- Email: [email protected]
Chapter 20 A Turbo Pascal Program for On-Line Analysis of Spontaneous Neuronal Unit Activity
E.J. Karjalainen (Editor), Scientific Computing and Automation (Europe) 1990 1990 Elsevier Science Publishers B.V., Amsterdam
231
CHAPTER 20
A Turbo Pascal Program for On-line Analysis of Spontaneous Neuronal Unit Activity L. GaAl and P. Molnir Pharmacological Research Centre, Chemical Works of Gedeon Richter Ltd., 1475 Budapest, P.O. Box 27, Hungary
1. Introduction Extracellular unit activity measurement is a widely used method in neuroscience and especially in pharmacological research. A large number of papers proves the power of this and related methods in the research [1-4]. Albeit the extracellular unit activity measurement is one of the simplest method of microelecuophysiology, the experimenter should have extensive experience using the conventional experimental setup and way of evaluation [ 11. There are no suitable programs available for IBM PCs to provide full support in cell identification and in recording of extracellular neuronal activity together with experimental manipulation and evaluation of data. On the other hand the experimental conditions should be varied according to the needs of tasks. A data acquisition and analysis program should fulfill the above mentioned requirements as much as possible [ 5 ] . The simplicity and high capacity of a measurement and the regulations of Good Laboratory Practice (GLP) require further solutions in the pharmacological research. These reasons prompted us to create a program which meets most of our claims. The main goals in the development of the program were: - to support the cell identification, - to allow real-time visualization and analysis of neuronal activity, - to record experimental manipulation, - measurement and evaluation according to GLP, - to crcate an easy-to-use and flexible program.
2. Methods and results The program, named IMPULSE, was written in Turbo Pascal 5.5, because it is a highlevel language, supports assembler-language routines and its graphical capabilities are
238
excellent [6]. The source code is wcll structured TABLE 1 and documcnted, thus, it may be modified to suit Hardware requirements of IMPULSE particular research needs. It contains carefully 3.0. optimized assembly-language routines for the IBM PC,XT, AT Compatible critical acquisition to allow high-speed sampling. computer A new fully graphic command interface was Graphics adapter and monitor developed that can be used by single-keystroke Hard disk commands or keyboard menu selection, in addi- 640 K memory tion, it contains a sophisticatcd context-sensitive TL-1 scries acquisition systcm help available at all times. This command inter- Epson compatible printer lace provides the easy creation of different menu slructurcs containing not only menu points but fill-in form ficlds, as well. The hardware requirements of the program are summarized in Table 1. The inputs of the program arc the following: - analogue signal of the exlracellular amplifier, - TTL pulse of an event detector, - TTL pulses of any instrument (e.g., iontophorctic pump), - keyboard.
A
/,DISCRIMINATION LEVEL
POSllRlGGER SAMPLING
TRIGGER
PRE- AND POSTTRIGGER SAMPLING
Figure 1. The problem of spontaneous spike sampling. Action potentials of spontaneously active neurons arc random. The detection of a spikc is done by an cvcnt dctcctor (or window discriminator). This device sends a triggcr, whcn the input signal cxcccds the discrimination level. Using the convcntional way of sampling (i.e., sampling startcd by the trigger), only a part of a spike could bc sampled (upper right figurc). To sample the whole spike, the prctriggcr part of the signal should also be acquisited (bottom right figure).
239
Y
S t a r t spike sampling
Figure 2. Screen dump of IMPULSE menu and screen. The left windows contain (downwards) the already sampled spike, firing pattern and interspike time distribution (IHTG) of a cerebellar Purkinje neuron. The right window is the active window showing the current spike. The amplitude of the current spike and the status of sampling mode are displayed in a dataline. (See text for the details of menu).
The first technical problem to be solved was the acquisition of spontaneous spikes by sampling the full signal (involving the pre-spike signal, as well). The principle of the problem is featured in Figure 1. Our solution was the use of a ring buffer. When the user starts the sampling of a spike, the output of the AD converter is loaded continuously into the ring buffer. The signal of the evcnt detector with certain delay stops the acquisition. At this time the ring buffer contains the pre- and posttrigger part of the signal. IMPULSE can be characterized by the following functions: 1. SPIKE: shape and amplitude of spontaneous action potentials with pretrigger intervals, averaging possibility. 2. PATTERN: time-series of spikes with real amplitudes. 3. IHTG: interspike time distribution of action potentials. 4. FHTG: frequency histogram of neuronal activity. 5. CONFIGURATION: parameters of hardware environment, sampling, displaying and protocols. The first thrcc functions serve the cell identification.
240
IHTG:
Store
ZOOH
Quit
S t a r t I H T G sanpling
Figure 3. Screen dump of IMPULSE dHTG> menu and screen. (Skc legend of Fig. 2 ) The active window shows the proceeding of IHTG sampling. Neurons of the same type can be characterized by certain distribution of interspike times, thus, IHTG is very helpful in the cell identification.
SPIKE function calls a menu (Fig. 2) which contains certain instructions for spike sampling. Spikes can be sampled continuously or one by one, and could be averaged and displayed with real or normalized amplitude in a window or full screen (Zoom). The amplitude is measured for every spike. Selected spikes can be stored for further evaluation. PATTERN function serves the visualization of the firing pattern of the neuron which is a very helpful information in cell identification. Spikes are represented by vertical lines (thc length is proportional with the amplitude of the spike), the time between successive spike is represented by the distance between lines. Sampled pattern can be stored for further evaluation. IHTG function (Fig. 3) is a new power in cell identification and also in the study of neuronal activity. It measures the time between successive spikes (interspike time) and displays the currcnt distribution of that. Sampled IHTGs can be stored for further evaluation. The main line of the experiment is the registration of the activity (i.e., the firing rate) of Uie neuron studicd, thus, ETRG function is focused. The number of spikes is integrated on a ccrtain timebase (bin) and is displayed continuously in form of frequency histogram (Fig. 4).
24 1
16.
0.
a
0 Rou: 324 Euent: 38 N.Drug: 9 HTG: m C l e a r scRoll Spike P a t t e r n Quit
75
Start thc w a s u r e n e n t o f spiking actiuity
Figure 4. Screen dump of IMPULSE menu and screen. Extracellular unit activity measurement of a noradrenergic locus coeruleus neuron. Upper windows show the spike and the firing pattern sampled simultaneously with the recording of FHTG. Markers above the histogram and the lines of different patterns in the histogram indicate the onset of drug administrations. See the text for details.
Other functions can be used simultaneously (e.g., the variation of spike shape and pattern can be displayed and stored as illustrated in Fig. 4). It means that the measurement is not restricted to the recording of firing rate but the registration of spike shape and variations of firing properties are also included. Expcrimental manipulation (generally drug administration) can be marked and connected to the FHTG record by two ways: 1. Keyboard input: the number keys represent the onset of different types of drug administrations. 2. ITL pulses of any instrument: the switch on and switch off of maximum five devices (e.g., iontophoretic pump). The markers are stored together with the FHTG data. They are displayed over the FHTG graph to facilitate the exact evaluation of effects (Fig. 4). The length of FHTG recording is limited by the free memory of the computer only (e.g., if the bin is one second, more than 5 hours continuous recording is possible). The parameters of the experiment (e.g., sampling rate, time base of FHTG, etc) can be varied in a wide range and stored in a configuration file. By means of the mentioned
242
CREATED BY
MATION
DATA
BASELINE
DRUG
DRUG EFFECT
PROTOCOL
PARAMETERS
PATTERN
DATA
COMMENTS
IHTG
GRAPHS
Figure 5. Main functions of the evaluation program, EVALEXT. EVALEXT evaluates the data created by IMPULSE. Data can b e edited and divided into intervals, which are inputs of statistical analysis. The recorded spikes, patterns and IHTGs can be displayed and printed. The results are presented in a protocol according the prescription of GLP.
32.
16.
0. 0
.
25
50
75
tine D i s t . : ?:40 LOCKED INTERUOL: Nextscreen Preuscreen Svitchcursor Junptonextdrug Locked Druglist accept Quit Enter in the CURSOR node rou,
233
I
38:40
Figure 6. Screen dump of EVALEXT menu and screen. Data can be divided into intervals. Markers above the histogram and the different patterns of the histogram represent che already dcfincd intervals. The two cursors allow an interactive selection of intervals according to the particular needs of the user. Statistical comparison is performed bctwccn the selccted intervals.
24 3
variability the program makes it possibile to satisfy broad range of requirements. The storage of parameters provides the complete registration of experimental conditions allowing the reproduction of an experiment with the same conditions. Data are stored on hard disk in standard binary files. The file contains the frequcncy histogram, markcrs, shapes of certain spikes, patterns and IHTGs selected and any comments. Neuronal activity and the effects of experimental manipulations recorded (hereafter referred to as drug adminismation) are cvaluated and printed in final GLP form by an independent program, EVALEXT. The main functions of EVALEXT are illustrated in Figure 5. Data can be edited (e.g., to remove artifacts) in graphical form. Drug adminisuation can be commented and also edited in a full window editor. Data can be divided into intervals. This is done either in an automatic way (predefined length of intervals connected to the markers) or in an interactive way using graphical selection of intervals (Fig. 6). The intervals represent the different states of neuronal
Baseline
;gjp
30.1
0.21
CLO-2
74.
CLO-4
CLO-8 WH-200 WH-400
100.0
5:1
32 f f6
I
18.7 #_____________-___-______________________----------------------------------KIBSO 37.2 1 .58 *
Press a key t o r e t u r n . . . PRINT: -Graph Data dEuice Q u i t
P r i n t r e s u l t s i n a protocol form on output d w i c e
Figure 7. Protocol of an extracellular unit activity experiment. The head of the protocol contains the description of the experiment. The results of ANOVA and the detailed effects of different drugs are presented. “Mean FR”: is an average value of spike numbers per bin for h e given interval. “SEM”: is the standard error of the mean. “Pcrccnt”: values are expressed as the pcrcentage of the baseline activity (considered as 100%). “SEP”: is the standard error of the “Percent” value. The stars indicate a significant difference from the baseline value at p<0.05 level (one sample t-test). “N”: is the number of bins in the given interval.
244
activity (ix., the baseline activity, the changes in firing rate caused by drug adminislration). EVALEXT performs statistical analysis (ANOVA followed by one sample t-test) of intervals (see above) and creates a protocol according to the rules of GLP (Fig. 7). The recorded spikes, patterns and IHTGs can be displayed and printed.
3. Discussion Main goals for dcveloping the program were to help the cell identification and to allow real-time analysis of neuronal activity and to record any experimental manipulation. Cell identification is a hard job, the experimenter should have extensive experience to attain it. IMPULSE provides full support in cell identification by the visualization of action potentials and firing properties of a neuron. It records and displays the firing activity and experimental manipulations. IMPULSE supports most of the experimental requirements of electrophysiology and related areas, especially pharmacology. Parameters of the experiment (e.g., sampling rate, time base of frequency histogram, etc.) can be varied in a wide range and stored in a configuration file allowing both for flexibility and standard measurements. The additional program EVALEXT performs statistical evaluation of changes in firing rate caused by experimental manipulations. Data can be edited and divided into intervals, which are the basis of statistical analysis (ANOVA followed by one sample t-test). The recorded spikes, patterns and IHTGs can be displayed and printed. Learning to use IMPULSE is very simple. It has a command interface that can be used either by single-keystroke commands or keyboard menu selection without any compromises between speed, flexibility and ease of use. In addition, IMPULSE has sophisticated context-sensitive help available at all times with the press of a key. Our laboratory has some years of good experience in using IMPULSE and the earlier versions [6-83. The Figures illustrate not only the program, but also some experimental results obtaincd by the use of IMPULSE (see the legends of Figures). The system can be useful in related electrophysiological methods (e.g., in measurement of ficld potentials and long term potentiations, etc.) as well, due to its flexible modular form.
References 1.
2. 3.
Thompson RF. Patterson MM, eds. Bioelectric recording techniques. New York: Academic Press: 1973. h e s RD. Microelectrode methods for intracellular recording and iontophoresis. London: Academic Press: 1981. Dingledine R, ed. BrainSlices. New York: Plenum Press: 1984.
24 5
4.
5. 6. 7. 8. 9.
Soto E, Vega R. A Turbo Pascal program for on line spike data acquisition and analysis using a standard serial port. J Neurosci Meths 1987; 19: 61-68. Kegel DR, Sheridan RE, Lester HA. Software for electrophysiological experiments with a personal computer. J Neurosci Meths 1985; 12: 317-330. Turbo Pascal Reference Guide, 1989. Gail L, Gro6 D, Pilosi 8. Dopamine agonists reverse temporarily the impulse flow blocking effect of gamma-hydroxy-butyric acid on nigral dopaminergic neurons. Neuroscience 1987; 22 (Suppl): 84. Gail L, Grod D, Pilosi 8. Effects of RGH-6141. a new ergot derivative on the nigral dopamine neurons in the rat. Pharmacol Res C o r n 1988; 20 (Suppl 1): 33-34. Gail L, Molnir P. Effects of vinpocetine on noradrenergic neurons in rat locus coeruleus. Eur J Pharmucoll990; (submitted).
231
CHAPTER 20
A Turbo Pascal Program for On-line Analysis of Spontaneous Neuronal Unit Activity L. GaAl and P. Molnir Pharmacological Research Centre, Chemical Works of Gedeon Richter Ltd., 1475 Budapest, P.O. Box 27, Hungary
1. Introduction Extracellular unit activity measurement is a widely used method in neuroscience and especially in pharmacological research. A large number of papers proves the power of this and related methods in the research [1-4]. Albeit the extracellular unit activity measurement is one of the simplest method of microelecuophysiology, the experimenter should have extensive experience using the conventional experimental setup and way of evaluation [ 11. There are no suitable programs available for IBM PCs to provide full support in cell identification and in recording of extracellular neuronal activity together with experimental manipulation and evaluation of data. On the other hand the experimental conditions should be varied according to the needs of tasks. A data acquisition and analysis program should fulfill the above mentioned requirements as much as possible [ 5 ] . The simplicity and high capacity of a measurement and the regulations of Good Laboratory Practice (GLP) require further solutions in the pharmacological research. These reasons prompted us to create a program which meets most of our claims. The main goals in the development of the program were: - to support the cell identification, - to allow real-time visualization and analysis of neuronal activity, - to record experimental manipulation, - measurement and evaluation according to GLP, - to crcate an easy-to-use and flexible program.
2. Methods and results The program, named IMPULSE, was written in Turbo Pascal 5.5, because it is a highlevel language, supports assembler-language routines and its graphical capabilities are
238
excellent [6]. The source code is wcll structured TABLE 1 and documcnted, thus, it may be modified to suit Hardware requirements of IMPULSE particular research needs. It contains carefully 3.0. optimized assembly-language routines for the IBM PC,XT, AT Compatible critical acquisition to allow high-speed sampling. computer A new fully graphic command interface was Graphics adapter and monitor developed that can be used by single-keystroke Hard disk commands or keyboard menu selection, in addi- 640 K memory tion, it contains a sophisticatcd context-sensitive TL-1 scries acquisition systcm help available at all times. This command inter- Epson compatible printer lace provides the easy creation of different menu slructurcs containing not only menu points but fill-in form ficlds, as well. The hardware requirements of the program are summarized in Table 1. The inputs of the program arc the following: - analogue signal of the exlracellular amplifier, - TTL pulse of an event detector, - TTL pulses of any instrument (e.g., iontophorctic pump), - keyboard.
A
/,DISCRIMINATION LEVEL
POSllRlGGER SAMPLING
TRIGGER
PRE- AND POSTTRIGGER SAMPLING
Figure 1. The problem of spontaneous spike sampling. Action potentials of spontaneously active neurons arc random. The detection of a spikc is done by an cvcnt dctcctor (or window discriminator). This device sends a triggcr, whcn the input signal cxcccds the discrimination level. Using the convcntional way of sampling (i.e., sampling startcd by the trigger), only a part of a spike could bc sampled (upper right figurc). To sample the whole spike, the prctriggcr part of the signal should also be acquisited (bottom right figure).
239
Y
S t a r t spike sampling
Figure 2. Screen dump of IMPULSE
The first technical problem to be solved was the acquisition of spontaneous spikes by sampling the full signal (involving the pre-spike signal, as well). The principle of the problem is featured in Figure 1. Our solution was the use of a ring buffer. When the user starts the sampling of a spike, the output of the AD converter is loaded continuously into the ring buffer. The signal of the evcnt detector with certain delay stops the acquisition. At this time the ring buffer contains the pre- and posttrigger part of the signal. IMPULSE can be characterized by the following functions: 1. SPIKE: shape and amplitude of spontaneous action potentials with pretrigger intervals, averaging possibility. 2. PATTERN: time-series of spikes with real amplitudes. 3. IHTG: interspike time distribution of action potentials. 4. FHTG: frequency histogram of neuronal activity. 5. CONFIGURATION: parameters of hardware environment, sampling, displaying and protocols. The first thrcc functions serve the cell identification.
240
IHTG:
Store
ZOOH
Quit
S t a r t I H T G sanpling
Figure 3. Screen dump of IMPULSE dHTG> menu and screen. (Skc legend of Fig. 2 ) The active window shows the proceeding of IHTG sampling. Neurons of the same type can be characterized by certain distribution of interspike times, thus, IHTG is very helpful in the cell identification.
SPIKE function calls a menu (Fig. 2) which contains certain instructions for spike sampling. Spikes can be sampled continuously or one by one, and could be averaged and displayed with real or normalized amplitude in a window or full screen (Zoom). The amplitude is measured for every spike. Selected spikes can be stored for further evaluation. PATTERN function serves the visualization of the firing pattern of the neuron which is a very helpful information in cell identification. Spikes are represented by vertical lines (thc length is proportional with the amplitude of the spike), the time between successive spike is represented by the distance between lines. Sampled pattern can be stored for further evaluation. IHTG function (Fig. 3) is a new power in cell identification and also in the study of neuronal activity. It measures the time between successive spikes (interspike time) and displays the currcnt distribution of that. Sampled IHTGs can be stored for further evaluation. The main line of the experiment is the registration of the activity (i.e., the firing rate) of Uie neuron studicd, thus, ETRG function is focused. The number of spikes is integrated on a ccrtain timebase (bin) and is displayed continuously in form of frequency histogram (Fig. 4).
24 1
16.
0.
a
0 Rou: 324 Euent: 38 N.Drug: 9 HTG: m C l e a r scRoll Spike P a t t e r n Quit
75
Start thc w a s u r e n e n t o f spiking actiuity
Figure 4. Screen dump of IMPULSE
Other functions can be used simultaneously (e.g., the variation of spike shape and pattern can be displayed and stored as illustrated in Fig. 4). It means that the measurement is not restricted to the recording of firing rate but the registration of spike shape and variations of firing properties are also included. Expcrimental manipulation (generally drug administration) can be marked and connected to the FHTG record by two ways: 1. Keyboard input: the number keys represent the onset of different types of drug administrations. 2. ITL pulses of any instrument: the switch on and switch off of maximum five devices (e.g., iontophoretic pump). The markers are stored together with the FHTG data. They are displayed over the FHTG graph to facilitate the exact evaluation of effects (Fig. 4). The length of FHTG recording is limited by the free memory of the computer only (e.g., if the bin is one second, more than 5 hours continuous recording is possible). The parameters of the experiment (e.g., sampling rate, time base of FHTG, etc) can be varied in a wide range and stored in a configuration file. By means of the mentioned
242
CREATED BY
MATION
DATA
BASELINE
DRUG
DRUG EFFECT
PROTOCOL
PARAMETERS
PATTERN
DATA
COMMENTS
IHTG
GRAPHS
Figure 5. Main functions of the evaluation program, EVALEXT. EVALEXT evaluates the data created by IMPULSE. Data can b e edited and divided into intervals, which are inputs of statistical analysis. The recorded spikes, patterns and IHTGs can be displayed and printed. The results are presented in a protocol according the prescription of GLP.
32.
16.
0. 0
.
25
50
75
tine D i s t . : ?:40 LOCKED INTERUOL: Nextscreen Preuscreen Svitchcursor Junptonextdrug Locked Druglist accept Quit Enter in the CURSOR node rou,
233
I
38:40
Figure 6. Screen dump of EVALEXT
24 3
variability the program makes it possibile to satisfy broad range of requirements. The storage of parameters provides the complete registration of experimental conditions allowing the reproduction of an experiment with the same conditions. Data are stored on hard disk in standard binary files. The file contains the frequcncy histogram, markcrs, shapes of certain spikes, patterns and IHTGs selected and any comments. Neuronal activity and the effects of experimental manipulations recorded (hereafter referred to as drug adminismation) are cvaluated and printed in final GLP form by an independent program, EVALEXT. The main functions of EVALEXT are illustrated in Figure 5. Data can be edited (e.g., to remove artifacts) in graphical form. Drug adminisuation can be commented and also edited in a full window editor. Data can be divided into intervals. This is done either in an automatic way (predefined length of intervals connected to the markers) or in an interactive way using graphical selection of intervals (Fig. 6). The intervals represent the different states of neuronal
Baseline
;gjp
30.1
0.21
CLO-2
74.
CLO-4
CLO-8 WH-200 WH-400
100.0
5:1
32 f f6
I
18.7 #_____________-___-______________________----------------------------------KIBSO 37.2 1 .58 *
Press a key t o r e t u r n . . . PRINT: -Graph Data dEuice Q u i t
P r i n t r e s u l t s i n a protocol form on output d w i c e
Figure 7. Protocol of an extracellular unit activity experiment. The head of the protocol contains the description of the experiment. The results of ANOVA and the detailed effects of different drugs are presented. “Mean FR”: is an average value of spike numbers per bin for h e given interval. “SEM”: is the standard error of the mean. “Pcrccnt”: values are expressed as the pcrcentage of the baseline activity (considered as 100%). “SEP”: is the standard error of the “Percent” value. The stars indicate a significant difference from the baseline value at p<0.05 level (one sample t-test). “N”: is the number of bins in the given interval.
244
activity (ix., the baseline activity, the changes in firing rate caused by drug adminislration). EVALEXT performs statistical analysis (ANOVA followed by one sample t-test) of intervals (see above) and creates a protocol according to the rules of GLP (Fig. 7). The recorded spikes, patterns and IHTGs can be displayed and printed.
3. Discussion Main goals for dcveloping the program were to help the cell identification and to allow real-time analysis of neuronal activity and to record any experimental manipulation. Cell identification is a hard job, the experimenter should have extensive experience to attain it. IMPULSE provides full support in cell identification by the visualization of action potentials and firing properties of a neuron. It records and displays the firing activity and experimental manipulations. IMPULSE supports most of the experimental requirements of electrophysiology and related areas, especially pharmacology. Parameters of the experiment (e.g., sampling rate, time base of frequency histogram, etc.) can be varied in a wide range and stored in a configuration file allowing both for flexibility and standard measurements. The additional program EVALEXT performs statistical evaluation of changes in firing rate caused by experimental manipulations. Data can be edited and divided into intervals, which are the basis of statistical analysis (ANOVA followed by one sample t-test). The recorded spikes, patterns and IHTGs can be displayed and printed. Learning to use IMPULSE is very simple. It has a command interface that can be used either by single-keystroke commands or keyboard menu selection without any compromises between speed, flexibility and ease of use. In addition, IMPULSE has sophisticated context-sensitive help available at all times with the press of a key. Our laboratory has some years of good experience in using IMPULSE and the earlier versions [6-83. The Figures illustrate not only the program, but also some experimental results obtaincd by the use of IMPULSE (see the legends of Figures). The system can be useful in related electrophysiological methods (e.g., in measurement of ficld potentials and long term potentiations, etc.) as well, due to its flexible modular form.
References 1.
2. 3.
Thompson RF. Patterson MM, eds. Bioelectric recording techniques. New York: Academic Press: 1973. h e s RD. Microelectrode methods for intracellular recording and iontophoresis. London: Academic Press: 1981. Dingledine R, ed. BrainSlices. New York: Plenum Press: 1984.
24 5
4.
5. 6. 7. 8. 9.
Soto E, Vega R. A Turbo Pascal program for on line spike data acquisition and analysis using a standard serial port. J Neurosci Meths 1987; 19: 61-68. Kegel DR, Sheridan RE, Lester HA. Software for electrophysiological experiments with a personal computer. J Neurosci Meths 1985; 12: 317-330. Turbo Pascal Reference Guide, 1989. Gail L, Gro6 D, Pilosi 8. Dopamine agonists reverse temporarily the impulse flow blocking effect of gamma-hydroxy-butyric acid on nigral dopaminergic neurons. Neuroscience 1987; 22 (Suppl): 84. Gail L, Grod D, Pilosi 8. Effects of RGH-6141. a new ergot derivative on the nigral dopamine neurons in the rat. Pharmacol Res C o r n 1988; 20 (Suppl 1): 33-34. Gail L, Molnir P. Effects of vinpocetine on noradrenergic neurons in rat locus coeruleus. Eur J Pharmucoll990; (submitted).