Corresponding author N. Offenhauser: University of Copenhagen, Department of Medical Physiology, The Panum Institute 12.5, Blegdamsvej 3, DK-2200 Copenhagen N, Denmark. Email: firstname.lastname@example.org
Functional neuroimaging relies on the robust coupling between neuronal activity, metabolism and cerebral blood flow (CBF), but the physiological basis of the neuroimaging signals is still poorly understood. We examined the mechanisms of activity-dependent changes in tissue oxygenation in relation to variations in CBF responses and postsynaptic activity in rat cerebellar cortex. To increase synaptic activity we stimulated the monosynaptic, glutamatergic climbing fibres that excite Purkinje cells via AMPA receptors. We used local field potentials to indicate synaptic activity, and recorded tissue oxygen partial pressure (Ptiss,O2) by polarographic microelectrodes, and CBF using laser-Doppler flowmetry. The disappearance rate of oxygen in the tissue increased linearly with synaptic activity. This indicated that, without a threshold, oxygen consumption increased as a linear function of synaptic activity. The reduction in Ptiss,O2 preceded the rise in CBF. The time integral (area) of the negative Ptiss,O2 response increased non-linearly showing saturation at high levels of synaptic activity, concomitant with a steep rise in CBF. This was accompanied by a positive change in Ptiss,O2. Neuronal nitric oxide synthase inhibition enhanced the initial negative Ptiss,O2 response (‘dip’), while attenuating the evoked CBF increase and positive Ptiss,O2 response equally. This indicates that increases in CBF counteract activity-induced reductions in Ptiss,O2, and suggests the presence of a tissue oxygen reserve. The changes in Ptiss,O2 and CBF were strongly attenuated by AMPA receptor blockade. Our findings suggest an inverse relationship between negative Ptiss,O2 and CBF responses, and provide direct in vivo evidence for a tight coupling between activity in postsynaptic AMPA receptors and cerebellar oxygen consumption.
Nerve tissue is among the highest oxygen-consuming tissues in the body. The oxygen is used for oxidation of glucose which is thought to provide almost all the energy needed by neurones to support brain activity (Siesjo, 1978). There is a tight coupling between neuronal activity, blood flow and glucose and oxygen consumption (Clarke & Sokoloff, 1994). The tight and robust neurovascular coupling (Iadecola, 2004) is driven by synaptic mechanisms (Lauritzen, 2005), and ensures that oxygen and glucose are supplied to the brain to match activity (Raichle, 1998). There is a strong interest in exploring the energetic costs of single elements of chemical neurotransmission in order to provide an energy budget for signalling in the grey matter of the brain (Attwell & Laughlin, 2001; Lennie, 2003). This, as well as an exact knowledge of the timing and the quantitative relationship between neuronal activities, blood flow and oxygen consumption, forms the basis for correct interpretation of signals obtained by functional neuroimaging in the working brain (Raichle, 2003; Obata et al. 2004).
The aim of the present work was to investigate oxygen consumption by examining the changes in tissue oxygen partial pressure (Ptiss,O2) during activation in a simple neuronal network in relation to excitatory synaptic activity and changes in blood flow. We chose the excitatory, monosynaptic glutamatergic climbing fibre–Purkinje cell synapse for that purpose. One of the advantages of this system is that synaptic excitation is robust, and that the cerebral blood flow (CBF) response develops slowly during stimulation; that is, over tens of seconds (Mathiesen et al. 1998). In comparison, the CBF response in the somatosensory system develops much faster; that is, of the order of hundreds of milliseconds (Nielsen & Lauritzen, 2001), which may cloud rapid initial decreases in Ptiss,O2. Therefore, the chances of observing changes in Ptiss,O2 independently of blood flow changes were greater in the cerebellar cortex. This allowed us to explore the basis for the activity-induced initial ‘dip’, and the following overshoot in tissue oxygenation. As far as we can ascertain, this is the first study to simultaneously measure tissue oxygenation changes and CBF responses to a wide range of synaptic activation and quantitatively assess their relationships. Specifically, we tested the hypothesis that clamping the CBF response by inhibiting activity-dependent vasodilatation would enhance the oxygen dip without affecting the evoked synaptic activity. At the same time we examined the hypothesis that the tissue oxygenation overshoot that follows the initial early oxygen dip depended on the rise in blood flow. Finally, we tested the hypothesis that blocking postsynaptic activity by glutamatergic AMPA receptor inhibitors would block the early oxygen dip.
Experiments were performed in 26 male Wistar rats (250–350 g). The study was in full compliance with the guidelines of the European Council's Convention for the Protection of Vertebrate Animals Used for Experimental and Other Scientific Purposes and was approved by the Danish National Ethics Committee. Anaesthesia was induced with isoflurane (5.0% for induction, 1.5% during surgery). After surgery, anaesthesia was switched to intravenous α-chloralose (bolus, 45 mg kg−1; supplement, 15 mg kg−1 (20 min) −1) for at least 1 h before data acquisition. Extra supplements of α-chloralose were given upon pilo-erection or increased blood pressure (> 10%). The trachea was cannulated for mechanical ventilation (30% O2–70% N2O during surgery; O2-enriched air thereafter) and the left femoral vein and artery for anaesthesia delivery and for continuous monitoring of mean arterial blood pressure (MABP) and periodic arterial blood gas sampling, respectively (MABP, 108 ± 1 mmHg; PCO2, 36.8 ± 0.3 mmHg; PO2, 123.9 ± 2.0 mmHg; pH 7.39 ± 0.004; mean ±s.e.m.; n= 26). Rectal temperature was monitored and maintained at 37 ± 0.5°C. Intraperitoneal catheters were inserted for application of 7-nitroindazole (7-NI). The animals were placed in a head holder with lignocaine (lidocaine) gel (2%) applied to the contact spots of the ear bars. An open cranial window preparation was placed over parts of the cerebellar vermis and medulla oblongata, and the brain was superfused with artificial cerebrospinal fluid (aCSF) as previously described (Mathiesen et al. 1998). At the end of the experiment animals were killed by an intravenous injection of air. The experimental setup is depicted in Fig. 1.
A single-barrelled glass microelectrode filled with 2 m saline (impedance, 2–3 MΩ; tip diameter, ∼2 μm) was lowered to a depth of 300–500 μm into the vermis of the cerebellar cortex. An Ag–AgCl ground electrode was placed in the neck muscle. Extracellular local field potentials (LFPs) and single unit activity (spikes) of Purkinje cells were recorded at the level of the Purkinje cell layer. Purkinje cells were identified by their ability to fire both simple and complex spikes spontaneously or the production of a complex spike at 5–8 ms after electrical stimulation of the inferior olive. The preamplified (× 10) signal was A/D-converted, amplified, filtered (spikes, 300–6000 Hz bandwidth; LFPs, 0.5–2400 Hz bandwidth) and digitally sampled using the Power 1401 interface (Cambridge Electronic Design, Cambridge, UK) connected to a PC running the Spike 2.5 software (Cambridge Electronic Design). Digital sampling rates were at 25 kHz for spikes and 5 kHz for LFPs.
Climbing fibre stimulation
A coated, bipolar stainless steel electrode (SNEX 200, Rhodes Medical Instruments, Inc., Woodland Hills, CA, USA; contact separation, 0.25 mm) was stereotaxically lowered into the caudal part of the inferior olive as described (Mathiesen et al. 1998). Positioning was optimized by means of the maximal response of LFPs to low-frequency stimulation (0.5 Hz) and by the occurrence of complex spikes in response to stimulation. Each Purkinje cell receives excitatory input from one climbing fibre that makes synaptic contact with the Purkinje cell soma and proximal dendrites. Climbing fibre stimulation induces release of glutamate, which activates Purkinje cells via postsynaptic AMPA receptors (Mathiesen et al. 1998), resulting in activation of Purkinje cells located within parasagittal bands (Dunbar et al. 2004). In optical imaging studies and multiple electrode recordings the width of these bands in the mediolateral direction ranged from 250 to 500 μm (Hanson et al. 2000; Fukuda et al. 2001). Stimulation was given as 0.2-ms square-wave pulses at 0.15 mA (ISO-flex, A.M.P.I., Israel) with increasing frequencies for 15 s. We used stimulation frequencies between 0.5 and 10 Hz; 10 Hz corresponds to the highest firing frequency of the inferior olive–climbing fibre system (Llinas & Volkind, 1973; Llinas et al. 2002) and evoked synaptic currents might be unreliable above this frequency (Silver et al. 1998).
Three sets of experiments were performed for this study. First, we examined the relationship between increases in excitatory synaptic activity and local changes in Ptiss,O2 and CBF within the activated region. To increase neuronal activity we stimulated climbing fibres with increasing frequencies of 0.5, 2, 5, 7.5 and 10 Hz (n= 10). We correlated the summed LFP (ΣLFP, see below) to evoked changes in Ptiss,O2 and CBF to obtain a quantitative relationship between these variables.
In the second part of the study, we clamped CBF during activation thereby diminishing the supply of O2 to the activated area. To this end, we applied the neuronal nitric oxide synthase (nNOS, type I NOS) inhibitor, 7-NI (40 mg kg−1i.p., in 5% cremophor; Sigma Chemicals, Vallensbaek, Denmark; n= 8). Climbing fibres were stimulated at 10 Hz and the responses compared before (CTR) and 30 min after drug application (7-NI).
Finally, we wanted to establish whether postsynaptic activity in Purkinje cells could explain the oxygen consumption found during activation. Thus, the third part of this study examined the effect of AMPA receptor blockade on activation-induced Ptiss,O2 changes. We superfused the cerebellum with the AMPA receptor antagonist 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX; 0.5 mm in aCSF; n= 8). Responses to stimulation at 10 Hz were compared before (CTR), during CNQX (CNQX), and after a wash out period (Wash out). Additionally, in five of the 7-NI-treated animals from the second part of the study, CNQX (0.5 mm) was applied after 7-NI had been given. In order to obtain complete recovery of responses, CNQX application was kept brief and no attempt was made to achieve a complete attenuation of signals in all animals. Only animals exhibiting a significant negative Ptiss,O2 response (area below 2 s.d. from baseline) were included in the CNQX experiments. Interstimulus interval in all three groups was 3 min.
Cerebellar cortical blood flow measurement
CBF was recorded continuously using laser-Doppler flowmetry (LDF). The LDF probe was at a fixed position ∼0.3 mm above the pial surface in a region devoid of large vessels (wavelengths, 780 nm; fibre separation, 250 μm; PeriMed, Järfälla, Sweden). The probe, measuring CBF changes down to a depth of 1000 μm, was placed as close as possible to the micro- and oxygen electrode. The LDF signal was smoothed with a time constant of 0.2 s (PeriFlux 4001 Master, PeriMed), sampled with 10 Hz, A/D converted and digitally recorded using Spike 2.5 software (Cambridge Electronic Design).
Tissue PO2 measurements
We used a modified Clark-type polarographic oxygen microelectrode (OX-10, Unisense A/S, Aarhus, Denmark) with a guard cathode for Ptiss,O2 measurements (Revsbech, 1989). The advantage of this electrode type is its small tip size (3–5 μm in this study) and its built-in guard cathode which removes all oxygen from the electrolyte reservoir. This enabled us to measure Ptiss,O2 over time (and different treatment conditions) with excellent long-term stability (signal drift, 0–0.5% h−1). The field of sensitivity is a sphere of 2 × tip diameter. The electrodes used in this study were constructed so that the 90% response time was < 1 s and the stirring sensitivity was nearly negligible at < 0.8% (Revsbech, 1989). Oxygen microsensors respond linearly to changes in oxygen concentration. Calibration of each electrode was performed in air-saturated and oxygen-free saline (0.9% at 37°C) before and after each experiment with reproducible oxygen measurements. Mean Ptiss,O2 was 38.4 ± 2.6 mmHg (mean ±s.e.m.; n= 26) and ranged from 12.7 to 64.4 mmHg. The oxygen electrodes were stepwise, vertically inserted in the cerebellar cortex and positioned in the same sagittal line and at the same cortical depth as the glass microelectrodes. The distance between the two electrodes was ∼100–200 μm. Both electrodes and the LDF probe recorded from the same cerebellar folia and were kept in the same position throughout the experiment. No linear drifts in baseline occurred during the experiments. The oxygen electrodes were connected to a high impedance picoammeter (PA 2000, Unisense A/S) sensing the currents of the oxygen electrodes. Signals were A/D converted and recorded at 100 Hz (Power 1401 and Spike 2.5 (Cambridge Electronic Design)). Off-line filtering using a low-pass filter of 0.3 Hz was used to remove noise induced by heartbeat and mechanical ventilation as described by Masamoto et al. (2003a).
Data analysis and statistics
Simultaneously recorded LFP, CBF and Ptiss,O2 signals were used for analysis. Stimulation-induced CBF and Ptiss,O2 responses were calculated as percentage change from baseline (20 s prior to stimulation). The magnitude of the response was estimated as the area under the curve above (CBF, positive Ptiss,O2 responses) or below (negative Ptiss,O2 responses) two standard deviations of baseline. This, in conjunction with the long baseline period, ensured that only statistically significant responses were included for further analysis. Individual animals had constant levels of baseline noise throughout the experiment. The initial slope (disappearance rate of oxygen in the tissue; −k) of the early Ptiss,O2 decrease is useful as an indicator of local oxygen consumption (Leniger-Follert, 1977). −k was calculated from the initial constant decrease phase of the negative Ptiss,O2 response (∼1–3 s after stimulation onset). −k was determined as the tangent of the steeply decreasing Ptiss,O2 curve during this period and is defined as the difference in tissue oxygen partial pressure (in percentage change from baseline) divided by time (in seconds). The initial slope was not calculated for the CNQX-treatment group because negative Ptiss,O2 responses were abolished or too small during CNQX application to obtain reliable initial slope calculations. Calculating oxygen consumption from the initial slope requires the determination of oxygen release from haemoglobin in the area of interest. Due to this, and the well-described spatial heterogeneity of brain Ptiss,O2 and local oxygen consumption (Leniger-Follert, 1977; Erecinska & Silver, 2001; Baumgartl et al. 2002; Masamoto et al. 2003b), no attempt was made to calculate local cerebral metabolic rate of oxygen (CMRO2) in absolute terms from our data. Another practical consequence of this spatial heterogeneity was that the temporal profiles of Ptiss,O2 responses were not averaged for groups of animals. The response magnitudes and the initial slopes of the oxygen signals were calculated for each animal from averaged responses of four to six stimulations per treatment condition or stimulation frequency. Calculations were performed by a custom-made analysis program based on Matlab 6.5 (MathWorks Inc., Natick, MA, USA). Responses in which the oxygen signal did not return to baseline (± 2 s.d.) within 75 s after stimulation onset were excluded.
The LFP is a potential change in the extracellular fluid induced by ion fluxes resulting from synchronized activity of ensembles of nerve cells (Llinas & Nicholson, 1974). LFPs in the cerebellar cortex are characterized by dendritic transmembrane ion fluxes (Nicholson & Llinas, 1971). Climbing fibre stimulation-induced LFPs consist of a negligible presynaptic component and a large excitatory postsynaptic potential which is followed by a smaller and slower positive potential (Eccles et al. 1967). A net influx of positive ions into the dendrites of Purkinje cells at the site of synaptic excitation produces a current sink and causes the negative deflection of the LFP. The following decline in negative potential and even its reversal to a positive wave are attributable to the somata and axons of the Purkinje cells, where a net efflux of positive ions acts as a current source (Eccles et al. 1966). We calculated LFP amplitudes as the difference between the large negative deflection and the baseline before stimulation onset. Additional off-line digital low-pass filtering (0.5–300 Hz) of the LFP signal ensured that no high frequency (spike) signals contaminated the LFP measurements. Mean LFP amplitudes were calculated for each frequency (0.5–10 Hz) or treatment condition (e.g. CTR, 7-NI, CNQX, Wash out) and the summed potential (ΣLFP) was calculated as the product of the LFP amplitude (in mV) and the stimulus rate (in Hz). The ΣLFP was used as an indicator of synaptic activity (Mathiesen et al. 1998).
To be able to compare data between animals, responses (LFP, CBF, Ptiss,O2) were normalized to the response evoked by 10 Hz stimulation. ΣLFP was correlated to the evoked CBF and Ptiss,O2 responses, either by least-squares linear regression analysis or by polynomial regression in cases where non-linearity was evident from the distribution of data. All data are expressed as means ±s.e.m. ANOVA on repeated measures followed by Bonferroni post hoc analysis was performed for statistical comparison of multiple treatment conditions. When appropriate for simple comparisons within groups (‘before-after’), a paired t test was used. Values were considered statistically significant at P < 0.05.
Co-localized synaptic activity, Ptiss,O2 and CBF were simultaneously measured (Fig. 1). We used local field potential recordings, a Clark-type polarographic oxygen microelectrode and a LDF probe to assess the interdependency of synaptic activity, Ptiss,O2 and CBF responses and to clarify the nature of the biphasic Ptiss,O2 signals evoked by neuronal activation.
We varied the input function by increasing the frequency of climbing fibre stimulation. Evoked responses of local Ptiss,O2 and CBF were frequency-dependent, as illustrated in Figs 1B, 2 and 3. The CBF responses were monophasic increments, whereas the Ptiss,O2 response pattern ranged from decreases alone at low frequencies to biphasic responses at higher stimulus frequencies (Fig. 2A). The biphasic responses consisted of an initial decrease in Ptiss,O2 (negative Ptiss,O2 response) followed by a slower and longer lasting positive peak (positive Ptiss,O2 response). Comparison of the time courses of the Ptiss,O2 and CBF signals showed that the decrease in Ptiss,O2 began before the increase in CBF was evident (Fig. 2A).
The disappearance rate of oxygen in the tissue (−k), reflecting early local oxygen consumption during activation, increased as a function of stimulation frequency, which is shown by the increasing initial slope of the negative Ptiss,O2 signal during the first seconds of stimulation (Fig. 2A inset, and Fig. 3A). In comparison, the magnitude of the area of the negative Ptiss,O2 response increased only up to stimulation frequencies of 5 Hz, after which it clearly levelled off or even decreased (Fig. 3B). The area of the positive Ptiss,O2 response paralleled the CBF increase at stimulation frequencies above 5 Hz (Fig. 3C and D).
Relationship of ΣLFP to Ptiss,O2 and CBF responses
As a measure of neuronal activation, synaptic activity was calculated for each stimulation train as the average field potential amplitude multiplied by stimulus frequency (ΣLFP). The LFP represents the spatially weighted sum of extracellular currents due to synchronous synaptic activity in a group of neurones in response to one stimulation impulse (Llinas & Nicholson, 1974). The summation of LFP amplitudes over time is therefore a plausible indicator of total synaptic activity evoked by repeated stimulation impulses and has been widely used to describe the inter-relationship of synaptic activity and CBF or metabolism (Mathiesen et al. 1998; Devor et al. 2003; Sheth et al. 2004; Thomsen et al. 2004). LFP amplitudes decreased at higher stimulation frequencies (Fig. 1C) due to reduced glutamate release (Silver et al. 1998; Foster & Regehr, 2004) or postsynaptic AMPA receptor saturation (Harrison & Jahr, 2003). In consequence, ΣLFP started to plateau at frequencies above 5 Hz (Fig. 3E).
The relationship between ΣLFP and the initial slope of the negative Ptiss,O2 response (−k) is shown in Fig. 2B (one animal) and Fig. 4A (all animals, normalized data). Evoked synaptic activity was found to be linearly related to −k and the data were well described by a linear regression passing through the origin (R2= 0.987; P= 3.37 × 10−6). This finding implies that during the early phase of activation, oxygen is consumed at all levels of synaptic activity.
The area of the negative Ptiss,O2 response was calculated to obtain a measure of the integrated effect of synaptic activity on Ptiss,O2 over time. However, in contrast to −k, the relationship between the area of the negative Ptiss,O2 response and synaptic activity (ΣLFP) was non-linear (Fig. 4B). The negative Ptiss,O2 area increased linearly at low or moderate levels of synaptic activity, but reached a plateau at higher levels. In comparison, neither the area of the positive Ptiss,O2 response nor the CBF response showed signs of saturation with increasing synaptic activity (Fig. 4C and D). The graphs in Fig. 4C and D show a non-linear relationship between synaptic activity and CBF, and positive Ptiss,O2 responses, respectively. They also illustrate that the positive Ptiss,O2 response followed the increase in CBF only at higher levels of synaptic activity.
The pattern of these responses suggested strong interaction between the evoked CBF responses and the magnitudes of the negative and positive Ptiss,O2 responses. The negative Ptiss,O2 response area increased as a function of ΣLFP up to the level of synaptic activity at which the magnitude of the CBF response increased sharply. At this point also, the positive Ptiss,O2 response appeared (Fig. 4B–D). We hypothesized that the activity-induced, biphasic Ptiss,O2 signal originates from increased oxygen consumption followed by increased oxygen supply due to the rise in CBF, and that this increase in oxygen supply is responsible for attenuating the negative Ptiss,O2 response.
Effect of attenuated CBF rise on Ptiss,O2 responses and synaptic activity
To validate the above hypothesis, we tried to clamp CBF during activation using 7-NI (40 mg kg−1), which is a relatively selective, irreversible nNOS inhibitor. 7-NI effectively attenuates stimulation-induced CBF responses (Cholet et al. 1996; Yang et al. 1999; Lindauer et al. 1999), without affecting either resting or stimulation-induced cerebral glucose utilization (Cholet et al. 1997). Furthermore, NOS inhibition has no effect on stimulation-induced neuronal activity (Akgoren et al. 1996; Lindauer et al. 1996) or on the basal CMRO2 (Iadecola et al. 1994; Chi et al. 2003). Thus 7-NI is a suitable pharmacological tool to investigate the effect of neuronal activation on tissue oxygenation changes under conditions of an attenuated CBF response.
In the present study, 7-NI did not result in significant changes of arterial blood gas parameters, MABP (+1.6 ± 1.6%) baseline CBF (−11.4 ± 6.3%), or baseline Ptiss,O2 (−0.8 ± 2.7%). In the following experiments, 10 Hz stimulations were given repeatedly under control conditions and in the presence of 7-NI (n= 8). This stimulation frequency of 10 Hz was chosen as it was found to produce robust biphasic Ptiss,O2 responses. Figure 5A shows an example of Ptiss,O2 and CBF responses before (CTR) and after administration of 7-NI. 7-NI attenuated the stimulation-induced CBF increase by 60.3 ± 4.6% (P= 0.003; Fig. 5B). At the same time, the negative Ptiss,O2 response increased fourfold (+403.4 ± 22.1%; P= 0.01), while the positive Ptiss,O2 response decreased by 78.1 ± 8.8% (P= 0.003; Fig. 5B). These findings confirmed our hypothesis, which predicted that 7-NI would enlarge the negative and reduce the positive Ptiss,O2 responses.
Interestingly, −k was not affected by 7-NI (+2.7 ± 6.9%; P= 0.71). This indicated that the initial slope of the negative Ptiss,O2 response was independent of the associated CBF rise (Fig. 6C). In comparison, both the positive and the negative Ptiss,O2 areas depended on the magnitude of the evoked CBF response. The positive Ptiss,O2 response was linearly correlated to the evoked CBF increase (Fig. 6A; linear regression analysis: y= 0.673x−37.2; R2= 0.67, P= 0.0001) which demonstrated that the positive Ptiss,O2 response in the cerebellum is indeed a consequence of increased oxygen supply. However, CBF had to rise above a certain threshold before a positive Ptiss,O2 response was evident. The negative Ptiss,O2 response decayed exponentially with increasing CBF (Fig. 6B). Of most importance, ΣLFP was not affected by NOS inhibition (−2.5%, P= 0.26; Fig. 5C). Thus, the observed changes in the Ptiss,O2 signal were not due to reduced synaptic activity. So far, our findings indicated that the origin of the positive Ptiss,O2 response is an increase in O2 supply derived from the activation-induced increase in CBF. At the same time, the negative Ptiss,O2 response is modulated by the increment in CBF.
Effect of glutamate receptor blockade on synaptic transmission, Ptiss,O2 and CBF responses
Here we examined whether postsynaptic activity in Purkinje cells could explain oxygen consumption during activation. Therefore, we studied the effect of AMPA receptor blockade on activation-induced Ptiss,O2 changes. AMPA receptor blockade has previously been shown to effectively inhibit glutamate-mediated synaptic transmission at the climbing fibre–Purkinje cell synapse (Mathiesen et al. 1998). Beside this direct action of CNQX on postsynaptic excitation, indirect actions might increase inhibition (Brickley et al. 2001) or decrease disinhibition of Purkinje cells (Satake et al. 2000) which could also affect the Purkinje cell output (spiking) activity.
We found that superfusion of the cerebellum with the AMPA receptor antagonist, CNQX (0.5 mm), had no effect on MABP, baseline CBF (+2.8 ± 2.7%) or baseline Ptiss,O2 (+2.2 ± 2.3%; n= 8). CNQX attenuated ΣLFP by 68.9 ± 7% (P < 0.001; Fig. 7C) and CBF responses by 69.0 ± 15.3% (P < 0.001; Fig. 7A and B; n= 8). At the same time, CNQX significantly inhibited both the negative (−81.3 ± 13.9%; P= 0.004) and the positive Ptiss,O2 responses (−84.4 ± 8.1%, P < 0.001; Fig. 7A and B). All effects of CNQX were reversible with no significant differences between control conditions and washout (Fig. 7). These findings supported the idea that the decrease in tissue oxygenation seen during activation depends upon postsynaptic activity in Purkinje cells. This was confirmed by additional experiments in which CNQX was applied after pretreatment with 7-NI (n= 5; Fig. 8).
The direct assessment of local changes in Ptiss,O2 in conjunction with CBF and electrophysiological recordings revealed a linear relationship between the disappearance rate of oxygen in the tissue and synaptic activity during the initial phase of neuronal activation in the cerebellar cortex. This increase in oxygen metabolism was evident at all levels of synaptic activity and depended on activation of glutamatergic AMPA receptors localized postsynaptically on Purkinje cell dendrites. Hence in vivo, early increases in neuronal aerobic metabolism depend on postsynaptic activity as previously shown in vitro for the hippocampus (Kasischke et al. 2004). During the initial phase of activation we observed a temporal uncoupling of oxidative metabolism and CBF as a decline in Ptiss,O2 was observed preceding the rise in CBF. Following the initial oxygen ‘dip’, Ptiss,O2 increased due to the rise in CBF. Application of 7-NI attenuated the CBF rise, while increasing the negative Ptiss,O2 response and decreasing the positive Ptiss,O2 response. These findings imply that, in the cerebellum, the activity-induced reduction in Ptiss,O2 is counteracted by an increase in oxygen supply due to the rise in CBF. The data also support the hypothesis that the positive Ptiss,O2 response is an overshoot caused by the evoked increase in CBF. The fact that oxygen consumption increased before CBF suggests a reserve capacity of oxygen in the tissue as proposed previously (Buxton, 2001; Mintun et al. 2001; Sheth et al. 2004).
Negative Ptiss,O2 and CBF response
In the present study, the typical Ptiss,O2 response was biphasic with an initial decrease followed by an increase. The magnitude of both phases varied with the level of activity and differed between animals, reflecting the well-known spatial heterogeneity of Ptiss,O2 profiles at rest and during activation (Gijsbers & Melzack, 1967; Silver, 1978; Lubbers et al. 1994; Erecinska & Silver, 2001; Masamoto et al. 2004). Despite this heterogeneity, initial Ptiss,O2 reductions preceding evoked CBF responses were observed at all stimulation frequencies. This is consistent with previous studies in the somatosensory and visual cortices in which early decreases in tissue PO2 (Ances et al. 2001) or intravascular oxygen content (Malonek et al. 1997; Jones et al. 2001) occurred before significant CBF responses. The present study confirmed this response pattern in the cerebellar cortex. Our findings strengthen the concept that early activity-induced increases in deoxyhaemoglobin in optical imaging studies (Jones et al. 2001; Devor et al. 2003) and the initial dip seen in functional MRI (fMRI) studies (Menon et al. 1995; Kim et al. 2000) reflect local increases in oxygen consumption. The area of the negative Ptiss,O2 response is thought to reflect changes in oxygen consumption and supply over time. The influence of activity-dependent increases in CBF for the Ptiss,O2 response has been hypothesized previously (Silver, 1978; Ances et al. 2001; Masamoto et al. 2003a; Thompson et al. 2003) but has not been proven by simultaneous measurements of Ptiss,O2, neuronal activity and CBF over a wide range of activities. For example Thompson et al. (2004) showed that activation of a small neuronal population induced a negative Ptiss,O2 change that was partly cancelled out by a positive Ptiss,O2 signal, which arose when a larger population of neurones was activated. They suggested that this was due to a rise in CBF, but CBF was not measured. We demonstrated by varying the synaptic input to a given population of neurones that the level of activity at which the negative Ptiss,O2 area no longer increases coincided with the level of activity at which the CBF response curve escalated (Fig. 4B and C). The non-linear relationship found between evoked synaptic activity and CBF responses in the cerebellum was similar to those previously found in the cerebral cortex (Devor et al. 2003; Sheth et al. 2004).
The hypothesis that the reduction of the negative Ptiss,O2 area was caused by increased oxygen supply due to greatly increased CBF responses was further tested in the 7-NI experiments. We found that the negative Ptiss,O2 area increased and the positive Ptiss,O2 area decreased as the magnitude of the activity-evoked CBF responses declined in the presence of 7-NI. Thus, this study clearly demonstrated an inverse relationship between the negative Ptiss,O2 area and the CBF response, and that at high levels of activation, the negative Ptiss,O2 area, and thereby estimates of oxygen consumption based upon this parameter, are confounded by increased O2 supply. We conclude that the non-linear relationship between the negative Ptiss,O2 area and synaptic activity is explained by increased CBF and oxygen supply and not by a reduction in oxygen consumption.
Tissue and intravascular deoxygenation during activation
The non-linearity between synaptic activity and negative Ptiss,O2 area in the present study raises important issues concerning the initial dip observed with optical imaging techniques and fMRI and whether it can accurately reflect the magnitude of underlying neuronal activity (Nemoto et al. 2004). PO2 in the tissue lies between capillary and mitochondrial PO2 levels and shows a high susceptibility to changes in oxygen consumption (Secomb et al. 2000). The increased oxygen concentration gradient in the tissue during activation (Vanzetta & Grinvald, 1999; Ances et al. 2001; Thompson et al. 2003) results in increased conversion of oxy- to deoxyhaemoglobin. Hence, deoxygenation signals from tissue and vasculature are linked in their response to neuronal activation. We found that the magnitude of the negative Ptiss,O2 area, representing tissue deoxygenation, reaches a plateau at high levels of neuronal activity (Fig. 4B). Thus, the initial dip representing intravascular deoxygenation cannot be expected to increase at high levels of neuronal activity when tissue deoxygenation does not. This implies that the initial dip calculated as peak response or area might not be capable of indicating increases in neuronal activity beyond a given threshold, which could explain some of the inconsistent findings in optical imaging studies correlating neuronal activity to the magnitude of the initial dip (Mayhew et al. 2000; Devor et al. 2003; Sheth et al. 2003, 2004; Nemoto et al. 2004).
Initial slope and activity-dependent increases in early oxygen consumption
As the relationship between synaptic activity and the area of the negative Ptiss,O2 response was confounded by CBF, we then assessed the initial slope of the negative Ptiss,O2 response (−k) to obtain a relative measure of activity-induced oxygen consumption. Local oxygen consumption has previously been determined in absolute terms in the brain cortex and hippocampus by using O2-disappearance measurements during brief periods of total brain ischaemia (Leniger-Follert, 1977; Buerk & Nair, 1993). In the present study, −k was independent of the associated CBF rises, implying that −k is a reliable indicator of early increases in oxygen consumption during activation in the cerebellum. Unlike the negative Ptiss,O2 response area, −k did increase as a linear function of synaptic activity (i.e. ΣLFP; Fig. 4A). This confirms previous findings using optical imaging techniques, which calculated that peak CMRO2 increased with growing stimulation intensities (Jones et al. 2001) and that the relationship between ΣLFP and increases in CMRO2 was linear (Sheth et al. 2004). Sheth et al. (2004) calculated that increased oxygen extraction from the vasculature occurs only above a given threshold of synaptic activity. In comparison, we found no such threshold in the cerebellum, indicating that tissue oxygen consumption increased at even the lowest levels of synaptic activity. This, in conjunction with the observed non-linear increase of CBF with synaptic activity, supports the idea that at rest a significant amount of readily available oxygen in the tissue exists which might serve as a tissue oxygen reserve (Buxton, 2001; Mintun et al. 2001). Importantly, the present study indicated that an increase in local oxygen consumption occurs in response to activation and that this increase is linearly related to the level of synaptic activity.
Activity-induced changes in Ptiss,O2 have been explored in a number of studies (Travis & Clark, 1965; Gijsbers & Melzack, 1967; Sick & Kreisman, 1979; Ances et al. 2001) but recently the relationship between neuronal activity and oxygen consumption has been assessed more directly by recording the two variables simultaneously in the same location. This has shown that negative Ptiss,O2 changes are highly specific for local increases in neuronal activity and quantitatively related to the amount of underlying neuronal activity (Thompson et al. 2003; Masamoto et al. 2003a), although the type of neuronal activity has not been established. Masamoto et al. (2003a) related neuronal activity induced by single activation tasks to voltage-sensitive dye recordings which, however, allowed no discrimination between pre- and postsynaptic activation. Thompson et al. (2003, 2004) focused on the local specificity of tissue deoxygenation signals in relation to spike recordings and confirmed a correlation between spiking activity and negative tissue PO2 responses during activation (Sick & Kreisman, 1979). As neurotransmission and synaptic input in most neuronal circuits are prerequisites for spike generation and are often linearly correlated to changes in output spiking activity (Devor et al. 2003; Jones et al. 2004), it is not possible to deduce by spike measurements alone whether glutamate release, postsynaptic processing or output firing is responsible for the increased oxygen demand during activation.
During climbing fibre stimulation, we observed a linear relationship between synaptic activity and −k as discussed above. Topical application of CNQX almost abolished synaptic transmission as well as the evoked negative Ptiss,O2 response, identifying postsynaptic AMPA receptor activation as necessary for early negative Ptiss,O2 changes. Astrocytes are not activated during electrical stimulation of Purkinje cell afferents as reported in an immunohistochemical study (Tian & Bishop, 2002). Thus, in the climbing fibre–Purkinje cell pathway, increases in oxygen consumption are coupled to activity in postsynaptic neurones. As the climbing fibre–Purkinje cell synapse is one of the most powerful excitatory synapses where glutamatergic transmission leads to activation of AMPA receptors on one of the widest dendritic trees, it has to be determined whether oxygen consumption in other brain regions is also mainly driven by postsynaptic activation. Our findings are consistent with a nuclear magnetic resonance spectroscopy study by Sibson et al. (1998), who showed that rates of neuronal CMRO2 and glutamate release and cycling are stoichiometrically related. However, our study cannot be used to differentiate between oxygen used for postsynaptic processing and oxygen used for generation of spikes. We conclude that the observed early increase in oxygen consumption during activation in vivo is of neuronal origin as has been previously shown in the hippocampus in vitro (Kasischke et al. 2004). Furthermore, the results of the present study demonstrate that neuronal oxygen consumption in the cerebellar cortex increases together with synaptic activity.
Neuronal activity and oxygen transport to tissue
There is still debate concerning the coupling between blood flow responses and oxygen consumption during activation in the brain. Oxygen-limitation models designed to explain this coupling have been based on the assumption that at rest all oxygen in the tissue is consumed and PO2 at the level of the mitochondria is zero (Buxton & Frank, 1997; Hyder et al. 1998; Vafaee & Gjedde, 2000). However during activation, we observed that oxygen consumption in the tissue increased before CBF did, and that in the presence of 7-NI, the magnitude of the negative Ptiss,O2 area increased despite the strong attenuation of the CBF response. These observations suggest that resting Ptiss,O2 is far from zero. Our findings support the existence of a tissue oxygen reserve and are in line with human positron emission tomography data (Mintun et al. 2001), calculations of resting tissue oxygenation levels in the brain (Secomb et al. 2000) and recent modelling of tissue oxygen delivery (Hudetz, 1999; Mintun et al. 2001). The results from our study imply that in the anaesthetized state, the cerebellum has an excess oxygen delivery at rest and that this tissue oxygen reserve is used during the initial phase and at low grades of activation. We suggest that up to a given level of neuronal activity, oxygen consumption can increase even under conditions of constant CBF. On the other hand, the finding of a supra-linear relationship between CBF and CMRO2 changes during activation provides experimental confirmation of the predictions derived from the oxygen-limitation models (Buxton & Frank, 1997; Vafaee & Gjedde, 2000). According to these models, oxygen diffusivity is limited (Gjedde et al. 1991) and during activation, increases in oxygen consumption are supported by disproportionately large increases in CBF (Buxton & Frank, 1997). Combined, our data support the recently proposed hypothesis that a buffer of oxygen availability in the tissue exists and that CBF increases during activation are regulated to maintain this buffer (Buxton, 2001).
In this study we examined the physiological basis of activity-induced changes in tissue oxygen tension in rat cerebellar cortex. We showed that Ptiss,O2 responses are the result of a temporally staggered interplay between neuronal oxygen consumption and evoked CBF increments both driven by postsynaptic activity. At the level of the climbing fibre–Purkinje cell synapse, tissue oxygen consumption is linearly coupled to synaptic excitation with no evidence of a threshold; thus, we found that tissue energy demands are supplied by oxidative metabolism at all levels of synaptic activity. We have also quantitatively assessed the relationships between synaptic activity and tissue oxygenation signals and CBF responses. The findings indicate a supralinear relationship between increases in oxygen consumption and CBF during activation and support the existence of a buffer of oxygen availability in brain tissue.
We thank Lillian Grondahl for expert technical assistance, and Jesper Grondahl for his sophisticated MatLab programming. This study was supported by The Lundbeck Foundation, the Danish Medical Research Council, Vera and Carl Michelsens legat, and The NOVO-Nordisk Foundation.