Human Subthalamic Neuron Spiking Exhibits Subtle Responses to Sedatives

• During deep brain stimulation implant surgery, microelectrode recordings are used to map the location of targeted neurons; anesthetic agents have the potential to alter neuronal activity.

• Using human microelectrode recordings from human neurons and from rat brain slices, electrophysiologic events were analyzed before and after administration of either propofol or remifentanil. Neither propofol nor remifentanil produce only clinical relevant alterations of subthalamic neuron discharge activity, suggesting minimal interference with deep brain stimulation implant surgery.

DEEP brain stimulation (DBS) has proven successful for the treatment of movement disorders in patients with Parkinson disease1–5and is gaining ground for the treatment of other brain disorders, including major depression6and obsessive-compulsive disorder.7Many centers use microelectrode recording to map the target site before implantation of the permanent stimulation electrodes. Mapping involves assessment of single-unit neuronal discharge responses to peripheral stimuli such as joint movement or tactile sensation. The use of sedatives during the microelectrode recording mapping procedure has usually been avoided because of the belief that neuronal firing patterns and/or responsiveness to peripheral stimuli could be adversely affected.8However, the anxiety and pain associated with the surgical procedure can lead to an increase in blood pressure, which leads to an increase in the risk of intracerebral hemorrhage.9,10It would be advantageous to use anxiolytic or sedative medications during DBS surgery if possible. The objective of our study was to provide a systematic characterization of the effects produced by propofol and remifentanil on discharge responses of subthalamic nucleus (STN) neurons during DBS implantation surgery. An ideal sedative for DBS implant surgery would not interfere with microelectrode mapping, i.e. , would not substantially alter STN discharge activity.

Only a few previous studies have reported the use of sedatives during DBS implantation surgery,11–13and none provided a systematic investigation of sedative effects produced on neuron discharge. A recent study of propofol effects on population activity in the STN suggested that the sedative appeared to interfere with microelectrode recording identification; however, single-unit recordings were not undertaken.14If sedatives are to be useful for DBS surgeries, then it will be essential to characterize the effects that are produced on single-unit discharge activity. The current study addressed this situation by comparing effects produced by propofol and an active control, remifentanil, on STN neurons recorded from patients undergoing DBS implantation surgery. In addition, we studied the effects produced by propofol on STN neurons recorded from rat brain slice preparations to characterize the cellular and synaptic actions that underlie effects seen on human STN neurons.

We studied 23 patients undergoing DBS electrode implant surgery for the treatment of refractory Parkinson disease, from September 2005 until August 2009, at the Stanford University Medical Center, in Palo Alto, California. All patients were fully informed of our study objectives and procedures and gave their consent to be included in the study before surgery. Our experimental protocol was approved by the Stanford University human subjects Institutional Review Board (Stanford, Santa Clara, California) and conformed to ethical guidelines of the Society for Neuroscience and American Society of Anesthesiologists. Of these 23 patients, 11 provided stable, discriminated single-unit neuron recordings that lasted for the duration of each experiment (12 to 15 min) and were included in our analysis. Two female patients and 9 males were studied. Age, weight, disease stage, and other identifying information was blinded to conform with the Health Information Portability and Accountability Act.

DBS procedures were performed using the Nexframe targeting platform (Medtronic, Inc., Minneapolis, MN) as previously described.15,16Briefly, magnetic resonance imaging was used without fiducial markers in place to minimize image distortion. On the day before surgery, skull-mounted fiducial markers were placed and volumetric computed tomography was done. These images were reconstructed using the Framelink software package (Medtronic, Inc.) and targets were chosen in the subthalamic nucleus bilaterally at 11–12 mm lateral to the midline, 5 mm inferior to the anterior commissure–posterior commissure plane, and 3 mm posterior to the midpoint. On the day of surgery, Nexframe targeting platforms were placed bilaterally, and microelectrode recording was carried out to map the target region.

As previously described,16a platinum-iridium microelectrode (D-ZAP, FHC Inc., Bowdoinham, MA) was advanced continuously starting from a point 15 mm above the expected target point. Signals were processed through an Axon Guideline 3000 system (Axon Instruments, Sunnyvale, CA) and fed to individual headphones through a Sennheiser wireless transmission system (Sennheiser, Old Lyme, CT). Each member of the surgical team was thus provided with audible feedback during electrode advancement and mapping. Recording electrode signals were amplified (10,000 times) and conditioned by filtering 0.1–10,000 Hz and with direct current offset, before being digitized at 10°K samples per second (Axon Instruments or Cambridge Electronic Design microelectrode recording systems, Cambridge, United Kingdom). Data were stored on a computer disk for subsequent analysis.

In general, one to three recording passes were made, and neuronal responses exhibiting high signal-to-noise ratio and good stability (less than 10% variation in spike frequency) were chosen for additional study. After stable unit recordings were obtained from the STN for 2 min and consistent with our Institutional Review Board approval, 11 patients undergoing STN-DBS were given bolus injections of either propofol (0.3 mg/kg) (N = 7) or remifentanil (0.8 g/kg) (N = 4), and STN action potential discharge activity was recorded for an additional 10 or more min after administration of the drug. No adverse drug effects were noted and none of the patients reported pain on injection.

After our experiments and any subsequent mapping, DBS electrodes were introduced to the target area and test stimulation was carried out to confirm therapeutic effect and ensure that there were no stimulation related side effects. After surgery, computed tomography was performed to further evaluate lead position, assess postoperative pneumocephalus, and rule out asymptomatic hemorrhage in all patients. The patients had their internal pulse generators implanted in a separate outpatient procedure the week after DBS lead placement.

Single-unit action potential discharges were discriminated using amplitude and time window thresholds set to a tolerance of ±10% of a manually predetermined spike template for each recording session, using Neuromatic software running under IgorPro (Wavemetrics, Lake Oswego, OR). This inevitably led to a loss of some data, especially for recording sites that exhibited a large cardioballistic artifact, because some unit discharges would fall below threshold with each heartbeat (see results), but at least 90% of unit discharge activity was captured for each neuron. Interspike discharge intervals were calculated from the discriminated unit data using IgorPro. Joint-interval analysis17,18was performed using the interspike interval data, also in IgorPro.

After approval by the Stanford University Institutional Review Board (Santa Clara, California), male Sprague-Dawley rats (approximately 200 gm; 21 days old) were deeply anesthetized with isoflurane, brains were removed, and 400-μm coronal slices were prepared and maintained as previously described.19The protocol adhered to the best- practice guidelines of the Society for Neuroscience, and the “Guide for the Care and Use of Laboratory Animals” from the Institute of Laboratory Animal Resources, Commission on Life Sciences, National Research Council (Washington, DC). Whole cell recordings of STN neurons were made under visual guidance. Electrode signals were amplified (1,000 times), filtered (0–20,000 Hz) and conditioned with voltage offset (Axon Instruments, MultiClamp 700A) before being digitized and recorded using IgorPro software (10°K samples/s) and stored on computer disk for subsequent analysis. Current clamp signals were analyzed using Neuromatic running under IgorPro; voltage clamp, especially spontaneous inhibitory postsynaptic currents (IPSCs), were analyzed using Minifit, also running under IgorPro.

Propofol in an emulsion (Diprivan) was obtained from AstraZeneca (Waltham, MA) for patient use during surgery. Pure propofol for brain slice studies was obtained from AstraZeneca, and stock solutions were prepared in solvent. All chemicals for physiologic solutions were reagent grade or better and obtained from Sigma Chemical Company (St. Louis, MO). Physiologic and drug containing solutions were prepared fresh before each experiment.

Drug effects were compared with control responses using ANOVA with a post-ANOVA, Newman-Keuls test, running under IgorPro. A criterion of P  less than 0.05 was taken to be representative of measures that were statistically significant. For all comparisons we used actual values, nonnormalized and not indexed to baseline. Because each patient (or brain slice) had an internal control before drug application to compare to drug effects, we are confident that the small sample sizes we used can provide meaningful comparisons in a two-tailed, repeated measures analysis.

During control recording periods, STN neurons discharged continuously at approximately 40 Hz (range = 32–67 Hz; 42 ± 9 (mean ± SD) for 11 neurons, each recorded from individual patients). Action potential wave forms were similar for all STN neurons recorded and consisted of a multiphasic, small positive, large sharp negative, large prolonged positive, and delayed negative/positive response, all lasting approximately 2.5 ms. Discharge patterns were highly variable, with mixed high and low frequency activity seen that ranged between 1.0 and 300 Hz. Often the highest frequency activity was grouped into short bursts of two to four spikes/burst.

A prominent cardioballistic artifact was evident for some recording sites (5 of 11), such as the one shown in figure 1. The 10-s long recording shown at the top of this figure (A  ¹), during the predrug recording of baseline activity (CONTROL), exhibits a cardioballistic effect producing ”burst” discharges occurring at approximately 1.0 Hz. This corresponds to the heart rate, i.e. , on each pulse the electrode moves away from the recorded cell and action potential amplitude decreases. Note that after administration of the drug (A  ²; PROPOFOL), the pulse artifact decreases secondary to the patient relaxing, with a decrease in their blood pressure. This cardioballistic artifact makes accurate discrimination of single units impossible, because some spikes will inevitably be lost to analysis when their amplitude falls below criteria (±10% peak-to-peak amplitude, ±0.5% peak to valley timing), using template matching. However, the result of this artifact is easily seen and contributes only minimally (and predictably) to subsequent analysis. It should also be noted that mean discharge rates (approximately 40 Hz) calculated using our discrimination techniques were comparable to those reported in previous studies20,21of patients with Parkinson disease and in normal monkey STN neurons.

After administration of a bolus intravenous (IV) injection of propofol (0.3 mg/kg), very little effect on discharge activity was seen (fig. 1B). When spike data were accumulated into 5-s bins for rate meter graphs, virtually no effect on discharge frequency was apparent (less than 2.0%), nor did spike discharge variability appear to change appreciably after propofol administration (fig. 1, A, B). Each rate meter graph is from separate recordings in individual patients. Thus, propofol appears to produce sedation with little alteration of STN neuronal discharge activity.

Upon closer inspection of STN action potential discharge (fig. 2), it was clear that propofol had no effect on action potential wave form characteristics; however, the highest frequency discharge activity seen in control recordings was shifted, resulting in a prolonged postdischarge refractory period. The refractory period was prolonged from approximately 4.0 ms in control to 5.5 ms in the presence of propofol (fig. 2; 4.34 ± 0.32 to 5.58 ± 0.24 ms; n = 11; P < 0.001). In addition, in most neurons (five of seven), propofol increased the background noise slightly, possibly by activating a population of as-yet uncharacterized small amplitude interneurons or by synchronizing STN activity. This is consistent with findings from a recent study using multiunit recordings from STN neurons.13 

To better characterize this subtle propofol effect on high-frequency discharge responses and to look at drug effects on the overall discharge pattern of each neuron, we used joint-interval analysis.17,18 

For joint-interval analysis, the time interval since the last discharge (leading interval) and the interval until the next action potential discharge occurred (trailing interval; fig. 3) were measured for each action potential in a 1-min recording. These measures were used to construct joint-interval analysis graphs that provide a useful way to visualize discharge activity of each neuron. This analysis is particularly good at characterizing different firing modes that are associated with different components of the circuit that control STN neuronal discharge (fig. 3A). Note the recurrent feedback inhibition from the globus pallidus to STN in the circuit contributes to the refractory period after each action potential (2.5–10 ms). In the joint-interval graphs shown in figure 3, B and Cthis short latency (short interval) discharge activity can be seen as discharges that occur within 10 ms after the previous action potential (box in lower left; 1–10 ms).

Propofol produced a marked reduction in this short-interval activity (from 48 spikes in control to 6 in propofol), without altering the overall discharge pattern of the neurons (compare fig. 3, B and C). Despite this large effect on short-interval discharge, there was no effect on the average discharge interval (dashed line connecting fig. 3, B and C) for most action potentials. Longer interval discharge activity (more than 200 ms), corresponding to the cardioballistic pulse artifact, also shifted to longer times as the heart rate slowed in this patient after propofol administration. More spikes were counted in these longer interval clusters because the pulse artifact amplitude was diminished as blood pressure also decreased after propofol administration (compare fig. 3, B and C), thus fewer spikes fell below our threshold for spike detection, because electrode movement was decreased on each heartbeat. Although we did not quantitatively assess sedation, we did ask patients if they felt better or more relaxed. All patients reported that they were more relaxed, and none of the patients lost their ability to respond to questions. We also noted a general decrease in blood pressure and heart rate after the bolus administration in each patient.

Each STN neuron we recorded had a distinct discharge pattern, evident in both control and propofol recordings; i.e. , propofol did not dramatically alter the overall discharge activity of these neurons. The joint-interval graphs shown in figure 4Ademonstrate these unique discharge patterns seen for four additional neurons, each recorded from individual patients. Note that propofol consistently reduced the short-interval discharge activity of each cell but did not alter the overall shape of each graph.

The time course for propofol's effect was evident from joint-interval graphs as the time to loss and recovery of short-interval discharge activity (fig. 4B). Propofol depressed short-interval activity within 1 min of IV administration. The peak effect was observed at approximately 2 min postinjection (123 ± 8 s; n = 11) and recovery began within 3 min postinjection. Full recovery was evident by 10 min (589 ± 64 s; n = 7 – four recordings were lost before full recovery). Also note that propofol did not alter the overall shape of the joint-interval graph for the neuron shown in figure 4B.

The propofol effect on short-interval discharge activity would be consistent with a prolongation of γ-aminobutyric acid (GABA)-mediated feedback inhibition from globus pallidus neurons (fig. 3A). The lack of effect observed on overall discharge frequency is consistent with a lack of effect on tonic GABA receptors in these STN neurons. To test whether these possibilities occur, we measured propofol effects on both tonic and synaptic GABA currents in STN neurons recorded in rat brain slices.

STN neurons were recorded using visual guidance and appeared as a homogenous population of round-shaped cells approximately 15–20 μm in diameter. They were readily identified electrophysiologically because they exhibited spontaneous discharge activity (1.0–5.0 Hz), as previously noted.22,23Propofol (2.0 μM) did not alter the discharge activity of STN neurons recorded from rat brain slices. When cells were artificially depolarized with injected current (fig. 5A), no effect on spike threshold or discharge frequency was seen. Similarly, no effect on hyperpolarizing responses to injected current was evident. We chose a propofol concentration of 2.0 μM to produce an effect site concentration of 0.4 μM at a depth of 0.2 mm (the recording depth in brain slices) based on previous experimental and diffusion modeling results.24A concentration of 0.4 μM is thought to correspond to an effect site concentration achieved by the sedative dose administered to patients.24 

Voltage clamp recordings revealed spontaneous inhibitory postsynaptic currents occurring at 4.2 ± 0.7 Hz (n = 1,640 events in 12 cells recorded from 5 slices; each slice from a different animal). These synaptic currents could be completely abolished using the GABA receptor/chloride channel blocker, picrotoxin (100 μM, fig. 5B), or the competitive GABA antagonist, bicuculline (10 μM, not shown). These currents exhibited fast rise times (0.9 ± 0.2 ms) and decay times (11.3 ± 4 ms) with amplitudes of 23.2 ± 11 pA for the inhibitory postsynaptic currents (IPSC) responses recorded from 12 cells.

Propofol did not significantly change IPSC frequency (3.9 ± 8 Hz; n = 12; P > 0.1 compared with control) nor did this sedative alter the rise time (0.9 ± 0.1 ms). Similarly, there did not appear to be any change in the need for holding currents to voltage-clamp STN neurons at −60 mV. Propofol did significantly prolong the decay time constant of IPSCs (17.8 ± 6 ms; P < 0.01 compared with control; fig. 5C). Summary data for propofol effects on resting membrane potential, membrane resistance, evoked spike discharge frequency, IPSC frequency, IPSC amplitude, IPSC decay time (τ) and holding current (I^m) have been normalized and are presented in figure 5D. Note that only effects on the decay time constant of IPSCs were significant (P < 0.01). The prolongation of IPSC decay time mapped nicely onto the increase in refractory period produced by propofol on human STN neuron discharge activity (fig. 5E), when time course was corrected for temperature differences (scale factor = 1.8; based on comparing IPSC responses recorded at 37°C vs.  22°C).25,26 

To determine whether the propofol-induced prolongation of refractory period was caused by a direct effect of this sedative on STN neuron GABA-mediated inhibitory currents, we used a non-GABA enhancing sedative, the opiate remifentanil, as an active control drug. When remifentanil was administed as an intravenous bolus (0.5 μg/kg in 5 s), similar to propofol, it also produced a rapid onset (approximately 1 min) and short-lasting period of sedation (less than 10 min). Unlike propofol, there was no change apparent for short-interval discharge activity of STN neurons (fig. 6A). In fact, either no change or a slight increase in short interval activity was noted in the four cells studied (1,880 ± 560 spikes in control, 2,150 ± 580 in remifentanil), but this change was not significant. Remifentanil did not dramatically alter the shape of joint-interval graphs, but a tendency to produce burst discharge was seen in two of the recorded cells, and this can be seen as clusters in the joint-interval graphs (fig. 6, middle graph, arrows ). Summary data for seven propofol and four remifentanil experiments are shown in figure 6B, and it is clear that the two sedatives produced significantly different effects on STN neuron short-interval discharge activity: propofol consistently reduced this activity (P  less than 0.004), whereas remifentanil did not.

This is thought to be the first study to systematically investigate sedative effects on STN neuron discharge activity recorded intraoperatively from humans. The results indicate that both propofol and remifentanil have only minimal effects on action potential discharge activity at the doses studied, i.e. , discharge frequencies changed less than 2.0% with either sedative. Our findings agree with a growing body of literature that indicates that sedation can be used in patients undergoing DBS electrode implant surgeries,1,12,13with little disruptive effect on neuronal discharge responses. Previous studies did not compare sedative effects on the same neurons before and after drug administration, so they lacked the internal control we had for each experiment. Similarly, previous work did not analyze single-unit discharge activities before and after drug administration, but focused on multiunit or local field potential responses. That said, our results agree with previous findings indicating that sedatives have little effect on STN discharge. We targeted drug levels for sedation based on STANPUMP models#,27,28and the protocol we used would be predicted to produced a peak effect site concentration of 1.34 μg/ml at 1 min, 40 s postinjection (approximately 0.4 μM), using the Schnider model and a k(e0) with a fixed time-to-peak of 1.6 min.28It should be noted that we did not administer the sedatives as an infusion, only as a bolus. This is comparable to a steady-state infusion dose of 50 μg/kg/min – a typical sedative dose in these patients. Of course, individual patients will exhibit varying degrees of sedation in response to this infusion dose, and this should only serve as an approximation for initially targeting appropriate doses for the clinical management of each individual, in any subsequent DBS surgeries. Data from the STANPUMP model closely paralleled our drug kinetic results (fig. 4B), with effect site concentrations peaking within 2 min and falling below effective levels within 10 min. Patients reported feeling sleepy, calm, and/or relaxed after the bolus injection, but none lost consciousness. Additional studies will be needed to determine how microelectrode mapping in this nucleus will be altered while using these sedatives, with voluntary or imposed joint movements driving evoked discharge; however, it is clear that spontaneous background discharge activities are not markedly disrupted. Sedation would be desirable in these patients because growing anxiety and discomfort are leading causes of concern during these surgeries and often contribute to increases in blood pressure that can increase intracranial bleeding.

STN neuron discharge patterns were quite variable, although mean discharge frequencies were very consistent. The variability can be seen in joint-interval graphs (figs. 3 and 4) comparing different neuron firing patterns, where unimodal, fanlike patterns, bimodal low- and high-frequency clusters, and multimodal discharge patterns can be seen. It was interesting that each neuron's discharge pattern remained mostly stable for the duration of each recording (10–15 min; fig. 4) even in the presence of propofol or remifentanil. It is not known whether these discharge patterns reflect the individual electrical properties of each cell, or its position within a neural circuit, or to what extent the Parkinson disease pathology contributes to these firing patterns. It was clear, however, that feedback GABA-mediated inhibition largely determines the limit for short-interval discharges, because this was consistently increased in each neuron after propofol administration (figs. 3 and 4). To our knowledge, this is the first report to demonstrate both the uniqueness and stability of STN neuron firing patterns. It will be useful to explore joint-interval analyses for comparing STN discharge patterns in animal models of Parkinson disease, as well as in control recordings of STN neurons, because this may reveal interesting discharge properties of the disease state.

Propofol produced an increase in refractory periods for STN neuron discharge, and this effect was completely consistent with a prolongation in the time course of feedback GABA-mediated inhibition from globus pallidus neurons. Drugs such as propofol have been shown to prolong GABA-mediated inhibition in a number of brain regions. In fact, this is the most consistently reported effect produced by propofol from a number of independent laboratories.19,29We confirmed that STN neurons receive GABA-mediated synaptic inhibition, using voltage clamp recordings from STN neurons in rat brain slices. We also showed that propofol prolongs the decay time constant of IPSCs recorded from these cells. Furthermore, the degree of IPSC prolongation was entirely consistent with the increase in postspike inhibition seen in human recordings (fig. 5E).

The concentration of propofol we applied to brain slices was chosen to approximate effect site concentrations thought to pertain in vivo  during sedation. We arrived at this finding by first measuring the brain concentrations achieved in rats after an anesthetic dose of propofol.24We then used both chemical measures and diffusion modeling to determine what the applied propofol concentration should be, to achieve this effect site concentration in our brain slices.24We then followed this up with physiologic measures of the ability of this applied concentration to achieve steady-state effects in our brain slices. We found that it required more than 5 h for propofol to achieve steady-state effects, because of the slow diffusion into brain slices. We found an excellent agreement between our chemical measures, model predictions, and actual physiologic measures.24That said, it should be noted as a major caveat that actual propofol effect site concentrations are not known–and have never been directly measured. Neither is it known whether an effect site concentration in rats is the same as in humans, but it is likely that the higher doses needed to sedate rats, compared with humans, are the result of both pharmacokinetic and pharmacodynamic differences between species. Our observation of the similar increase in time course of inhibition seen in both human recordings and rat brain slice responses lends support to the idea that our model of propofol kinetics appears to closely approximate real conditions.24 

Surprisingly, STN neurons do not appear to express tonic GABA receptors, because propofol did not depress current-evoked discharge in current clamp recordings from rat neurons (fig. 5A), nor were holding currents changed in voltage clamp (fig. 5D), nor was there any indication that propofol produced a general depression of human STN neuronal activity (figs. 1, 3, and 4). Propofol has been shown to potently activate tonic GABA channels in neurons from hippocampus,19,30thalamic,31and neocortical regions.32Even GABAergic inhibitory interneurons have been shown to express propofol-sensitive tonic GABA currents.30CA1 pyramidal neuron,30spinal neuron,33and neocortical neuron34spontaneous discharge activity is markedly depressed by propofol acting through enhanced tonic GABA currents. To the best of our knowledge, STN neurons are the only cells that do not have these tonic currents. Thus, some of the results from our study may not generalize to other brain regions where neurons do express tonic currents. However, the propofol-induced prolongation of GABA-mediated IPSCs would be expected to occur throughout the brain.29 

This finding raises an interesting observation–propofol's actions on STN neurons can be explained by local actions on these neurons, even though it is obvious that this sedative would affect neurons providing input to the STN, especially layer V cortical pyramidal neurons2as well as neurons downstream from the STN that project back to STN (e.g. , globus pallidus GABAergic neurons). It appears that effects on other brain regions do not markedly affect STN neurons, or that whatever effects do contribute to altered STN discharge are largely masked by the local effects produced directly on the STN cells. Another interpretation could be that compensatory responses to propofol can balance effects throughout the brain; for example, perhaps the tonic inhibition of pyramidal neurons is balanced by tonic inhibition of inhibitory interneurons such that the net excitatory drive from neocortex to STN neurons remains constant, especially for the relatively low doses used in our study.

Different chemical/pharmacologic classes of drugs produced different effects on STN action potential discharge activity. Propofol depressed short-interval discharge activity, whereas remifentanil did not. These observations are consistent with a growing body of literature indicating that depressants appear to operate through quite different mechanisms to achieve a similar degree of sedation.19,26There appear to be different pharmacologic routes to the endpoint of sedation. Additional studies will need to focus on the differences between propofol's and remifentanil's actions on STN neurons, especially regarding the burst discharge clusters we observed after administration of the opiate. Opiates have been shown to produce disinhibition by depressing GABA-mediated synaptic inputs to pyramidal neurons,35so it is possible that a similar effect is occurring on STN neurons or on neurons upstream from these cells.

The authors acknowledge the support of the residents and fellows who facilitated obtaining the informed consent of each patient and who assisted with drug administration and patient care throughout these experiments.