Rapid fluctuations in extracellular brain glucose levels induced by natural arousing stimuli and intravenous cocaine: fueling the brain during neural activation

Abstract

Glucose, a primary energetic substrate for neural activity, is continuously influenced by two opposing forces that tend to either decrease its extracellular levels due to enhanced utilization in neural cells or increase its levels due to entry from peripheral circulation via enhanced cerebral blood flow. How this balance is maintained under physiological conditions and changed during neural activation remains unclear. To clarify this issue, enzyme-based glucose sensors coupled with high-speed amperometry were used in freely moving rats to evaluate fluctuations in extracellular glucose levels induced by brief audio stimulus, tail pinch (TP), social interaction with another rat (SI), and intravenous cocaine (1 mg/kg). Measurements were performed in nucleus accumbens (NAcc) and substantia nigra pars reticulata (SNr), which drastically differ in neuronal activity. In NAcc, where most cells are powerfully excited after salient stimulation, glucose levels rapidly (latency 2–6 s) increased (30–70 μM or 6–14% over baseline) by all stimuli; the increase differed in magnitude and duration for each stimulus. In SNr, where most cells are transiently inhibited by salient stimuli, TP, SI, and cocaine induced a biphasic glucose response, with the initial decrease (−20–40 μM or 5–10% below baseline) followed by a reboundlike increase. The critical role of neuronal activity in mediating the initial glucose response was confirmed by monitoring glucose currents after local microinjections of glutamate (GLU) or procaine (PRO). While intra-NAcc injection of GLU transiently increased glucose levels in this structure, intra-SNr PRO injection resulted in rapid, transient decreases in SNr glucose. Therefore, extracellular glucose levels in the brain change very rapidly after physiological and pharmacological stimulation, the response is structure specific, and the pattern of neuronal activity appears to be a critical factor determining direction and magnitude of physiological fluctuations in glucose levels.

although glucose is the primary source of energy for the active brain and its proper delivery to brain cells is essential for normal neural functions (Fox et al. 1988; Siesjo 1978), our knowledge of how and why its brain levels are changed remains controversial. It is known that neural activation results in enhanced glucose utilization in brain cells (Sokoloff et al. 1977; Siesjo 1978), suggesting possible decreases of its extracellular levels. On the other hand, neural activation is accompanied by rapid increases in local cerebral blood flow (CBF) (Fox and Raichle 1986; Grubb et al. 1974; Hamadate et al. 2011; Hoge et al. 2005; Kong et al. 2004; Martin et al. 2006; Paulson et al. 2010), pointing at the enhanced entry of glucose into brain tissue from the peripheral blood, where its concentration is much higher (De Vries et al. 2003; Fellows and Boutelle 1993; Silver and Erecinska 1994). However, it remains unclear how this balance is maintained under physiologically relevant conditions and how it is changed during neural activation.

Previous microdialysis studies have revealed that various arousing stimuli such as tail pinch or restraint increase extracellular glucose levels in various brain structures (Fellows et al. 1993; Fellows and Boutelle 1993; Osborne et al. 1997), suggesting that inflow of glucose during neural activation overcomes its consumption. However, electrochemical studies produced more controversial results; increases, decreases, and down-up glucose fluctuations were found in the striatum and hippocampus after arousing stimuli (Lowry et al. 1998; Lowry and Fillenz 1997; Newman et al. 2011). While the reasons for these differential responses are not clear, they could be related to structure-specific differences in neuronal activity. Since local CBF could change in brain tissue on a scale of seconds (Devor et al. 2011; Hirano et al. 2011), area-specific differences in glucose response could be related at least in part to differences in temporal resolution of measurements. To resolve this apparent discrepancy, we employed selective enzyme-based glucose sensors coupled with high-speed amperometry to evaluate the pattern of physiological fluctuations in brain extracellular glucose levels in awake, freely moving rats. Our measurements were conducted with 1-s temporal resolution, thus revealing phasic glucose fluctuations that could be lost or missed with slow time-resolution techniques.

Animals were exposed to three types of natural stimuli with different salience and arousing potential as well as intravenous (iv) cocaine at a low, self-administering dose (1 mg/kg). While a brief tone is a weak sensory stimulus that induces transient cortical EEG desynchronization (Kiyatkin and Smirnov 2010; McClung et al. 1976–1977; Sasaki et al. 1996) without evident behavioral effects, tail pinch (TP), social interaction with male conspecific (SI), and iv cocaine are complex arousing stimuli that induce metabolic neural activation as well as relatively strong and prolonged behavioral and physiological responses (Kiyatkin et al. 2002; Kiyatkin and Brown 2005). Although cocaine differs from natural stimuli by the mechanisms of its action, its iv administration induces rapid neural activation as assessed by monitoring of EEG and EMG (Kiyatkin and Smirnov 2010) as well as single-unit activity in the nucleus accumbens (NAcc) (Kiyatkin and Brown 2007). To evaluate structural specificity of brain glucose response, our measurements were conducted in two basal ganglia structures, NAcc and substantia nigra pars reticulata (SNr), which have profound differences in spontaneous impulse activity and its changes induced by arousing stimuli. Most accumbal neurons are silent or sporadically fire under quiet resting conditions and show robust excitations after exposure to salient sensory stimuli and during motor activation (Kiyatkin and Brown 2007; Kiyatkin and Rebec 1996). In contrast, most SNr neurons are autoactive, have high rates of basal impulse activity and typically show transient inhibitions after sensory stimulation and movement activation (DeLong et al. 1984; Deniau et al. 2007; Windels and Kiyatkin 2004). To assess the relationships between glucose response and behavioral activation, electrochemical measurements were supplemented by recording of animal locomotion.

Although our primary goal was to examine physiological fluctuations in extracellular brain glucose levels, several control studies were also conducted to establish the reliability of electrochemical measurements of glucose in behaving animals and assess the possible mechanisms underlying rapid, differential fluctuations in glucose levels found in the NAcc and SNr. Since stimulus-induced changes in electrochemical currents generated by glucose sensors in vivo could be potentially affected by concomitant changes in other physical and chemical factors (i.e., other oxidizable substances, pH, brain temperature), parallel recordings were conducted with enzyme-free, glucose-null sensors to exclude these nonspecific contributions. Since glucose entry from the peripheral blood could be a possible mechanism for its increase in brain tissue, we also examined how extracellular levels of glucose in the NAcc and SNr are affected by rapid increases in blood glucose levels induced by its iv administration. Finally, to directly assess the contribution of neuronal activity in mediating rapid fluctuations in brain glucose found in the NAcc and SNr, we examined how glucose levels in these structures are affected after neuronal activation or inhibition induced by local microinjections of glutamate (GLU) and procaine (PRO), respectively.

METHODS

Subjects and surgical preparations.

Twenty-six male Long-Evans rats (Charles River Laboratories) weighing 460 ± 40 g at the time of testing were used in this study. Rats were individually housed in a climate-controlled animal colony maintained on a 12:12-h light-dark cycle (lights on at 0700), with food and water available ad libitum. All procedures were approved by the National Institute on Drug Abuse-Intramural Research Program Animal Care and Use Committee and complied with the Guide for the Care and Use of Laboratory Animals (NIH Pub. No. 865-23). Maximal care was taken to minimize the number of experimental animals and their possible discomfort or suffering at all stages of the experiment.

This study describes the results of three in vivo experiments performed in freely moving rats. In these experiments, we examined changes in NAcc and SNr extracellular glucose levels induced by natural arousing stimuli (experiment 1, 12 rats), iv glucose injections (experiment 2, 6 rats), and local microinjections of either GLU or PRO near the glucose sensing area (experiment 3, 8 rats). Each rat was implanted under general anesthesia (Equithesin 0.33 ml/100 g ip; dose of pentobarbital sodium 32.5 mg/kg and chloral hydrate 145 mg/kg) with either one (n = 17) or two (n = 8) BASi cannulas (Bioanalytical Systems, West Lafayette, IN) for future insertions of the sensor(s) in either the medial part of the NAcc (NAcc shell) or SNr. Typically, target coordinates were for the NAcc: AP +1.2 mm, ML ± 0.8 mm, and DV +7.2–7.4 mm and for SNr: AP−5.6 mm, ML ± 2.2 mm, and DV +7.8 mm, according to the stereotaxic atlas of Paxinos and Watson (1998). The guide cannula hubs were fixed to the skull with a head mount constructed from dental acrylic that was secured with three stainless steel bone screws. When not in use, stainless steel obdurators were inserted into the cannulas to prevent occlusions. During the same surgical procedure, 18 rats were also implanted with a chronic jugular catheter, which ran subcutaneously to the head mount and was secured to the same head assembly. Rats were allowed a minimum of 4 days of postoperative recovery; jugular catheters were flushed daily with 0.2 ml of sterile saline.

Glucose and glucose-null sensors and their in vitro calibration.

Commercially produced glucose oxidase-based biosensors (Pinnacle Technology, Lawrence, KS) were used to measure changes in extracellular glucose concentration ([glucose]) in the brain. These sensors are prepared from Pt-Ir wire of 180-μm diameter, with a sensing cavity of ∼1-mm length on its tip and a sensing area of ∼0.56 mm². The active electrode is incorporated with an integrated Ag/AgCl reference electrode. On the active surface of these sensors that is held at a potential of +0.6 V versus the reference electrode, glucose oxidase enzyme converts glucose to glucono-1,5-lactone and hydrogen peroxide, which is detected as an amperometric oxidation current (Hu and Wilson 1997). The potential contribution of ascorbic acid to the measured current is competitively reduced by colocalizing ascorbic acid oxidase enzymes on the active surface of the sensor. This enzyme converts ascorbic acid to nonelectroactive dehydroascorbate and water. In addition, a negatively charged Nafion polymer layer under the enzyme layer serves to exclude the contribution of endogenous anionic compounds. An outer polyurethane membrane is cast over the entire sensor surface, serving to limit diffusion of glucose to the enzymatic layer as well as to provide additional screening of interferents.

To validate the reliability of glucose measurements in vivo, we also used glucose-null sensors (Pinnacle Technology), which were identical to glucose sensors except for the absence of glucose oxidase. Therefore, when used under identical in vivo conditions, these electrodes are exposed to the same chemical and physical factors (i.e., various electroactive interferents, temperature, pH, etc.) as glucose sensors but are insensitive to glucose. Therefore, the difference in currents detected by glucose and glucose-null sensors (“current differentials”) in the same brain areas under identical conditions provides the best possible tool to reveal actual changes in extracellular glucose levels. Both types of sensors were tested in vitro with the same protocol before and after in vivo recording, and they were used in identical in vivo experiments. Although some of our rats had two implants for targeting both structures of interest, only one sensor (either glucose or glucose-null) was used in each recording session to avoid between-sensor cross talk that could occur during simultaneous use of two sensors of this design.

In vitro calibrations of both types of sensors were conducted in PBS (pH 7.3; t° 23°C) by incrementally increasing the concentration of glucose (Sigma-Aldrich) from 0 to 1, 2, 3, 4, and 5 mM followed by a single addition of ascorbate (250 μM). As shown in Fig. 1, glucose sensors used in this study (n = 22) produced incremental current rises with increases in [glucose]. Substrate sensitivity of glucose sensors varied from 3.5 to 16 nA/1 mM (mean for the first 1 mM test = 7.11 ± 0.84 nA; SD = 3.92 nA) and slightly decreased with increases in [glucose]. The increase was slightly deviated from linear within the entire concentration scale (see dashed lines in Fig. 1), but within the concentrations between 0 and 2 mM (the range of brain glucose levels according to previous studies; see discussion) it was highly linear. Glucose sensors showed very small changes in currents with addition of ascorbate at 250 μM (mean 0.28 ± 0.10 nA). Electrodes with low initial glucose sensitivity (<2 nA/1 mM glucose) or poor glucose/ascorbate selectivity were rejected from further use. As shown in Fig. 1, the mean postrecording calibration curve was almost identical to that before recording, except for a 1- to 2-nA shift below the initial curve. Surprisingly, after 7–8 h of staying in the brain, the sensitivity and selectivity of glucose sensors remained virtually unchanged (postrecording calibration: 1 mM glucose, 6.11 ± 0.77 nA, SD 3.62 nA and 250 μM ascorbate, 0.29 ± 0.07 nA). In a sample of glucose sensors (n = 4), we also determined their sensitivity to dopamine, another potential contributor to glucose oxidation currents. While these sensors were sensitive to dopamine at large, 1 μM concentrations (0.15–0.70 nA), current changes induced by dopamine at 100 nM were within a background noise; these levels exceed maximal changes in dopamine induced by arousing stimuli as reported in both microdialysis and electrochemical studies (Abercrombie et al. 1989; Kiyatkin and Gratton 1994; Owesson-White et al. 2012; Wightman et al. 2007).

Fig. 1.

Results of prerecording and postrecording in vitro calibrations of glucose and glucose-null sensors used in this study. Glucose sensors (n = 22) both before and after in vivo recordings showed gradual increases in electrochemical currents following 5 repeated applications of glucose (1 mM) and much weaker responses to application of ascorbate (250 μM). During postrecording in vitro calibrations, both basal currents and responses to glucose proportionally decreased, but sensitivity and selectivity remained the same. In contrast, glucose-null sensors were fully insensitive to glucose but showed a comparable response to ascorbate (see text for mean values).

Calibration of glucose-null sensors (see Fig. 1) revealed that they generated no current after repeated applications of glucose and showed an equally small current response to application of 250 μM ascorbate (0.25 ± 0.26 nA). In contrast to glucose sensors, glucose-null electrodes showed lower background currents (1.00 ± 0.11 nA vs. 2.66 ± 0.31 nA, P < 0.001) at the start of prerecording calibrations (see basal value in Fig. 1); these currents further decreased during postrecording calibrations (0.26 ± 0.02 nA vs. 1.00 ± 0.11 nA).

In addition to routine pre- and postrecording calibrations of glucose and glucose-null sensors, we also conducted two in vitro control tests. Since electrochemical currents following long-term recordings in vivo show a consistent negative trend in amplitude, in the first test we examined the changes in electrochemical currents detected by glucose-null electrodes in vitro at stable ambient temperatures (22–23°C) for 8 h, the approximate duration of our behavioral experiment. These data were used for correlating tonic changes in electrochemical currents detected by both glucose and glucose-null sensors in vivo. Since temperature could affect both the basal electrochemical currents and sensor sensitivity, in the second in vitro test we determined current response to 1 mM glucose at 22°C and 37°C, an approximate temperature in deep brain structures of awake rats (Kiyatkin 2010). Substrate sensitivity of glucose sensors (n = 6, double measurements) was strongly temperature dependent, with almost doubling of current response at 37°C vs. 21–22°C (8.31 ± 0.28 nA vs. 4.29 ± 0.12 nA, increase of 95.55 ± 8.30% or ∼0.25 nA for 1°C). This temperature coefficient was used for transferring current data (nA) obtained with glucose sensors in the behavioral experiment into concentration values (μM). By monitoring basal electrochemical currents after slow increases and decreases in temperature within physiological range (±3°C), we also determined that both glucose and glucose-null sensors are temperature sensitive, producing ∼0.08-nA current change with 0.5°C temperature change (n = 6; 0.083 ± 0.007 nA).

Experimental protocol.

All in vivo electrochemical procedures occurred in an electrically insulated chamber (38 × 47 × 47 cm) located in a larger open-faced cabinet. The cage was illuminated continuously by a dim 20-W light bulb, a room-wide air filter fan provided background noise, and ambient temperature was maintained within 23–24°C. Two speakers were placed above the cage for providing audio stimuli. The bottom of the cage was covered with wood chip bedding, which remained in place during habituation and recording of each rat. The cage was equipped with four infrared motion detectors (Med Associates, Burlington, VT), which were used for monitoring animal locomotion. Prior to recording, rats were habituated to the testing environment for a minimum of 6 h/day for 3 consecutive days.

At the beginning of each experimental session, rats were minimally anesthetized (<2 min) with isoflurane and a calibrated sensor (either glucose or glucose-null) was inserted into the brain through the guide cannula. The rat was then placed into the testing chamber, and the sensor was connected to the potentiostat (model 3104, Pinnacle Technology) via an electrically shielded flexible cable and a multichannel electrical swivel. Additionally, the injection port of the jugular catheter on the head mount was connected to a plastic catheter extension that allowed stress-free drug delivery from outside the cage, thus minimizing possible detection of the iv drug injection by the rat. Testing began a minimum of 135 min after insertion of the sensor when the baseline current was relatively stabilized (see results). In the main behavioral experiment, rats were exposed to four types of stimuli: 5-s sound (880 Hz, 75 dB), 3-min TP, 3-min social interaction with a novel male conspecific of a similar age and weight (SI), and iv cocaine injection (1 mg/kg in 0.2 ml over 20 s). For TP, a wooden clothespin was attached to the base of the tail; in contrast to other studies that employed a metal clamp, this is a mild form of TP that is not harmful to the animal. For SI trials, a novel male rat was introduced into the cage for 3 min. Each rat was tested with all stimuli twice per session (except for a single sound), with each stimulus presented in random order and separated by at least 45 min (natural stimuli) or 60 min (cocaine). Thus during each session the rat was tested with an audio stimulus, two pairs of arousing stimuli (TP and SI), and two cocaine injections. At the end of each session, rats were lightly anesthetized (Equithesin, 0.6–0.7 ml by slow iv injection, ∼2 min) and disconnected from the potentiostat, and the biosensor was carefully removed for postrecording calibration. Rats were allowed to recover from anesthesia, and their jugular catheters were flushed with sterile saline before they were returned to the animal colony. Rats with bilateral cannula implants were tested a second time, after one free day. In these rats, the order of stimulus presentation was reversed.

Rats for experiment 2 (n = 6) were prepared similarly to the main behavioral experiment 1 and were implanted with one cannula (to reach either NAcc or SNr) and a chronic iv catheter. In this experiment, glucose currents were monitored during one recording session, when rats received three iv glucose injections (15 mg in 0.3 ml over 30 s, dissolved in saline). This dose of glucose (15 mg), after its distribution within the entire circulatory system of the rat (volume ∼30 ml), should increase blood glucose levels by ∼3 mM. The time interval between consecutive glucose injections was at least 90 min.

Rats for experiment 3 (n = 8) were implanted with a modified cannula connected in parallel with a small-diameter stainless steel needle (28 gauge, external diameter 360 μm) that allowed us to inject chemical substances into brain tissue near the sensor's sensing area (∼500 μm). This needle had an internal volume of ∼0.5 μl and was filled during the surgery with artificial cerebrospinal fluid, which was used as a vehicle to dissolve the test substances. The top part of the needle was equipped with a miniature plastic cap that allowed the content of the needle to be moved into brain tissue (at 0.1-μl steps) after its implantation, thus preventing its blockade. On the day of an experiment, under minimal isoflurane anesthesia, a glucose sensor was inserted into brain tissue and the top of the intracranial needle was connected to a catheter extension filled with the solution of test substance and secured to the electrical cable. This catheter extension was then connected via another tubing to a pump (model A-99, Razel Scientific Instruments, St. Albans, VT) that allowed slow delivery of the substance of interest (0.1 μl over 10 s) during electrochemical recordings. Because of the prefilling with a neutral solvent, the design of this parallel sensor cannula-intracranial injection device allowed us first to test the effects of a control vehicle injection (10 s) and then to examine the effects of the substance of interest; dose was regulated by changing the injection's duration. PRO, a short-acting local anesthetic drug of the amino ester group (procaine hydrochloride; 1–10% weight/volume solution, 0.2–1.0 μl), and GLU (l-glutamate monosodium salt, 1–5 mM solution, 0.2–0.5 μl) were used to induce local neuronal inhibition and activation, respectively. As shown previously, dorsal and ventral striatal neurons in awake, unrestrained rats are very sensitive to local applications of PRO and GLU, showing dose-dependent inhibitions and excitations, respectively (Kiyatkin and Rebec 1999, 2000). In contrast, SNr neurons have a lower sensitivity to GLU (Windels and Kiyatkin 2004), but they are effectively inhibited by PRO because of blockade of Na⁺ channels. Several microinjections (n = 3–8) of either GLU or PRO were typically performed during the session in rats well habituated to the testing environment. Near the end of the session, rats received a single intraperitoneal injection of Equithesin (1 ml) to examine the pattern of glucose response during the transition from a waking state to general anesthesia.

Sensor placement verification.

After experimental sessions were completed, rats were deeply anesthetized with Equithesin and transcardially perfused initially with room temperature PBS (pH 7.4) followed by 10% formalin. Brains were sectioned on a cryostat to a thickness of 45 μm, and some of them were stained with cresyl violet. The location of the sensors within the NAcc shell and SNr was verified with the stereotaxic atlas of Paxinos and Watson (1998).

Data analysis.

Electrochemical data were sampled at 1 Hz (i.e., mean current over 1 s) with the PAL software utility (version 1.5.0, Pinnacle Technology) and analyzed with two time resolutions. Slow changes in electrochemical current were analyzed with 60-s quantification bins for 5 min before and 20 (sound), 40 (TP and SI), and 60 (cocaine) min after stimulus onset. Rapid current changes were analyzed with 4-s bins for 60 s before and 300 s after stimulus onset. Although data could be analyzed with a finer temporal resolution, the 4-s bin appears to be optimal for detecting rapid current changes within the timescale of our stimuli while simultaneously reducing the contribution of noise. First, data obtained with glucose and glucose-null sensors were analyzed separately. Because our sensors differed in background currents and substrate sensitivity in vitro, the currents were transformed into relative changes, taking a basal value (16 and 300 s prestimulus for rapid and slow changes, respectively) as 0. Then, we determined current differentials (i.e., mean changes generated by glucose sensors minus mean changes generated by glucose-null sensors), which show the component of current changes determined by glucose. These data were calibrated in micromolar [glucose] based on the results of in vitro tests (electrode substrate sensitivity corrected for 37°C).

Although a within-subject double-sensor recording (glucose/glucose-null) from the same (or symmetrical) recording areas could be viewed as the best approach to reveal a specific contribution of glucose to electrochemical currents, only one sensor was used in each recording session and current differentials were determined by analyzing mean data obtained with each type of sensor. This approach appears to be the most appropriate with the sensors of this design (relatively large size, inbuilt reference electrode), because glucose and glucose-null sensors during simultaneous use have different baseline currents and could electrically cross talk with each other, thus affecting the measurements from each sensor.

In addition to stimulus-related changes in electrochemical currents, we also analyzed changes in basal currents (determined at different time points preceding exposure to stimuli) recorded with both glucose and glucose-null sensors in vivo as well as glucose-null sensors in vitro. Comparison of these changes is important for evaluating the origins of slow, tonic current drift seen in behavioral experiments as well as for estimations of basal glucose levels in tested structures. Animal activity was quantified as the number of infrared beam brakes per minute and analyzed as mean changes for each experimental procedure.

One-way ANOVA with repeated measures [followed by Fisher least significant difference (LSD) post hoc tests] was used for evaluating statistical significance of changes in electrochemical currents and locomotion. Between-group comparisons were conducted by using either two-way ANOVA or Student's t-test where appropriate, and the latency of the glucose response was determined based on the first data point significantly different from baseline (P < 0.05, Fisher test). Standard correlation and regression analyses were used to determine the relationships between specific and nonspecific changes in electrochemical currents and between changes in glucose and other recorded parameters.

RESULTS

Slow changes in glucose electrochemical currents: specific and nonspecific contributions, basal glucose levels in NAcc and SNr.

Figure 2 shows changes in basal electrochemical currents detected by glucose and glucose-null sensors in both structures during the entire behavioral experiment as well as current values recorded by glucose-null sensors in vitro (22–23°C, for ∼8 h, the duration of the in vivo experiment). Since changes in basal currents detected by glucose-null electrode were superimposable in the NAcc and SNr, they were combined into one group.

Fig. 2.

Tonic changes (means ± SE) in electrochemical currents (nA) recorded by glucose and glucose-null sensors in vivo during behavioral experiment as well as in vitro at stable ambient temperatures. Basal electrochemical currents in both nucleus accumbens (NAcc) and substantia nigra pars reticulata (SNr) decrease gradually during ∼8-h in vivo recordings. Despite quantitative differences in current amplitude, similar gradual decreases were also seen for glucose-null sensors both in vivo and in vitro (see text for details). The decrease was exponential in each group, but within the time of behavioral tests (120–440 min; dotted lines) it was well approximated by a linear correlation (r = coefficient of correlation).

Basal currents detected by glucose sensors in vivo were consistently larger than those detected by glucose-null sensors. Since glucose-null sensors during in vivo recording are exposed to the same physical and chemical factors as glucose sensors, it is reasonable to assume that these differences are due to the contribution of glucose. Taking into account these differences (8.03 nA and 5.17 nA for NAc and SNr, respectively) and in vitro sensor sensitivity (7.62 nA/1 mM and 6.51 nA/1 mM for probes used in the NAcc and SNr, respectively) corrected for 37°C (14.84 nA and 12.69 nA), basal [glucose] is 540 μM and 407 μM in the NAcc and SNr, respectively. Since the current differences throughout the 8-h experiment became smaller (end of recording: NAcc 6.25 nA and SNr 4.12 nA) but the sensor sensitivity (determined during postrecording calibrations) remained relatively stable, it appears that basal glucose levels slightly decreased during the session, being equal to 422 μM (−22%) and 325 μM (−20%) at the session end in the NAcc and SNr, respectively. Although electrochemical currents recorded by both glucose and glucose-null sensors decrease exponentially during long-term in vivo recording (see curve for Null, in vitro, Fig. 2), within the time of our behavioral tests (140–460 min) this decrease could be adequately described as linear.

Stimulus-induced changes in NAcc [glucose].

To evaluate changes in NAcc glucose levels induced by various arousing stimuli, we compared the mean changes in electrochemical currents recorded by glucose and glucose-null sensors after exposure to each stimulus. Data were analyzed with both low (1-min bin; Fig. 3) and high (4-s bin, Fig. 4) temporal resolution. In both cases, we then determined current differentials (difference in currents detected by glucose and glucose-null sensors), which provide a more accurate measure of stimulus-induced glucose fluctuations by excluding nonspecific contributions detected by glucose-null sensors. These latter data were calibrated in micromolar concentration units.

Fig. 3.

Slow time course analysis (60-s bins) of changes in NAcc glucose levels induced by natural arousing stimuli and iv cocaine. A: relative changes in electrochemical currents (nA) recorded by glucose and glucose-null sensors after exposure to audio stimulus, tail pinch, social interaction, and iv cocaine injection. Values significantly different from baseline (Fisher test, P < 0.05) are shown as filled symbols. B: changes in glucose concentrations ([glucose], μM) determined as the difference in mean currents detected by glucose and glucose-null sensors (see methods for details). Significant changes (Student's t-test, P < 0.05) in currents are shown as filled symbols. n, Number of tests with glucose electrodes in each group. Vertical dashed lines show moments of stimulus presentation.

Fig. 4.

Rapid time course analysis (4-s bins) of changes in NAcc glucose induced by natural arousing stimuli and iv cocaine. A: relative changes in electrochemical currents (nA) recorded by glucose and glucose-null sensors after exposure to audio stimulus, tail pinch, social interaction, and iv cocaine injection. Values significantly different from baseline (Fisher test, P < 0.05) are shown as filled symbols. B: changes in [glucose] (μM) determined based on differences in mean currents detected by glucose and glucose-null sensors. Significant changes (Student's t-test, P < 0.05) in currents are shown as filled symbols. Vertical dashed lines show duration of stimulus presentation.

Low temporal resolution analysis (Fig. 3A) revealed that TP, SI, and cocaine induced significant changes in NAcc oxidation currents recorded by glucose sensors (F_10,340 = 4.09, F_11,371 = 5.61, F_10,340 = 5.42, respectively, P < 0.001), while an audio stimulus induced no changes (F_5,125 = 1.29, P = 0.19). Although electrochemical currents recorded by glucose-null sensors also slightly changed following these procedures (see Nonspecific contributions to glucose currents and their possible cause), these changes were incomparably smaller and current differentials (Fig. 3B) revealed significant and strong changes in glucose levels for all procedures expect the sound. While glucose levels rapidly increased (40–60 μM) during TP and SI and after cocaine injection, the pattern of subsequent changes differed for each stimulus. The increase induced by TP was stronger and more prolonged than that induced by SI, and it was not followed by a reboundlike decrease that appeared at ∼20 min from the start of SI. The glucose increase induced by cocaine was maintained for ∼15 min, but glucose levels decreased below baseline from ∼50 min after injection.

High temporal resolution analysis (Fig. 4A) revealed that all stimuli significantly changed oxidation currents recorded by glucose sensors (audio F_5,185 = 1.81, P < 0.02; TP F_10,349 = 2.91, P < 0.01; SI F_11,371 = 7.64, P < 0.01; cocaine F_10,340 = 5.42, P < 0.001) with minimal changes produced by glucose-null sensors. Analysis of current differentials (Fig. 4B) revealed that the audio stimulus produced a rapid, brief increase in glucose (∼30 μM or 6% over baseline) that peaked at the second data point (6 s) and was maintained for ∼20 s. TP induced a similarly rapid (0–4 s) but larger increase in glucose (60 μM or +11%), which peaked at 6–10 s after the start of procedure, was maintained within the entire duration of stimulation, and remained elevated within the entire analysis interval. While glucose levels also rapidly increased at the start of SI (peak ∼60 μM at 6–14 s), they then decreased toward prestimulus baseline but phasically increased again after the guest rat was removed from the cage. Equally rapid (latency 6 s) glucose rise also occurred during iv cocaine injection, peaking at ∼70 μM at the injection end. Then, glucose levels relatively decreased and began to increase slowly again from 30–40 s after the start of injection to the second lower peak (∼40 μM).

Stimulus-induced changes in SNr [glucose].

A similar analysis paradigm was applied to our SNr recordings (Fig. 5 and Fig. 6). After slow time course analysis (Fig. 5), we found that TP, SI, and cocaine but not sound induced significant changes in oxidation currents recorded by glucose sensors (audio F_4,104 = 0.32, P = 0.99; TP F_9,309 = 3.60, P < 0.001; SI F_9,309 = 3.19, P < 0.001; cocaine F_9,309 = 3.45, P < 0.001), with minimal changes detected by glucose-null sensors. In contrast to rapid increases seen in the NAcc, glucose levels significantly decreased (−20–40 μM or −5–10% of baseline) after the initiation of TP, SI, and cocaine injection. These decreases were slower and reached their nadirs immediately after the end of TP and SI (4 min) and at approximately the same time (4–7 min) after cocaine injection. In each case, glucose levels began to rebound after their decreases. Similar to the initial decreases, which differed in magnitude and duration, the poststimulation increases also differed in magnitude depending on the stimulus, being significantly greater versus baseline for TP and not significant for SI and cocaine. Finally, in each case, glucose levels showed a third fluctuation, a decrease below baseline from ∼30 min after stimulus. In contrast to robust current changes detected by glucose sensors, slow increases in electrochemical currents were seen with glucose-null sensors; these changes were generally similar for TP, SI, and cocaine injection (see Nonspecific contributions to glucose currents and their possible cause).

Fig. 5.

Slow time course analysis (60-s bins) of changes in SNr glucose levels induced by natural arousing stimuli and iv cocaine. A: relative changes in electrochemical currents (nA) recorded by glucose and glucose-null sensors after exposure to audio stimulus, tail pinch, social interaction, and iv cocaine injection. Values significantly different from baseline (Fisher test, P < 0.05) are shown as filled symbols. B: changes in [glucose] (μM) determined based on differences in mean currents detected by glucose and glucose-null sensors. Significant changes (Student's t-test, P < 0.05) in currents are shown as filled symbols. n, Number of tests with glucose electrodes in each group.

Fig. 6.

Rapid time course analysis (4-s bins) of changes in SNr glucose induced by natural arousing stimuli and iv cocaine. A: relative changes in electrochemical currents (nA) recorded by glucose and glucose-null sensors after exposure to audio stimulus, tail pinch, social interaction, and iv cocaine injection. Values significantly different from baseline (Fisher test, P < 0.05) are shown as filled symbols. B: changes in [glucose] (μM) determined based on differences in mean currents detected by glucose and glucose-null sensors. Significant changes (Student's t-test, P < 0.05) in currents are shown as filled symbols.

High-resolution analyses (Fig. 6) revealed that SNr glucose levels slowly decreased during TP and SI and after cocaine injection (TP F_9,309 = 1.61, P < 0.02; SI F_9,309 = 12.81, cocaine F_9,309 = 16.77, both P < 0.001) but were not changed significantly by audio stimulus. The decrease occurred with ∼30-s latencies after initiation of TP and SI and reached its nadir after the end of procedures, and then glucose levels began to increase slowly toward the prestimulus baseline. In both cases, decreases were significant versus changes recorded by glucose-null sensors and the decrease with SI was more rapid and greater (∼40 μM or −11%) than that with TP (∼20 μM or −6%). Interestingly, a transient increase in glucose was seen with all stimuli at their onset (20–30 s). When analyzed within the first 240 s from the stimulus onset, this increase did not reach the level of statistical significance. However, the immediate increase at the start of SI became significant when analyzed for the first 60 s after stimulus onset (F_9,109 = 2.26, P < 0.02). Cocaine also decreased SNr glucose levels; the decrease was more rapid but not greater than that seen with TP and SI.

Relationships between changes in NAcc and SNr glucose levels and locomotion.

As shown in Fig. 7B, TP, SI, and cocaine significantly increased locomotor activity. The increases were about the same for each stimulus, but they were more rapid and shorter for TP and SI than for cocaine. When motor activity increased, glucose levels increased rapidly in the NAcc but decreased more slowly in the SNr (Fig. 7A). While correlation between glucose levels and locomotor activity was statistically not significant in the NAcc for both TP and SI (r = 0.19 and 0.18, respectively), it was significant and strong for both these stimuli in the SNr [r = (−)0.612 and (−)0.773 for TP and SI, respectively; P < 0.01]. In contrast, cocaine-induced changes in glucose levels in both structures tightly correlated with drug-induced locomotor activation [NAcc r = 0.734 and SNr r = (−)0.805, P < 0.01].

Fig. 7.

Changes in NAcc and SNr glucose (A), locomotor activity (B), nonspecific electrochemical currents recorded by glucose-null sensors (C), and NAcc temperatures (D) induced by tail pinch (left), social interaction (center), and iv cocaine (right) in freely moving rats. Graph shows onset and offset of stimulation (vertical dashed lines) and onset of cocaine injection; basal values are shown as horizontal dotted lines. Filled symbols mark values significantly different from baseline (Fisher post hoc test, P < 0.05). Data for locomotion, nonspecific currents, and brain temperature are shown with SE. Dashed line in C shows the negative trend in baseline recorded by glucose-null sensors.

Nonspecific contributions to glucose currents and their possible cause.

As shown in Fig. 7C, TP, SI, and cocaine significantly increased electrochemical currents recorded by glucose-null sensors (F_6,216 = 2.94, 3.68, and 1.79, respectively; each at least P < 0.05). Although these current changes were much weaker than those detected by glucose sensors (∼0.1 nA vs. 0.6–0.8 nA; see Figs. 3 and 5), they began to increase during TP and SI and after cocaine injection, peaked at 10–15 min, and slowly decreased afterwards. Despite the up-down pattern of the recorded currents, adjustment for a tonic decrease in baseline currents detected by glucose-null electrodes in vitro (dashed lines in Fig. 7C) revealed that they are clearly monophasic. Importantly, these changes were delayed from more rapid fluctuations in glucose levels (Fig. 7A) and rapid increases in locomotion (Fig. 7B). However, these nonspecific current changes were tightly related to changes in NAcc temperature [Fig. 7D; original data were obtained previously (Brown and Kiyatkin 2005; Kiyatkin et al. 2002) and reanalyzed for this study], which was also slowly increased by TP, SI, and cocaine. A tight correlation between these parameters (TP r = 0.849, SI r = 0.799, cocaine r = 0.942; each P < 0.001) and proportional changes in electrochemical currents with temperature increases found in vitro (∼0.08 nA/0.5°C; see methods) suggest brain temperature increases as a major nonspecific contributor to changes in electrochemical currents induced by arousing stimuli. In fact, this correlation is even stronger if changes in electrochemical currents are corrected for slow decreases in baseline (dashed lines in Fig. 7C), which were also seen with glucose-null sensors in vivo at stable recording temperatures.

Finally, changes in glucose levels seen in the NAcc and SNr (Fig. 7A) were independent of each other after exposure to natural arousing stimuli [TP r = (−)0.209 and SI r = (−)0.35] but tightly correlated after cocaine injection [r = (−)0.887, P < 0.001].

Changes in NAcc and SNr [glucose] induced by rapid increases in blood glucose.

Since blood glucose levels greatly exceed those in the brain's extracellular space, rapid increases in extracellular glucose found in the NAcc after sensory stimulation could be explained by an increased transport of glucose from the peripheral blood to the extracellular compartment due to accelerated local CBF. To assess how quickly a rapid change in blood levels of glucose affects its levels in the brain, we examined dynamics of changes in [glucose] in the NAcc and SNr after its passive iv administration. The dose of injected glucose (15 mg) was calculated to increase blood glucose levels by ∼3 mM after its full distribution within the entire blood volume (30 ml). Such relatively modest hyperglycemia, well within the range occurring after food consumption (Dunn-Meynell et al. 2009), was chosen to keep blood glucose levels within natural physiological fluctuations.

As shown in Fig. 8, A and B, the increase in blood glucose levels resulted in a significant rise in extracellular [glucose] in both structures (NAcc F_6,216 = 5.16 and SNr F_4,154 = 3.23; both P < 0.001). When calculated with 1-min bins, glucose levels grew from the second minute after the injection start, peaked at ∼7.5 min (NAcc) and ∼6 min (SNr), and then slowly decreased toward baseline, being increased for 19 and 14 min after injection, respectively. While the response pattern was similar in both structures, the increase in the NAcc was larger in amplitude (∼100 μM) than that in the SNr (∼65 μM). Since glucose levels in the NAcc are slightly higher than in the SNr (540 vs. 410 μM, see above), the relative change was about the same in both structures (NAcc +18% and SNr +16%). Rapid time course analysis (Fig. 8, C and D; NAcc F_6,216 = 14.30, SNr F_4,154 = 4.95; both P < 0.001) revealed that glucose levels in both structures began to increase slowly within the injection duration (latency to significant change 14 s and 62 s for NAcc and SNr, respectively), showing a small peak at 80–100 s after the injection start (a presumed direct contribution of blood glucose due to sensor-induced weakening of the blood-brain barrier), and then continued to increase slowly to the second, large peak at much later times. Assuming that our glucose injections induce an ∼3 mM increase in blood levels, it appears that extracellular brain glucose levels are relatively resistant to physiological blood hyperglycemia being increased only within 2–3% (or ∼17% with respect to extracellular brain levels) after 30–40% increase in blood levels.

Fig. 8.

Slow (A and B) and rapid (C and D) changes in [glucose] (μM) in NAcc (A and C) and SNr (B and D) after iv injection of glucose (15 mg) in freely moving rats. Values significantly different from preinjection baseline (Fisher post hoc test, P < 0.05) are shown as filled symbols. First dashed vertical line at 0 min shows the start of iv glucose injection, and second dotted line marks concentration peaks. Arrows in C and D show the first, weakly defined, brain glucose peak at 80–100 s that could reflect a minor contribution of strong rise in blood glucose after its iv injection.

Changes in NAcc [glucose] after direct neuronal activation induced by local glutamate microinjections.

Excitation, a typical response of NAcc neurons to various environmental challenges, could be a possible cause for rapid increases in extracellular glucose levels found in this structure. This mechanism was tested by examining changes in this parameter during direct neuronal activation induced by local microinjection of GLU.

Several original examples of changes in glucose current obtained in this experiment are shown in Fig. 9A. We found that local application of GLU consistently increased glucose levels in each of three experiments; these increases occurred with variable latencies and were transient, and their magnitude was not larger than that seen with natural stimuli. At low GLU doses (1–5 mM, 0.2 μl), increases in glucose currents were smaller in magnitude and typically followed by their decreases below baseline (Fig. 9A, see rats M35-1 and M34-1); these responses increased in magnitude and duration with subsequent injections of GLU at higher doses (0.5 μl). These GLU-induced increases were similar in their parameters to the increases occurring naturally during spontaneous movement activity (Fig. 9A, see rats M34-2 and M37-5). Typically, the increases were weak with the first GLU injection, grew at the second and third injections, and decreased or disappeared with subsequent injections. At these later tests, we also observed two cases of phasic decreases (Fig. 9A, rat M35-3), which were seen when the rat was behaviorally active before the start of GLU injection and glucose levels were already increased. GLU tests did not induce apparent changes in animal behavior. Interestingly, in most GLU tests we also observed transient glucose peaks at the start of the injection (see top 3 examples in Fig. 9A); these increases tightly correlated with the sound of the click when the pump was switched on. An averaged response for all cases of initial GLU applications (injections 1–3 for each session; see Fig. 10A), revealed that GLU levels began to increase during the injection time, peaked at ∼120 s (∼100 μM, close to the phasic peak seen with arousing stimuli), and decreased toward the preinjection baseline for the next 8 min (F_6,432 = 4.51, P < 0.001). Interestingly, an averaged response also revealed a phasic, sound-related rise in glucose when the pump was switched on. Although this increase was nonsignificant within the entire 4-min analysis interval, it was highly significant within the first 12 s from the start of microinjection (F_6,41 = 5.03, P < 0.001).

Fig. 9.

Original examples of changes in NAcc glucose electrochemical currents (nA) detected in NAcc (A) and SNr (B) after local glutamate (GLU; A) and procaine (PRO; B) microinjections near the glucose sensing area. Each graph represents an original record of electrochemical currents (1-s time resolution) before and after GLU or PRO microinjections (2 vertical dashed lines mark the injection duration). Concentration of GLU or PRO, injected volumes, and number of injections within a session (second number after the dash) are shown in each graph. Solid horizontal lines below electrochemical records (rats M34 and M37) show periods of motor activity. Other explanations are in the text.

Fig. 10.

Mean changes in extracellular glucose levels (μM) after intra-NAcc microinjections of GLU (A), intra-SNr microinjections of PRO (B), and intraperitoneal injections of Equithesin (C). Two dashed vertical lines in each graph show the timing of injections. Horizontal dashed lines at 0 μM show basal, prestimulation values. Values significantly different from baseline [Fisher least significant difference (LSD) post hoc test, at least P < 0.05] are shown as filled symbols.

Changes in SNr [glucose] after direct neuronal inhibition induced by local procaine microinjections.

Transient decreases in SNr extracellular glucose levels could be related to phasic decreases in neuronal activity typically occurring in this structure after salient sensory stimulation and in association with movement activation. To test this possible mechanism, we examined changes in SNr glucose currents after intra-SNr microinjections of PRO.

We found that PRO injection consistently decreased SNr glucose currents (see original examples in Fig. 9B). Similar to the GLU-induced responses, PRO-induced decreases in glucose current were highly variable in their latency, magnitude, and duration, depending upon the rat, the concentration and dose of procaine, the injection number within a session, and the animal's activity state/glucose levels at the time of the test. Typically, the decreases, especially those induced by PRO at low doses (see top example in Fig. 9B), were relatively small in magnitude (0.2–0.8 nA or 20–80 μM) and followed by subsequent glucose increases above baseline. This down-up response pattern was typically maintained with larger doses of PRO, but the durations of both decreases and subsequent increases were more prolonged. In some rats, intra-SNr PRO injections at larger doses induced contralateral rotational behavior, which was evident when glucose levels were at lower levels but disappeared when they increased.

When all tests were averaged (n = 11), we found that PRO has a significant, strong effect on SNr glucose levels (F_10,340 = 14.87, P < 0.001; see Fig. 10B). These levels decreased rapidly within the microinjection duration (latency 10 s) and reached plateau for the first time at ∼40–50 μM (2–4 min) and for the second time at 70–80 μM (7–8 min). This second decrease in mean values originated from a few cases of PRO injections at larger doses, when a down-up glucose response pattern was eventually transformed into glucose decrease (see bottom example in Fig. 9B).

Brain glucose response induced by a general anesthetic drug.

In contrast to the relatively small-magnitude physiological fluctuations in extracellular glucose found in the NAcc and SNr, general anesthesia induced by a mixture of chloral hydrate and pentobarbital induced a powerful increase in [glucose] in both structures (Fig. 10C; NAcc F_2,92 = 12.50 and SNr F_2,92 = 46.24, both P < 0.001). This increase began with 2–4 min onset latencies, peaked at ∼20–25 min, and slowly decreased toward baseline afterward. The increase was strong in both structures, but it was about twofold greater in the SNr (∼388 μM or +95% of baseline) than in the NAcc (∼227 μM or +42% of baseline). Although this robust rise in glucose is consistent with previous microdialysis data obtained in the hippocampus and striatum with different anesthetic drugs (Canal et al. 2005; Fellows et al. 1992; Osborne et al. 1997), its magnitude clearly exceeds much weaker, structure-specific glucose fluctuations elicited by natural arousing stimuli.

DISCUSSION

The primary goal of this study was to examine how extracellular glucose levels fluctuate in response to natural arousing stimuli and iv cocaine administration in the awake, freely moving rat and to understand their underlying mechanisms. In contrast to microdialysis, which allows glucose levels to be directly sampled from brain tissue typically on a scale of 5–10 min, high-speed amperometry with glucose-selective sensors has a second-scale resolution, making it possible to evaluate glucose fluctuations at the temporal scale comparable to neuronal activity. To further examine the possible role of neural activity in determining the glucose response, measurements were conducted in two brain structures that have profound differences in basal neuronal activity and responsiveness. Finally, we directly tested the role of neuronal activity in mediating rapid and differential changes in extracellular glucose by measuring glucose levels after local drug-induced neuronal excitation and inhibition.

Reliability of electrochemical evaluation of brain glucose levels in behaving animals.

Although electrochemistry, because of its excellent temporal resolution, appears to be a better tool than microdialysis to detect real-time fluctuations in brain neurochemicals, the specificity of such measurements is always problematic. Enzyme-based glucose microsensors that allow such measurements were developed in the 1990s (Hu and Wilson 1997; Lowry et al. 1998; Netchiporouk et al. 1996; see Wang 2008 for review), but their use in behavioral experiments has been limited and data are controversial. The glucose sensors (Pinnacle Technology) used in this study were quite substrate sensitive in vitro, producing much larger currents (10–15 nA/1 mM at 37°C) than other oxidizable neurochemical substances at their basal or response range concentrations (i.e., ascorbate ∼0.5 nA/250 μM; dopamine 0.05 nA/100 μM). Taking into account that glucose levels in the brain's extracellular space are much higher (500–1,500 μM; Dunn-Meynell et al. 2009; Fellows and Boutelle 1993; McNay et al. 2001) than those of other possible chemical substances (i.e., GLU ∼0.4–4.0 μM, dopamine ∼5–50 nM), high sensor sensitivity and selectivity is an important basis for assuming that current changes detected by these sensors in vivo reflect fluctuations in extracellular glucose levels.

To control for measurement selectivity, our recordings with glucose sensors were compared with those obtained with glucose-null sensors that are exactly the same in their design but have no glucose oxidase enzyme. Since these sensors are equally sensitive to everything (i.e., chemical interferents, pH, oxygen, temperature, etc.) except glucose, their use in vivo provides the best possible tool to eliminate the influence of nonspecific biological and physical changes affecting glucose currents. As shown in this study, basal currents detected in both structures by glucose sensors slowly decreased during the in vivo experiment (see Fig. 2). However, a similar decrease in current baseline occurred with glucose-null sensors. After correcting for this nonspecific effect by subtracting values generated by glucose-null sensors, we found that basal extracellular glucose levels in both structures appear to be relatively stable during an ∼8-h recording interval. The use of glucose-null sensors also allowed us to evaluate basal levels of glucose, which were determined to be ∼540 μM and ∼407 μM in the NAcc and SNr, respectively. While these values appear to be lower than the early estimates, a similar range of “low” values has been determined in the striatum and hippocampus with zero-net flux microdialysis (0.35–0.71 mM; Fellows et al. 1992; Lowry et al. 1998; McNay et al. 2001). These quantitative data, however, should be considered with certain caution because our glucose sensors are equipped with a polymer layer that limits glucose diffusion to the active surface and possibly undervalues its basal concentrations. However, this factor equally affects stimulus-induced glucose responses, making it possible to evaluate accurately the range (i.e., relative change) of physiological fluctuations of glucose levels in brain tissue.

Finally, we confirmed an important role of naturally occurring brain temperature fluctuations as a factor affecting electrochemical measurements in awake, behaving animals. As shown in this study, TP, SI, and cocaine induced consistent increases in electrochemical currents (∼0.1 nA) detected by glucose-null sensors, and these slow changes tightly correlated with NAcc temperature increases (0.6–0.8°C) induced by these stimuli (see Fig. 7, C and D). These current increases, moreover, were about the same as those induced in vitro by changing temperature, suggesting a primary contribution of this biological variable in their generation. However, this temperature contribution was not significant for stimulus-induced glucose responses because of relatively large brain [glucose] and large-magnitude changes in oxidation currents. In contrast, it could be a serious complicating factor for other neuroactive substances that are present in the brain at much lower concentrations (i.e., GLU), generating smaller fluctuations in oxidation currents (see Wakabayashi and Kiyatkin 2012).

Rapid, structure-specific changes in brain glucose levels induced by arousing stimuli and cocaine.

Extracellular glucose levels change rapidly and differently in the NAcc and the SNr. In the NAcc, glucose levels phasically increased after an audio stimulus, at the start of TP and SI, and within the duration of iv cocaine injection. While equally rapid (latency 2–6 s), the increase was weaker and transient with an audio stimulus but stronger and much more prolonged with SI, TP, and cocaine. Although glucose levels after the initial peak relatively decreased when the recorded rat actively interacted with the guest, a second, smaller phasic increase occurred when the guest was removed from the cage and the recorded rat showed active exploratory behavior. Glucose levels also rapidly peaked at the start of TP and declined relatively when the rat was engaged in chewing behavior but were maintained at increased levels for ∼12 min. A biphasic NAcc glucose increase was even more evident with cocaine, which induced a rapid, injection-related rise that was followed by a slowly developing second increase. A weak, reboundlike decrease in glucose levels was found from ∼20 min after SI and ∼40 min after cocaine injection, but this effect was not seen with TP within the 40-min analysis interval. While highly significant, these stimulus-induced glucose increases were relatively small in magnitude, within 6–13% of basal levels. Interestingly, extracellular glucose levels in the NAcc were also relatively resistant to large (∼3 mM) passive increases in blood glucose levels, rising by 100 μM or <20% of baseline. In contrast to relatively small physiological fluctuations, NAcc glucose levels strongly increased during general anesthesia induced by a mixture of chloral hydrate and pentobarbital (227 μM or +42% of baseline).

Glucose fluctuations in the SNr showed a different, more complex pattern. While an audio stimulus did not change glucose levels, these levels tonically decreased during TP and SI, when the rat was engaged in chewing behavior or actively interacting with the guest, reaching nadirs after the end of both procedures. Then glucose levels rebounded for ∼20 min; this effect was especially evident with TP, which induced a weaker initial decrease. A similar decrease followed by a less pronounced increase also occurred after iv cocaine, when the rats were engaged into locomotor activation. In all these cases, glucose levels negatively correlated with locomotion, with the tightest correlation for cocaine that induced the most robust motor activation. Similar to the NAcc, these physiological fluctuations in the SNr were within 5–10% of baseline, with a slightly larger elevation (∼18%) following iv glucose injection. Despite down-up fluctuations in response to natural stimuli, SNr glucose levels strongly increased during general anesthesia (388 μM or +95% of baseline), with twice more increase than that seen in the NAcc.

These graded glucose responses also correlated with the known physiological effects induced by these procedures. The increase was weak and short with an audio stimulus that induced rapid transient cortical EEG desynchronization and weak EMG activation with no locomotor response (Kiyatkin and Smirnov 2010), and it was much stronger and more prolonged with TP, SI, and cocaine, all of which induced powerful and prolonged motor activation, strongly elevated brain and muscle temperatures, and decreased skin temperature (Brown and Kiyatkin 2005; Kiyatkin et al. 2002).

Neuronal activity as a critical factor mediating the pattern of extracellular glucose response.

It is known that most GABA-containing NAcc neurons are silent or have low, sporadic impulse activity under quiet resting conditions, showing rapid, GLU-mediated excitations after sensory stimulation and in association with motor activity (Kiyatkin and Rebec 1996; Rebec 1998; Schneider 1991). For example, a 5-s TP in freely moving rats induced strong increases in impulse activity of most accumbal cells (70–80%), with a significant elevation of their mean rate for the entire cell population (Kiyatkin and Brown 2007). The NAcc glucose levels in the present study followed the same pattern (see Fig. 11, A and B), increasing significantly within the first 4 s and peaking at 6–10 s after the start of TP. A similarly rapid but weaker and more transient glucose peak was induced by an audio stimulus, which also phasically excites most striatal and accumbal neurons (Kiyatkin and Rebec 1996). A rapid rise in NAcc glucose induced by iv cocaine was also tightly related to phasic activation of accumbal neurons (Fig. 11, C and D), which appeared with a ∼7-s latency and peaked at the end of injection duration. Therefore, neuronal activation could be a critical factor determining rapid increases in NAcc extracellular glucose levels, which occur via rapid acceleration of local CBF (Ances 2004; Fellows and Boutelle 1993; Fellows and Lowry 1998; Grubb et al. 1974; Hoge et al. 2005; Kong et al. 2004) and enhanced glucose entry from the peripheral blood, where its concentration is five to eight times higher than in brain tissue (Abi-Saab et al. 2002; Barros et al. 2005; De Vries et al. 2003; Fellows and Boutelle 1993). While the exceptionally rapid change in extracellular glucose found in this study could be viewed as surprising, similar-scale latencies were recently reported for cortical Po₂, another neural activity-CBF-regulated parameter, which began to increase at ∼2 s and peaked at 4–8 s after a brief electrical stimulation (Devor et al. 2011). Both of these effects were slightly shifted in time from more rapid CBF response that typically begins to occur with up to 300- to 500-ms onset latencies after a brief stimulation (Hirano et al. 2011; Tian et al. 2010). This delay is well explained because a certain time is always necessary for the transport of both oxygen and glucose through the blood-brain barrier into brain tissue. Although glucose could be efficiently transported through the blood-brain barrier by means of the GLUT-1 transporter (Barros et al. 2005; Qutub and Hunt 2005), its exceptionally rapid rise is still surprising, suggesting a possible direct neural modulation of glucose transport via GLUT-1 (Duelli and Kuschinsky 2001). Our control experiment with iv glucose delivery revealed that NAcc extracellular glucose levels are relatively resistant to passive increases in its blood levels, increasing more slowly and peaking at 5–7 min after injection, well after glucose peaks in blood plasma.

Fig. 11.

Rapid changes in NAcc glucose levels induced by tail pinch (A) and iv cocaine injection (C) shown on the same timescale with changes in impulse activity of NAcc neurons induced by these stimuli (B and D) in awake, freely moving rats. Data in B and D are modified from our previous publication (Kiyatkin and Brown 2007). Filled symbols in A and C show values significantly different from baseline. Changes in discharge rate are shown as imp/0.5 s (B) or imp/s (D); values significantly different from baselines are shown by bold horizontal line. n, Number of tests in the groups. Vertical dotted lines show onset and offset of tail-pinch and duration of iv injection. Note that the duration of tail pinch was 5 s in the neuronal experiment but 180 s in the present electrochemical experiment.

While simultaneous monitoring of neuronal activity, local CBF, and glucose fluctuations could be viewed as the best approach to clarify the mechanisms underlying differential glucose responses in different brain areas, because of the technical limitations of each approach this combined multiparameter recording at the same temporal and spatial resolution and under physiologically relevant conditions is not feasible at this time. Moreover, reliable evaluation of each individual parameter is a technically challenging task, and their combination inevitably decreases the precision of each measurement. To overcome these problems and directly assess the role of neuronal impulse activity in mediating the glucose response, electrochemical measurements were combined with local microinjections of GLU and PRO, two pharmacological tools to activate and inhibit neuronal activity near the glucose sensing area. NAcc neurons are known to be exceptionally sensitive to GLU, showing short-latency phasic excitations with its local application (Kiyatkin and Rebec 1999). As an endogenous substance, GLU is rapidly removed from brain tissue by reuptake, thus allowing several repeated measurement tests in the same recording session in the awake, freely moving rat. Using this approach, we were able to demonstrate that selective activation of NAcc neurons results in rapid, transient rises in local glucose levels (see Fig. 9A, Fig. 10A). Although GLU microinjection cannot mimic naturally occurring neuronal excitation, the NAcc glucose responses induced by local GLU application were comparable in their magnitude (50–150 μM) to those seen with natural arousing stimuli. Therefore, it appears that neuronal activation that either occurs naturally because of GLU release or is induced by local GLU application is able to rapidly increase glucose levels to certain limits that possibly reflect the response capacities of local CBF. Consistent with the known neuronal depolarization inactivation induced by GLU at higher iontophoretic currents (Kiyatkin and Rebec 1999), GLU-induced glucose rises became eventually inverted into decreases with either larger GLU doses or when the next GLU injections were made while the glucose levels were still elevated because of the previous drug effect.

It is more difficult to explain the mechanisms mediating the more complex, multiphasic fluctuations in glucose levels found in the SNr. However, the available data on functional properties of SNr neurons as well as the results of our tests with local PRO microinjections also suggest neuronal activity as a critical regulating factor. In contrast to silent or sporadically active NAcc neurons, large-somata, GABA-containing SNr neural cells are autoactive, having high rates of basal impulse activity in awake, quiet resting conditions (Richards et al. 1997; Waszcak and Walters 1983; Windels and Kiyatkin 2004). In contrast to predominantly excitatory responses seen in accumbal cells, SNr neurons are typically inhibited by sensory stimuli in association with movement activation (DeLong et al. 1984; Schultz 1986; Schwartz et al. 1984) due to a powerful increase in GABA input from the active striatum and globus pallidus (Deniau et al. 2007). Although direct data on impulse activity response of SNr neurons to TP, SI, and cocaine injection are currently absent, the initial, transient decreases in SNr glucose levels elicited by these procedures could be related to transient decreases in impulse activity of SNr neurons. In contrast to the rapid, afferent input-related glucose increases in the NAcc, these decreases occurred with longer latencies and were correlated with robust increases in motor activity evoked by each of these stimuli.

The role of neuronal activity as a regulating factor was confirmed by local microinjections of PRO that resulted in rapid, relatively modest (20–80 μM or 5–10% below baseline), and transient decreases in SNr glucose levels (see Fig. 9 and Fig. 10B). Although all these data point to a relative decrease in local CBF and diminished entry of glucose in brain tissue as a possible cause for transient decreases in glucose levels induced in the SNr by strong arousing stimuli, this mechanism needs to be clarified by further experiments. Importantly, these decreases in SNr glucose levels were of smaller amplitudes, much slower, and more transient than the increases seen in the NAcc, and they were always followed by reboundlike increases, which were especially evident with TP. This down-up response pattern is not surprising in the light of known reboundlike excitations of SNr neurons consistently seen after both naturally occurring and GABA-induced inhibitions (Windels and Kiyatkin 2004). Therefore, despite a transient drop in glucose levels, which could be related to movement-related inhibition of neuronal activity, these levels quickly rebound because of a possible postinhibitory increase in activity of SNr cells and subsequent enhancement of local CBF. Such a mechanism is supported by our tests with PRO, which at lower doses after the initial fall in glucose levels induced a powerful, tonic glucose rise (see Fig. 9).

Interestingly, transient, weak increases in extracellular glucose levels (5–10 μM or ∼2% of baseline) were revealed in the SNr at the start of all natural stimuli. Although this weak effect was evident only with high-speed resolution analysis and in most cases it did not reach statistical significance, it is possible that a transient glucose peaklet is the earliest SNr response to sudden changes in afferent input triggered by all stimuli. Typically, the rats posed immediately after the start of SI and TP before engaging in active interaction with the guest or active chewing of the clothespin. The initial glucose peaklet was absent after cocaine injection, which always induced an immediate motor activation. In this case, decrease in glucose levels occurred with much shorter latencies than with all natural stimuli.

Conclusions and functional implications.

Use of enzyme-based glucose sensors combined with high-speed amperometry revealed that basal extracellular glucose levels in awake, freely moving rats are maintained at relatively stable levels (∼500 μM) but fluctuate continuously during normal physiological activity and are changed with second-scale latencies after natural arousing stimulation and iv cocaine administration. The magnitude and duration of these physiological changes depend upon the nature of the stimulus and its arousing potential, and they are structure specific, multiphasic, but relatively small in magnitude (5–10% of baseline). Extracellular glucose levels are relatively resistant to the large passive increases in blood glucose levels but strongly increased during general anesthesia. Since the pattern of glucose response drastically differed in the NAcc and the SNr, two structures with different neuronal activity and responsiveness to afferent challenges, it appears that neuronal activity could be a critical factor determining the pattern of changes in extracellular glucose levels. While a combined, multiparameter evaluation of glucose levels, neuronal activity, and local CBF in behaving animals remains a future challenge, the critical role of neuronal activity in mediating the glucose response was confirmed by our microinjection experiments. Extracellular glucose levels in the NAcc rapidly, transiently increased after local applications of GLU near the recording site, while a transient drop in SNr glucose levels occurred during neuronal inhibition induced by PRO. Importantly, the magnitude of these drug-induced changes was comparable to those induced by natural stimuli.

Hence, it appears that in contrast to relatively uniform metabolism-related decreases in glucose levels in brain cells (Sokoloff 1992), extracellular glucose levels are regulated primarily by neuronal activity via rapid modulation of local CBF. Because of structure-specific alterations in neuronal activity and local CBF, physiological fluctuations in extracellular glucose levels differ in different brain locations, reflecting redistribution of available energetic resources among the areas with different activity and metabolic requirements. Therefore, structure-specific change in neural activity is not only the cause of multiple physiological and behavioral responses induced by arousing stimuli but also a factor inducing, via rapid modulation of CBF, efficient delivery of oxygen and glucose to active brain cells to organize and maintain these responses. Via this basic mechanism, the brain is able to anticipate its future metabolic needs and provide necessary energetic resources to functionally active brain areas.

GRANTS

This study was supported by the Intramural Research Program of the National Institute on Drug Abuse.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: E.A.K. conception and design of research; E.A.K. and M.L. performed experiments; E.A.K. and M.L. analyzed data; E.A.K. and M.L. interpreted results of experiments; E.A.K. and M.L. prepared figures; E.A.K. drafted manuscript; E.A.K. and M.L. edited and revised manuscript; E.A.K. and M.L. approved final version of manuscript.