Improving Data Acquisition for Fast-Scan Cyclic Voltammetry

Described is an improved data acquisition system for fast-scan cyclic .... (E) A control signal determines when the PCI-1200 will be triggered to upda...
2 downloads 0 Views 114KB Size
Download PDF
Anal. Chem. 1999, 71, 3941-3947

Improving Data Acquisition for Fast-Scan Cyclic Voltammetry Darren J. Michael, Joshua D. Joseph, Michaux R. Kilpatrick, Eric R. Travis, and R. Mark Wightman*

Department of Chemistry, University of North Carolina, Chapel Hill, North Carolina 27599-3290

Described is an improved data acquisition system for fastscan cyclic voltammetry (FSCV). The system was designed to significantly diminish noise sources that were identified in previously recorded FSCV measurements for the detection of neurotransmitters. Minimized noise is necessary to observe the low concentrations of neurotransmitters that are physiologically important. The system was based on a high-speed, 16-bit AD/DA acquisition board that allowed high scan rates and better resolved the small faradaic currents which remained after background subtraction.Irregularities that occur when independent timing sources are used for generation of the voltage waveform and collection of the current can create large noise artifacts near the voltage limits during FSCV. These were eliminated by the use of a single acquisition board that generated the voltage waveform and collected the current. Noise from frequency drift of the power line was eliminated through the use of a phase-locked loop. To demonstrate the improved performance of the system, data were collected using carbon-fiber microelectrodes in a flow injection analysis system and in brain slices. This new data acquisition system performed significantly better than another system previously used in our laboratory without these features. The improved detection limits of the new system allowed clearly resolved current spikes featuring pre-release “feet” to be recorded adjacent to individual mast cells following chemical stimulation. When combined with false-color plots, the low-noise system facilitated identification of dopamine release in a freely moving animal. Carbon-fiber microelectrodes can serve as electrochemical sensors of several neurotransmitters. Detectable molecules include the catecholamines (dopamine, norepinephrine, epinephrine), 5-hydroxytryptamine (serotonin), and histamine.1 Use of the electrodes with amperometry provides high temporal resolution with very low limits of detection. Recent reports2-5 have demon* To whom correspondence should be addressed. Tel.: (919) 962-1472. Fax: (919) 962-2388. Email: [email protected]. (1) Travis, E. R.; Wightman, R. M. Ann. Rev. Biophys. Biomol. Struct. 1998, 77-103. (2) Zhou, Z.; Misler, S. Proc. Natl. Acad. Sci. U.S.A. 1995, 92, 6938-42. (3) Jaffe, E. H.; Marty, A.; Schulte, A.; Chow, R. H. J. Neurosci. 1998, 18, 354853. (4) Bruns, D.; Jahn, R. Nature (London) 1995, 377, 62-5. (5) Urena, J.; Fernandez-Chacon, R.; Benot, A. R.; Alvarez de Toledo, G.; LopezBarneo, J. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 10208-11. 10.1021/ac990491+ CCC: $18.00 Published on Web 08/17/1999

© 1999 American Chemical Society

strated amperometric detection of release from small vesicles containing only a few thousand molecules. However, amperometry provides little chemical information to identify the detected molecules. In contrast, fast-scan cyclic voltammetry (FSCV) allows identification of detected molecules as well as observation of their temporal behavior. Recently,6 FSCV was used to monitor release from small vesicles containing only 100 000 molecules. However, the limits of detection for FSCV must be improved before it will be possible to monitor even smaller events such as occur in some neurons. FSCV also has been used to monitor electrically stimulated release within isolated slices7-13 of brain tissue, electrically stimulated release in various regions of the intact brain14,15 of anesthetized animals, and naturally evoked release during behavior.16,17 The behavioral experiments clearly demonstrate the low concentrations that must be monitored by FSCV in physiologically relevant situations. With an awake animal, there is a greater amount of noise generated during recordings because the preparation is not stationary. Identification and subsequent reduction or elimination of systematic noise is essential for interpreting data from these experiments. Monitoring electroactive molecules with fast-scan cyclic voltammetry (FSCV) is accomplished with voltage scans of a few milliseconds in duration that are repeated at regular intervals. Thus, instrumentation with a high band-pass must be used. Because the concentration of the species of interest is often in the nanomolar range, high sensitivity must be achieved. At single cells, the restricted number of molecules present in the small volume of a single vesicle is quickly diluted, causing the rapid decay of the electrochemical signal. For in vivo and brain slices, (6) Kozminski, K. D.; Gutman, D. A.; Davila, V.; Sulzer, D.; Ewing, A. G. Anal. Chem. 1998, 70, 3123-30. (7) Kelly, R. S.; Wightman, R. M. Brain Res. 1987, 423, 79-87. (8) Bull, D. R.; Palij, P.; Sheehan, M. J.; Millar, J.; Stamford, J. A.; Kruk, Z. L.; Humphrey, P. P. J. Neurosci. Methods 1990, 32, 37-44. (9) Rice, M. E.; Oke, A. F.; Bradberry, C. W.; Adams, R. N. Brain Res. 1985, 340, 151-55. (10) Bunin, M. A.; Prioleau, C.; Mailman, R. B.; Wightman, R. M. Anal. Chem. 1998, 70, 1077-87. (11) O’Connor, J. J.; Kruk, Z. L. J. Neurosci. Methods 1991, 38, 25-33. (12) Mitchell, K.; Oke, A. F.; Adams, R. N. J. Neurochem. 1994, 63, 917-26. (13) Palij, P.; Stamford, J. A. Brain Res. 1992, 587, 137-46. (14) Millar, J.; Stamford, J. A.; Kruk, Z. L.; Wightman, R. M. Eur. J. Pharmacol. 1985, 109, 341-8. (15) Garris, P. A.; Wightman, R. M. Neuromethods, Vol 27: Voltammetric Methods in Brain Systems; Boulton, A. A., Baker, G. B., Adams, R. N., Eds.; Humana Press: Totowa, NJ, 1995. (16) Rebec, G. V.; Christensen, C., Jr.; Guerra, C.; Bardo, M. T. Brain Res. 1997, 776, 61-7. (17) Rebec, G. V. Alcoholism: Clin. Exp. Res. 1998, 22, 32-40.

Analytical Chemistry, Vol. 71, No. 18, September 15, 1999 3941

active removal of molecules via protein transporters (4 µM/s maximally for dopamine18) also necessitates rapid monitoring. In previous work, we have described optimization of electrodes, potentiostats, and operating procedures to achieve rapid, low-noise, high-sensitivity measurements.19,20 The requirements of high sensitivity and rapid measurements demand that every effort is made to reduce noise, especially those sources that arise as artifacts of the data acquisition system. Two noise sources have been particularly problematic for fast-scan cyclic voltammetry. The first occurs when independent timing is used for generation of the voltage waveform and collection of the current. Any timing inaccuracies will lead to large current artifacts near the potential limits of the voltage scan. This noise source is eliminated when a single data acquisition board is used to generate the voltage waveform and monitor the current. Second, line frequency noise still occurs in FSCV data despite attempts to eliminate it. Previously, the repetition rate for FSCV has been either 60 Hz or one of its harmonics. This procedure will remove any 60 Hz noise upon background subtraction. However, slight drifts in line frequency during the course of FSCV recordings cause a low frequency noise source to appear in the data. By using a phase-locked loop to control timing, it was possible to greatly diminish line frequency noise. In this work we describe an improved data acquisition system for fast-scan cyclic voltammetry. In addition to reducing systematic noise, the hardware has been selected to provide better temporal and digital resolution-essential requirements for increasing the scan rate used for FSCV.21 The new system is shown to have reduced noise relative to previously used approaches. This is demonstrated by high-resolution recording of chemical events at isolated mast cells in culture during the “foot”, a feature that precedes some exocytotic events. In addition, recordings of dopamine fluctuations in the brains of freely moving animals are shown with improved signal-to-noise ratios. EXPERIMENTAL SECTION Software. The data acquisition software was written in LabVIEW (National Instruments, Austin, TX), as was an analysis package featuring false-color representation of data. Transform (Fortner Software LLC, Sterling, VA), a multidimensional plotting program, was used to generate the plots shown herein. A nearestneighbor smoothing routine was used for some false-color images. They are labeled accordingly. The maximum rate of acquisition is 100 kHz for single-electrode experiments and 91 kHz for multiple-electrode experiments (91 kHz/no. of channels ) rate per channel). A maximum of eight channels can be collected in one experiment. Each channel may be configured with its own software-programmable ADC gain. However, there is only one voltage waveform available from the DAC for all of the recording electrodes. Data are stored as a 16-bit array. For data processing, the array is arranged such that the number of rows is equal to the number (18) Jones, S. R.; Garris, P. A.; Kilts, C. D.; Wightman, R. M. J. Neurochem. 1995, 64, 2581-9. (19) Wiedemann, D. J.; Kawagoe, K. T.; Kennedy, R. T.; Ciolkowski, E. L.; Wightman, R. M. Anal. Chem. 1991, 63, 2965-70. (20) Howell, J. O.; Kuhr, W. G.; Ensman, R. E.; Wightman, R. M. J. Electroanal. Chem. 1986, 209, 77-90. (21) Hsueh, C.; Bravo, R.; Jaramillo, A. J.; Brajter-Toth, A. Anal. Chim. Acta 1997, 349, 67-76.

3942 Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

Figure 1. Overview of data acquisition system. Each box represents one component of the data acquisition system. The PCI-MIO-16XE10, PCI-1200, and PC-TIO-10 are all National Instruments computer interface boards. The letters shown correspond to the timing events illustrated in Figure 2. See text for details.

of cyclic voltammograms collected during the recording session. If there is only one channel for collection, each column represents the current at a different voltage value in the working electrode waveform. If there are n > 1 channels configured, then every nth column represents the next voltage value in the working-electrode waveform. Each group of n consecutive columns from the beginning represents a single working electrode voltage for a set of n channels. During data collection, there are eight current-versus-time plots that can be updated. The data for these plots are taken from a single voltage, i.e., a column in the array. Because the channel displayed for each plot is software-configurable, multiple plots can be used to display multiple potential values from the same channel. The analysis portion displays cyclic voltammograms and currentversus-time plots along with false-color plots. Typically, the currentversus-time plots generated for analysis represent the average of the current over a range of potentials, i.e., the current trace is the average of a few columns from the original array. Likewise, the cyclic voltammograms are often generated by averaging several rows of data from the original array. Thus, the signal-tonoise ratio appears better for the data displayed for analysis, given the benefits of averaging. Hardware. (a) National Instruments Data Acquisition System. Hardware included three computer interface boards available from National Instruments (PCI-MIO-16XE-10, PC-TIO10, and PCI-1200), two breakout boards also available from National Instruments (SCB68 and LP-50), and four integrated circuits. The circuits were the following: a phase-locked loop (CD4046, Harris Semiconductor, Melbourne, FL), a three-channel analogue multiplexer (CD74HCT4053, Texas Instruments, Dallas, TX), an operational amplifier (LM324, Texas Instruments), and an AND gate (CD74HCT08, Texas Instruments). Figure 1 provides an overview of the essential components of the system. More detailed information is given in the Supporting Information for this article. The gray rectangles represent the computer interface boards. The phase-locked loop is connected to a low-power transformer to monitor line frequency and controls the timing of the system.

Figure 2. Simplified timing diagram. (A) The output from the phaselocked loop (shown) follows line frequency (∼60 Hz). (B) The output from the divide-by-N is typically the phase-locked loop frequency divided by 6 (∼10 Hz, as shown). (C) The output of the waveform from the PCI-MIO-16XE-10 DAC is triggered by each rising edge of the square wave from the divide-by-N. (D) A high-level gate is used to control when data acquisition occurs. Acquisition is only enabled when the gate level is high. (E) A control signal determines when the PCI-1200 will be triggered to update its DAC. (F) The output from the PCI-1200 is phased to occur between updates of the DAC on the PCI-MIO-16XE-10 to prevent cross-talk.

A simplified timing diagram is given in Figure 2. The square wave from the phase-locked loop (Figure 2A) is sent to the PCTIO-10 for software-programmable frequency division (Figure 2B). The PCI-MIO-16XE-10 generates the voltage waveform (Figure 2C) and monitors the current for the electrochemical cell through a potentiostat. During an experiment, the triangle waveform will be continually applied to the working electrode even though data are not always acquired. A gate signal (Figure 2D) is used to control when data are recorded. A similar gate signal (Figure 2E) is used to determine when the PCI-1200 will generate a second voltage waveform which is typically used for external electrical stimulation (Figure 2F). The system has been adapted to allow an animal to deliver stimulations to itself according to the intracranial self-stimulation protocol.22 Previously, we had used an external arbitrary waveform generator for this function. However, with development of an external timing circuit, it is possible to allow the same National Instruments hardware which normally generates the stimulus to continue to do so in the self-stimulation experiment. (b) LabMaster Data Acquisition System. Locally written software (VA.exe) was used to control the LabMaster (Scientific Solutions, Mentor, OH) acquisition board (12-bit, 40 kHz). Working-electrode waveforms were generated either using one of the DACs available on the board or by the potentiostat via a digital trigger generated by the LabMaster board. Electrochemistry and Microelectrodes. Microelectrodes were prepared as described elsewhere.23,24 Briefly, both microcylinder and microdisk electrodes were prepared from 5-µm and 10-µm carbon fibers (T650 and P55, respectively, Amoco Corp., Greenville, SC), respectively. Epoxy sealed the glass and carbon (22) Garris, P. A.; Kilpatrick, M.; Bunin, M. A.; Michael, D.; Walker, Q. D.; Wightman, R. M. Nature (London) 1999, 398, 67-9. (23) Kawagoe, K. T.; Zimmerman, J. B.; Wightman, R. M. J. Neurosci. Methods 1993, 48, 225-40. (24) Cahill, P. S.; Walker, Q. D.; Finnegan, J. M.; Mickelson, G. E.; Travis, E. R.; Wightman, R. M. Anal. Chem. 1996, 68, 3180-6.

fiber and was allowed to cure for 18-24 h at 100 °C and then for 3 days at 150 °C. Microdisk electrodes were subsequently beveled at 45° (K. T. Brown Type, Sutter Instrument Co., Novato, CA). In brain slice experiments, the microdisk electrodes were also coated with Nafion, using a 2.5% solution in pure 2-propanol (Aldrich, Milwaukee, WI). Otherwise, microdisk and microcylinder electrodes were soaked in isopropyl alcohol prior to use. An EI400 potentiostat (Cypress Systems, Lawrence, KS) was used for all measurements. For fast-scan cyclic voltammetry, a 300 V/s triangle waveform from a rest potential of -400 mV to 1000 mV and back to -400 mV was normally used. All voltages are relative to a silver/silver chloride reference electrode (Cypress Systems, Lawrence, KS). The repetition rate for this waveform was ∼10 Hz, i.e., either 10 Hz from a counter or the output of the phase-locked loop (line frequency) divided by six. Current was low-pass filtered using a 2 kHz Bessel filter. For FSCV at mast cells, the voltage was scanned from +0.1 to 1.4 V and back to 0.1 V at 800 V/s.25 The waveform was repeated roughly every 33 milliseconds. The current was low-pass-filtered using a Bessel filter with a 5 kHz cutoff frequency. Animals. Male Sprague Dawley rats (250-300 g, Charles River, Wilmington, MA) and male C57 black mice (20-30 g, Jackson Laboratory, Bar Harbor, Maine) were used in the experiments outlined below. All animals were housed under conditions of controlled temperature and lighting with food and water available ad libitum. Animal care was in accordance with the Guide for Care and Use of Laboratory Animals (NIH Publication 86-23) and was approved by the Institutional Animal Care and Use Committee of the University of North Carolina-Chapel Hill. Brain Slices. Brain slices from mice were prepared as previously described.7 Four hundred micrometer-thick slices were isolated and incubated in oxygenated artificial cerebrospinal fluid. A working electrode was placed in the caudate nucleus approximately 75 µm into the slice. A stimulating electrode was placed on the surface of the slice, approximately 100 µm from the working electrodes. Biphasic stimulation pulses (300 µA, 2 ms per phase) were delivered via a pair of optically isolated, constant-current stimulation units (Neurologs, Digitimer, Welwyn Garden City, England). Mast Cell Analysis. Mast cells were isolated from adult mice by peritoneal lavage.25 Mice were anesthetized and sacrificed. Five milliliters of physiological buffer (150 mM NaCl, 1.2 mM MgCl2, 5 mM glucose, 10 mM Hepes, 2 mM CaCl2, pH 7.4) was then introduced into the peritoneal cavity of each animal. The abdomen was massaged for two minutes and the buffer recovered through an incision along the midline. The solution was centrifuged at 200 g for 5 min. The supernatant was discarded, and the pellet was resuspended in media (Dulbeco’s Modified Eagle Medium/Ham’s F-12, Life Technologies, Inc., Grand Island, NY). Two-milliliter aliquots were plated on 35-mm plastic cell culture dishes (Falcon 3001, Becton Dickson, Franklin Lakes, NJ) and incubated (37 °C, 5% CO2) before analysis. Cells were used within 36 h of plating. Mast cells were readily distinguished using phase contrast microscopy. Microelectrodes were placed 1 µm from single cells. (25) Pihel, K.; Hsieh, S.; Jorgenson, J. W.; Wightman, R. M. Anal. Chem. 1995, 67, 4514-21.

Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

3943

Table 1: Comparison of Noise Levels system new system previous system a

points per CV

average noise (pA)

standard deviation

n

932 300

13.5a

2.5 0.81

6 6

20.6

Significantly different (t-test, 0.05).

To induce vesicle release, a 10-s pressure injection (8 psi, Picospritzer, General Valve Corp., Fairfield, NJ) of 0.5 µM calcium ionophore A23187 was delivered from a 10-µm (i.d.) glass micropipet positioned 30-50 µm from the cell. Freely Moving Animal Experiments. A complete description of the experimental apparatus used for freely moving animals has been published.22,26 The reference, auxiliary, and stimulation electrodes were permanently implanted in the brain while the animal was anesthetized. The animal was allowed to recover for several days and then a fresh carbon-fiber electrode was lowered into the caudate nucleus. Stimulations were generated by the experimenter who pressed an external switch to deliver a stimulus. Each stimulus consisted of a train of 24 biphasic pulses (1 millisecond per phase, 80 µA) delivered at 60 Hz. An external circuit was used to ensure that stimulus pulses did not occur during the voltage scan of the working electrode. Statistics. To compare the two systems as shown in Table 1, current traces recorded in brain slices were used. The standard deviation of a five-second baseline collection before application of the stimulus was used to quantify noise. Several traces were analyzed for each system, and statistics are shown. A t-test was used to compare the noise levels from the systems. RESULTS AND DISCUSSION Identification of Noise in FSCV Data. As shown in our earlier work,25,27 simultaneous examination of the data from a series of cyclic voltammograms allows for a qualitative evaluation of all of the electrochemically detectable changes that occur during the sampling interval. An example of such a representation is given in Figure 3. In this instance, dopamine (1 µM) was introduced into the flow injection apparatus at 5 s, and voltammograms were recorded at the outlet with a carbon-fiber microelectrode. The voltammograms were acquired according to procedures normally used in our laboratory: voltammograms were initiated by a computer-generated trigger sent to an external waveform generator, scans were repeated at 100-ms intervals, and data were recorded with a 12-bit analog-to-digital converter (ADC). The plot was generated by subtracting from the entire data set the average of all of the cyclic voltammetric data acquired in the three seconds before dopamine was introduced into the apparatus. The false-color representation of the background-subtracted voltammetric current reveals several features. The simultaneous emergence of the peak for dopamine oxidation and that for the reduction of the dopamine-o-quinone generated by the initial voltage scan is readily apparent. However, other more subtle (26) Garris, P. A.; Christensen, C., Jr.; Rebec, G. V.; Wightman, R. M. J. Neurochem. 1997, 68, 152-61. (27) Michael, D.; Travis, E. R.; Wightman, R. M. Anal. Chem. 1998, 70, 586A92A.

3944 Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

Figure 3. Flow injection analysis of 1 µM dopamine. Dopamine was introduced to a flow injection analysis system after 5 s of background current was collected. The figure illustrates the typical backgroundsubtracted current which results when using the potentiostat to generate the voltage waveform. Time varies along the x-axis and voltage along the y-axis. The starting point for the voltammogram is indicated by the asterisk. The background-subtracted current is encoded according to the false-color bar at the bottom of the figure. Select contour lines were added to highlight the oxidative and reductive current peaks. The current-versus-time trace was taken as a 100 mV average of the current near the peak oxidative potential for dopamine. The cyclic voltammogram represents the average of 10 cyclic voltammograms at the peak current response.

features are also present. Large and irregular current spikes are seen at both the beginning and the apex of the voltage scans. In addition, there are diagonal streaks in the current. (Note the upper right and lower left corners of Figure 3.) These noise sources, clearly revealed in the false-color plots, have been characterized and removed as described in the following sections. The Voltage Waveform as a Noise Source. During a single voltage scan a major source of noise in the measured current arises from noise in the applied potential. This is because the electrode double layer differentiates the applied potential. Waveforms generated by computer-controlled digital-to-analog converters (DACs) often have superimposed digital noise originating from the computer.20 For this reason, we have normally used an external waveform generator with the scan initiated by a digital pulse from the computer. When multiple subtracted scans are to be examined, however, it becomes important that the time of the voltage scan and data acquisition coincide. Slight temporal fluctuations, or jitter, in the timing between initiation of the voltage scan and the start of data acquisition can lead to artifacts. In background-subtracted cyclic voltammetry this is particularly apparent at the initial and switching potentials because this is where the largest and most rapid change in the background current occurs. This leads to the large current excursions seen in the background-subtracted current data both in Figures 3 and 4A and in data published by a number of laboratories that use FSCV.28,29 To ensure synchrony between data acquisition and waveform generation, a single computer interface board can be used to (28) Iravani, M. M.; Muscat, R.; Kruk, Z. L. Neuroscience 1996, 70, 1025-37. (29) Rice, M. E.; Richards, C. D.; Nedergaard, S.; Hounsgaard, J.; Nicholson, C.; Greenfield, S. A. Exp. Brain Res. 1994, 100, 395-406.

Figure 4. Background-subtracted currents at microcylinder electrodes. The false-color images use the same convention as outlined in the legend for Figure 3. (A) The current was collected using the potentiostat to generate the voltage waveform. (B) One computer interface board was used with a single timing source to generate the voltage waveform and collect the current. (C) As in (B) except a phase-locked loop was used to control timing.

generate the voltage waveform and acquire the data. Data obtained with a computer-generated waveform are shown in Figure 4B. The artifacts at the switching potential are absent. To remove the digital noise from the applied waveform, the waveform from the computer was generated to be four times larger than desired and then was passed through a voltage divider that reduced its amplitude 4-fold, simultaneously diminishing the computer-generated digital noise.20 In addition, it was low-pass filtered with a time constant of 120 µs. At a scan rate of 300 V/s, this filter did not appreciably distort the waveform. The time constant was lowered to 25 µs for scanning the electrode at 800 V/s. With computer generation of the waveform, the noise in each cyclic voltammogram is more uniform across all potentials. (part A vs part B, Figure 4). Line Noise. A major noise source in many electrochemical experiments is line noise and its harmonics. To minimize this, practitioners of FSCV have traditionally acquired cyclic voltammograms at a repetition rate corresponding to line frequency or one of its lower harmonics. In principle, this should make stationary, in time, any line frequency noise sources relative to the start of the voltage scan allowing removal upon background subtraction. However, while the average line frequency over a 24-h period is exact, it can vary during the day depending on the power load. This results in the low-frequency diagonal features present in Figures 3 and 4 (A/B), i.e., the interference from line noise has been aliased to a low frequency, but not zero. Note that the diagonal lines actually change direction when comparing parts A and B of Figure 4, indicating a change in line frequency.

Figure 5. Comparison of data collected with various acquisition systems. Two-pulse, 10 Hz stimulations (300 µA, 2 ms per phase) were applied to brain slices. The format for the false-color images is outlined in the legend for Figure 3. The concentration-versus-time trace was obtained by converting the average current between 550 and 650 mV to concentration via a postexperiment calibration factor. The circles indicate the start and finish of the stimulation. (A) The data were collected using Scientific Solutions hardware with VA.exe. The scan rate was 300 V/s with 300 points per cyclic voltammogram. (B) The data were collected using National Instruments hardware with LabVIEW software. The scan rate was 300 V/s with 932 points per cyclic voltammogram.

To remove this noise source, we used the output of a phaselocked loop to trigger waveform generation and data acquisition. The output from a low-current transformer was used to capture line frequency and generate a square wave of the same frequency that was divided to the desired frequency (typically about 10 Hz) via software control. This approach immobilizes the noise source that generates the diagonal bands. The noise becomes a standing wave in the data that is removed by subtraction (Figure 4C). However, the limited response time of the phase-locked loop can lead to jitter of these standing waves. Fortunately, the position of the jitter within the data set can be altered by changing the scan rate. System Comparisons. We compared our new data acquisition system to another data acquisition system that is used for FSCV. The latter employs the LabMaster ADC, DAC board for generation of waveforms for both the working electrode and stimulating electrode along with system timing. To compare the systems, cyclic voltammograms were collected from the same electrode in the same brain slice with DAC generated waveforms. A pair of electrical pulses (10 Hz, 300 µA biphasic pulses) were delivered via the stimulating electrode. At the time of the stimulus, dopamine rapidly appears at the electrode and then is restored to baseline levels by uptake. Our new data acquisition system had significantly better noise levels than the other system (Figure 5, Table 1). The threedimensional plots reveal the noise sources apparent in each of the systems. The Labmaster based system exhibits diagonal lines which slant down from left to right (Bottom, Figure 5A) that are due to the line frequency drift described above. In addition, there is much more high-frequency noise than with the newer system. Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

3945

Figure 6. Exocytosis monitored at a single murine mast cell. The format for the false-color image is consistent with the description in Figure 3. A single exocytotic event is shown following application of a secretagogue. The concentration-versus-time trace was obtained by averaging the current in a 100 mV window around the peak oxidative potential for 5-hydroxytryptamine (+550 mV) and converting to concentration via a post-experiment calibration constant. The pair of voltammograms on the left of the current spike are taken from the foot of the spike. The voltammogram on the left in this pair was generated from a single voltage scan. Next to it is the average of 10 cyclic voltammograms. The voltammograms on the right of the current spike were taken from the peak of the current spike. The voltammogram on the left in the pair is from a single voltage scan while the one on the right is the average of two. The voltage limits for all voltammograms were +100 and +1400 mV. The color image was smoothed.

Note that the difference in noise is entirely due to the hardware/ software employed; all of the collected data used the same potentiostat. The top of Figure 5 shows individual current-versus-time curves for each of the systems. The current was averaged over the identical voltage window for each of the systems (550-650 mV), so there are approximately 3× more points averaged in the current-versus-time trace for the newer system. It had roughly half the noise of the data collected with the previously available acquisition system. Interestingly, the noise levels for the new data acquisition system did not significantly change when the point density of each cyclic voltammogram was reduced from 932 to 300 points or the scan rate was increased from 300 to 410 V/s. This suggests that in the brain slice preparation the noise is not white since, in that case, the signal-to-noise ratio would improve with a greater point density. Signal Resolution. The new FSCV system has been used to monitor chemical release in several biological systems. Because of the decreased digitization limits of the ADC, it is possible to monitor very small faradaic currents which can occur during FSCV. The large background that occurs during FSCV normally limits the gain of the ADC and the current-to-voltage converter. Increasing the bit resolution of the ADC can improve the limits of detection of the system as long as system noise is small. Figure 6 shows background-subtracted current recorded at a single mast cell. During exocytosis at a mast cell, histamine and 5-hydroxytryptamine (5-HT, serotonin) are released simulta3946 Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

Figure 7. In vivo collection of data. Cyclic voltammograms were collected for 10 s while a freely moving animal was trained to deliver external stimuli to itself. All of the bar presses which occurred were delivered externally by the experimenter and are indicated by the black bars; the width indicates the duration of each stimulation train. The presentation scheme for the false-color image is outlined in Figure 3. The concentration trace above the color plot represents the average of the current between 550 and 650 mV which was converted to concentration via a postexperiment calibration factor. Several cyclic voltammograms are displayed above the trace. Each is the average of two cyclic voltammograms taken from the region indicated by the arrow. The horizontal bar in each cyclic voltammogram represents zero current. Labels for the voltage limits were -0.4 and 1.0 V for all. The scale bar for the cyclic voltammograms is shown in the top right. The color image was smoothed.

neously from vesicles.25 Oxidation of 5-HT results in a peak around +550 mV while histamine is oxidized at the scan reversal location. There is a reductive peak at the end of the reverse scan that has contributions from both analytes. Occasionally, there will be a pre-release current associated with a spike.30,31 Referred to as a foot, it is interpreted as the opening of the fusion pore between the vesicle and plasma membrane and results in a small, steady-state current. The magnitude of the foot shown in Figure 6 is about 20 pA, or 30 nM. The data were collected with a current-to-voltage amplifier gain of 2 nA/V and a ADC range of (10 V. With the 16-bit PCI-MIO-16XE-10, the above conditions corresponded to 0.610 pA/bit. Thus, the 20-pA signal corresponds to roughly 33 bit levels. Had the signal been collected with a 12-bit board, however, the signal would only correspond to two-bit levels. The foot may have been observable in the currentversus-time trace, but with much less resolution. Evaluation of Data Collected from Freely Moving Animals. When evaluating data collected using FSCV in freely moving animals, identity of the source of the signal is crucial. During the experiment, the animal is continually moving, often vigorously as in the case of bar pressing or other behavioral tasks. This adds to the noise because the electronics for the preamplifier are mounted on the head of the animal.26 Our previous results in freely moving animals contained the noise sources described above (for example, see Figure 2A of ref 22). With the improved data (30) Chow, R. H.; von Ruden, L.; Neher, E. Nature (London) 1992, 356, 60-3. (31) Wightman, R. M.; Schroeder, T. J.; Finnegan, J. M.; Ciolkowski, E. L.; Pihel, K. Biophys. J. 1995, 68, 383-90.

acquisition system (Figure 7), the diagonal features and the artifacts at the switching potentials are not present; in addition, there is an overall 2-fold reduction in noise. Traditionally, investigators using FSCV have examined the temporal changes in current over a potential range where dopamine is oxidized (line trace in Figure 7). However, when small concentration changes occur, such as in this example, it is difficult to distinguish them from noise. More information is obtained when individual cyclic voltammograms are shown for each current maximum. Examination of all of the data, in the form of color plots, provides the most information as revealed in Figure 7. During the period shown in Figure 7, several stimulations (indicated by the solid bars) were delivered to the animal by the experimenter while training the animal to bar press. The training session lasted several minutes, and 10 s of the data are presented in the figure. The stimulating electrode was positioned so that the stimulation would depolarize dopamine neurons. Backgroundsubtracted cyclic voltammograms (average of two) are shown above the current trace with arrows indicating the time they were recorded. At each of these time points the current at the potential where dopamine is oxidized is significantly larger than the baseline. However, examination of the cyclic voltammograms at each of these points reveals that many are noise artifacts. Cyclic voltammograms 1, 2, 3, and 6 are identical to those for dopamine (compare with Figure 3). Cyclic voltammograms 4 and 8 are clearly not due to dopamine since they contain no peaks, but the current is significantly different from zero (indicated by the horizontal bar). Cyclic voltammograms 5 and 7 have peaks, but are not identical to that for dopamine. For cyclic voltammogram 5, the reductive peak relative to the oxidative peak is smaller than anticipated. However, because it occurs directly after a stimulation, it is likely dopamine. Cyclic voltammogram 7 resembles dopamine but is superimposed with a broad wave between -200 and +400 mV. All of these features become more readily interpreted when the entire data set is examined via the color plot. Cyclic voltammograms 4 and 8 clearly arise from noise that occurs at all potentials and thus do not have a chemical origin. The portion of the color plot for the time when cyclic voltammogram 7 was recorded, directly after a stimulation, reveals that the change in dopamine occurs at the same time as a background change. The change may be caused by background drift or by a change in the ionic composition of the extracellular environment. The occurrence of the stimulations in this example provides a marker as to when dopamine changes should occur. However, as studies move to more subtle stimulations, e.g., food reward, novel events, etc., it will become more difficult to predict where dopamine changes should occur, making false-color images even more important for data analysis. Other investigators have used FSCV in vivo and simply reported ratios of reduction current to oxidation current as an

indicator of chemical information.32 Given the sensitivity of FSCV to changes in the ionic composition of the solution surrounding the electrode, we feel that it is essential to present at least individual cyclic voltammograms from areas of interest. Likewise, it is important for those using high-speed chronoamperometry to report ratios of reduction current to oxidation current so that some indicator of analyte is available.33,34 CONCLUSIONS Because of the small faradaic currents generated when monitoring biological samples with FSCV, it is essential that the system noise be as low as possible. The acquisition and analysis system described here has better noise characteristics and greater capabilities than a similar acquisition system we have used extensively. Although direct comparison of current-versus-time traces reveals only a factor of 2 improvement in noise, the multidimensional plots clearly reveal the improved quality of the data due to the reduction of systematic noise which does not occur uniformly throughout the data. In addition to improving the quality of FSCV data, this new data acquisition allows FSCV to be used with scan rates that were previously inaccessible with available software/hardware. The increased update rate and 16-bit resolution of the acquisition board are crucial for this increase in scan rate. The 100 kHz update rate allows the electrode to be scanned at 2000-4000 V/s with enough point resolution to accurately describe the current profile at the working electrode. The increased number of bits is necessary because large background current is generated by fast scan rates. Without sufficient bit resolution, the small faradaic currents generated by analytes will be obscured by digitization. Thus, any advantage gained by scanning faster is lost because of insufficient digital resolution. Work is continuing to identify the ideal FSCV parameters for the detection of dopamine. ACKNOWLEDGMENT The authors thank Steven Woodward (Staff Member, Chemistry Department, University of North Carolina, Chapel Hill) for his technical support and discussions. Financial support for this work was provided by N.I.H. (DA10900). SUPPORTING INFORMATION AVAILABLE Details in the form of text and a scheme are given for the National Instruments Data Acquisition System.

Received for review May 10, 1999. Accepted July 9, 1999. AC990491+ (32) Xi, Z. X.; Fuller, S. A.; Stein, E. A. J. Pharmacol. Exp. Ther. 1998, 284, 15161. (33) Blaha, C. D.; Phillips, A. G. J. Neurosci. Methods 1990, 34, 125-33. (34) Richardson, N. R.; Gratton, A. J. Neurosc. 1998, 18, 9130-8.

Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

3947