid to block ionotropic glutamate receptors as shown by absence of fEPSCs on sensor trace. Note reduction in extracellular adenosine during second hypoxic episode, which is no different from the reduction seen in the absence of kynurenic acid (Fig. 7a).

Glutamate receptor dependence

As hypoxia is known to cause release of glutamate, and activation of glutamate receptors can cause release of adenosine (Latini and Pedata 2001), we tested whether adenosine and inosine release during hypoxia depended upon activation of glutamate receptors by using the broad-spectrum antagonist kynurenic acid (2–5 mm). Slices were subjected to an initial hypoxic episode in control aCSF and adenosine and inosine release recorded. Kynurenic acid was then applied for at least 10 min, which inhibited synaptic transmission by 95 ± 1% (n = 8, see absence of fEPSCs on sensor trace in Fig. 7b). A second hypoxic episode was then delivered. We found that in the presence of kynurenic acid, adenosine and inosine were still released during hypoxia (Fig. 7b). The amount of purine release was less than during the first hypoxic episode (37 ± 9% reduction of sensor current, n = 8), but this is almost certainly due to the depletion of release between consecutive episodes that we have already demonstrated. Indeed in interleaved control experiments performed without the addition of kynurenic acid after the first episode, the reduction of the sensor current recorded during the second hypoxic episode was 31 ± 5% (n = 8) and was not significantly different from the reduction occurring in the presence of kynurenic acid. We therefore conclude that kynurenic acid had no effect on adenosine release and that release of purines during hypoxia is mechanistically independent of the release of glutamate, a conclusion in agreement with an earlier HPLC study (Pedata et al. 1993).

Discussion

Comparison with previous direct measurements of adenosine release

We have previously investigated the release of adenosine during hypoxia using the first generation adenosine sensor, which was placed on the surface of the hippocampal slice (Pearson et al. 2001). At the time, the Mk-1 sensor offered unprecedented spatial and temporal resolution, but suffered from the limitations of size (500 µm across) and that it detected adenosine from the surface of the hippocampal slice. These limitations, as discussed extensively by Dale et al. (2000), could potentially have distorted the temporal profile and magnitude of adenosine release during hypoxia. In addition, the possibility has been raised that surface measurements might be altered by a layer of dead tissue on the surface of the slice (Latini and Pedata 2001).

Early studies quoted by Latini and Pedata (2001) gave an estimate of 50–100 µm of dead tissue at the cut surfaces of slices from adult rat cerebellum (Garthwaite et al. 1979; Professor John Garthwaite, personal communication). However improvements in slice cutting methodology in the subsequent two decades demonstrably yield slices with viable superficial neurones. To confirm this we sent hippocampal slices (as prepared for this study) for histological analysis (courtesy of Mr David Goodwin and Professor John Garthwaite, UCL, London, UK). These slices possessed a much shallower layer of damage of approximately 35 µm, essentially one to two CA1 pyramidal cell bodies deep. Furthermore, the similarity between our previous findings with the Mk-1 sensor on the surface of the slices and the present results with the Mk-2 inserted into the slice confirms the validity of our previous conclusions. The subtle differences between surface and interior measurements likely reflect the difference in proximity to the sites of adenosine release rather than the influence of the thin layer of dead surface tissue.

Thus, the close association between adenosine release and the depression of excitatory synaptic transmission (Dale et al. 2000), the proportional relationship between duration of metabolic stress and amount of adenosine release (Dale et al. 2000) and the reduced adenosine release in response to repeated hypoxia (Pearson et al. 2001) have all been confirmed with measurements of adenosine release within the slice. However, the greater temporal and spatial resolution offered by embedding the degradative enzymes within a polymer matrix around a thin platinum wire has allowed us to make novel observations concerning the release of adenosine during metabolic stress, most notably the post-hypoxic purine efflux (PPE).

Post-hypoxic purine efflux

An unexpected observation made both in single-ended and differential recording mode was that the return to normoxia was always associated with a dramatic increase in adenosine production. This was only seen, however, with sensors inserted into the slice and not if the sensor was placed on the surface of the slice. This may reflect that this phenomenon is very local in nature and when the recordings are made on the surface of the slice, the temporal profile of the abrupt rise is distorted due to diffusional barriers. Clearly, since the post-hypoxic signal was not seen on the null sensor, it cannot reflect release of non-specific electroactive substances such as ascorbate or hydrogen peroxide. Furthermore, although seen on the inosine sensor, from where it might have arisen from adenosine metabolism, the PPE was still seen when the inosine signal had been subtracted to yield an adenosine-specific, coformycin-sensitive signal. This allows us to conclude that the signal reflects a surge of adenosine release.

The mechanism underlying the abrupt post-hypoxic release of adenosine is unclear. It does not reflect hypoxia-induced gross cellular lysis as transmission recovered fully, even that measured locally via the sensor. More likely the PPE results from an event initiated by the re-introduction of oxygen. Such events that could lead to a surge of adenosine production might include post-ischaemic release of nitric oxide (Kojima et al. 2001), changes in intracellular pH (Sheldon and Church 2002), and changes in the activity of adenosine transporters or enzymes involved in adenosine production or metabolism. The functional significance of this event remains to be established, although a surge of adenosine release at a time when reoxygenation may otherwise promote metabolism, neuronal activity, glutamate release and possibly intracellular free-radical production, may extend the period of synaptic depression and the neuroprotective actions of adenosine. Interestingly, Latini et al. (1999) have used indirect methods (competitive antagonism of A1 receptors) to estimate changes in extracellular adenosine during and after the induction of ischaemia in hippocampal slices. Close inspection of their Fig. 7 reveals a time-course of inferred extracellular adenosine remarkably similar to our records, including a surge of adenosine release reminiscent of the PPE on return to control aCSF.

Relationship between adenosine and inosine production during hypoxia

Our data give new information about the relationship between adenosine and inosine production during hypoxia. Both are clearly produced and the question arises as to what extent inosine results from the extracellular breakdown of previously released adenosine, or is itself released directly. We have provided several lines of evidence which indicate that ecto-adenosine deaminase plays an important role in curtailing the availability of extracellular adenosine. First, inosine is produced with a delay compared to adenosine, suggesting that it is the breakdown product of adenosine. Second, the normalized adenosine + inosine and inosine records show that relatively less inosine is produced during hypoxia and relatively more in the period following hypoxia – compare the sizes of the PPEs and the initial trajectory of the sensor currents in the normalized records (Fig. 5). Third, the application of coformycin greatly reduces the signal recorded by the inosine sensors (and eliminates the differential adenosine signal). This suggests that most, but perhaps not all, of the inosine production during hypoxia depends upon ecto-adenosine deaminase. The role that ecto-adenosine deaminase plays in regulating the actions of extracellular adenosine is discussed extensively by Latini and Pedata (2001).

The prominent extracellular production of inosine raises the possibility that it could have a physiological role. Interestingly, inosine has been described as a trophic agent (Petrausch et al. 2000) and as an agonist of adenosine A3 receptors (Tilley et al. 2000). Production of inosine from adenosine may therefore help to prime regenerative processes following pathophysiological insults such as hypoxia or ischaemia (Rathbone et al. 1999). The agonist action of inosine at A3 receptors could also have several downstream consequences for physiological function (see below).

Hysteresis between adenosine release and the depression of synaptic transmission

The depression of synaptic transmission during the onset of hypoxia scales with the release of adenosine (Dale et al. 2000). Correspondingly, the hypoxic depression of synaptic transmission is greatly attenuated by adenosine A1 receptor antagonists (Fowler 1989; Pearson and Frenguelli 2000; Gervitz et al. 2001; Pearson et al. 2001) and in A1 receptor knockout mice (Johansson et al. 2001). Our differential adenosine measurements record an almost pure adenosine signal (as shown by the virtually complete blockade of this signal by coformycin). This selectivity together with the improved temporal resolution of the microelectrode biosensors, in conjunction with simultaneous recordings of excitatory synaptic transmission, has revealed that the decay of the adenosine signal and the recovery of the fEPSP do not occur in parallel. Instead, we have observed that synaptic transmission recovers at a rate faster than the decay in the concentration of extracellular adenosine would predict. This surprising result requires consideration.

The rapid depression of synaptic transmission during the onset of hypoxia may reflect additional non-adenosine-dependent mechanisms, which distort the apparent role of adenosine in the rapidity of the initial depression leading to a leftward shift in the concentration–inhibition relationship. Support for this hypothesis comes from the observation that both muscarinic M2 (Coelho et al. 2000) and GIII metabotropic glutamate receptors (de Mendonça and Ribeiro 1997) contribute to the hypoxic depression of excitatory synaptic transmission in area CA1. However the magnitude of the adenosine-independent contribution to synaptic inhibition is small compared to that of adenosine and the rate of depression in the presence of an adenosine A1 receptor antagonist such as DPCPX to block the contribution of adenosine is very much slower than in the control (Latini et al. 1999; Pearson et al. 2001). We therefore think it unlikely that the non-adenosine-dependent mechanisms could sufficiently speed the onset of synaptic depression to account for our data. Alternatively, an adenosine-dependent mechanism can be advanced, namely that the incongruous recovery of synaptic transmission in the face of elevated extracellular adenosine reflects desensitization of A1 receptors by either adenosine A2A (Lopes et al. 2002) or A3 (Dunwiddie et al. 1997) receptors activated by the high concentrations of extracellular adenosine released during hypoxia. In addition, as inosine is a potent agonist at adenosine A3 receptors (Tilley et al. 2000; Fredholm et al. 2001), the high concentrations of inosine released from the slice during hypoxia would be expected to activate the A3 receptor, further promoting desensitization of A1 receptors.

Our use of microelectrode biosensors has revealed several novel aspects of adenosine release during metabolic stress in the mammalian CNS. By combining these measurements with simultaneous recording of physiological activity, we have explored the relationship between adenosine production and excitatory synaptic transmission in greater detail. Such a combination allows questions of the spatial and temporal characteristics of adenosine release to be asked and answered with unprecedented resolution and will help to resolve many questions which as yet remain unanswered regarding the production and release of adenosine (and down-stream metabolites) in the mammalian brain and its physiological implications.

Acknowledgements

We are grateful to the Cunningham Trust, the Wellcome Trust, Tenovus Tayside and the Anonymous Trust for financial support and to Mr David Goodwin and Professor John Garthwaite (UCL, London, UK) for detailed histological analysis of our hippocampal slices.

Ancillary