Anal. Chem. 1993, 65, 1882-1887
1802
Amperometric Monitoring of Chemical Secretions from Individual Pancreatic @-Cells Robert T. Kennedy,*$+ Lan Huang,t Mark A. Atkinson,$and Paula Dusht Department of Chemistry, University of Florida, Gainesville, Florida 3261 I, and Department
of Pathology,
Box 5-275, J . Hillis Miller Health Center, University of Florida, Gainesville, Florida 32610
The goal of the work was to develop and test an amperometric method for measuring insulin secretion from individual pancreatic @-cells. The electrode used was a carbon fiber microelectrode modified with a polynuclear ruthenium oxide/ cyanoruthenate film. The chemically modified electrode allowed anodic detection of insulin in physiological buffers with a detection limit of 0.5 pM. To measure secretion, an electrode was positioned 1 pm away from a @-cellthat had been stimulated with K+ or glucose. Recordings made from the cells consistedof a series of current spikes averaging 38 ms full width at half-height. The spikes decreased in height and increased in width as the electrode was pulled away from the cell. Spikes were only observed if a modified electrode was used and its potential was sufficient to oxidize insulin. The area under the spikes correspond to approximately 600 zmol of insulin, which is within the expected range for vesicular insulin content. Spikearea was independent of stimulation method. The results support the hypothesis that the electrode was anodically detecting a substance secreted from the cells by exocytosis. The results support, but do not prove, that insulin was the primary substance detected.
INTRODUCTION Secretion of neurotransmitters and hormones is a complex process that is central to many important biological functions. One of the most significant systems is secretion of insulin from P-cells of the pancreas. @-Cellssynthesize insulin and store it in vesicles of approximately 500-nm diameter.' When blood glucose levels are high, glucose is transported to the interior of the cell and metabolized. One or more of the metabolic products leads to depolarization of the cell, which allows Ca2+ to enter the cell through voltage-dependent protein channels. Ca2+ triggers exocytosis where vesicle membranes fuse with the cell wall and vesicle contents are subsequently expelled to the cell exterior. Many steps involved in insulin secretion, including exocytosis and the pathways that lead to depolarization of 6-cells, are not understood at the chemical level.24 Secreted insulin acts to maintain glucose levels within a narrow concentration range. In many forms of diabetes, insulin is produced and stored ~
+ Department
of Chemistry. Department of Pathology. (1) Orci, L.; Vassali, J. D.; Perrelet, A. Sci. Am. 1988, 260, 85. (2) Rajan, A. S.; Aguilar-Bryan, L.; Nelson, D. A.; Yaney, G. C.; Hsu, W. H.; Kunze, D. L.; Boyd, A. E. Diabetes Care 1990, 13, 340. (3) MacDonald, M. J. Diabetes 1990, 39, 1461. (4) Robertson, R. P.; Seaquist, E. R.; Walseth, T. F. Diabetes 1991,40, 1. f
normally, but it is not secreted p r ~ p e r l y .New ~ methods to measure insulin secretion may be useful in developing a greater understanding of the secretion mechanism and the pathology of diabetes. Spatial and temporal resolution are both important in measuring secretion. The temporal pattern of hormone release is important in evaluating dynamics of the process; therefore, high time resolution is desired. Spatial resolution sufficient to monitor secretion from single cells, cell clusters, and whole pancreas may also be useful. Measurements from single, isolated cells may be especially advantageous in determining how @-cellsinteract with each other and how the behavior of individual cells add together to generate observed in vivo insulin patterns.6s7 Available methods for measuring insulin secretion have limited temporal and spatial resolution. Insulin secretion from cell clusters and single cells has been measured using a reversed hemolytic plaque assay.6,i This immunological technique requires a development procedure and is not suitable for dynamic measurements. Dynamic measurements are usually performed by holding the tissue of interest in a flow system and collecting fractions. The amount of insulin in the fractions is determined by radioimmunoassay. Insulin secretion from as few as 150 000 @-cellshas been evaluated with 30-s time resolution using this technique.8 The objective of this research is to develop a method for the time-resolved measurement of insulin secreted from single, living @-cells. The approach is to use a microvoltammetric electrode as a sensor. Sensors of this type have been extensively used for i n vivo measurements of easily oxidized neurotransmitter^.^-^^ More recently, several groups have begun to study the possible utility of electrochemical sensors for probing the chemical environment within and around single cells.'2-18 Ewing's group has used microelectrodes to measure intracellular dopamine, oxygen, and glucose levels in individual snail neuron^.^^-^^ In another relevant example, groups led by Wightman and Neher have measured the -__________-(5) Unger, R. H. Science 1991, 251, 1200. ( 6 ) Salomon, D.; Meda, P. Exp. Cell. Res. 1986, 162, 507. ( 7 ) Bosco, D.; Orci, L.; Meda, P. Exp. Cell Res. 1989, 284, 72. (8) Longo,E. A.; Tornheim, K.; Deeney, J. T.; Varnum, B. A.;Tillotson, D.; Prentki, M.; Corkey, B. E. J . Biol. Chem. 1991, 266, 9314. (9)Millar, J.; Stamford, J. A.; Kruk, Z. L.; Wightman, R. M. Eur. J. Pharm. 1985, 109, 341. (10) Suaud-Chapnv. _ . M. F.; Mermet, C.: Gonon, F. J. Neuroscience 1990, 34, 411. (11) Marsden, C. A,; Joseph, M. H.; Kruk, Z. L.; Maidment, N. T.; O'Neill, R. D.; Schenk, J. 0.;Stamford, J . A. Neuroscience 1988,25,389. (12) Chien, J. B.; Wallingford, R. A.; Ewing, A. G. J . Neurochem. 1990, ii4, 633. (13) Lau, Y. Y.; Abe, T.; Ewing, A. G. Anal. Chem. 1992, 64, 1702. (141Abe, T.; Lau, Y. Y.; Ewing, A. G. Anal. Chem. 1992, 64, 2160. (15)Leszczyszyn, D. J.; Jankowski, J. A,; Viveros, 0. H.; Diliberto, E. d.: Near, J. A.; Wightman, R. M. J. Riol. Chem. 1990, 265, 14736. (16) Wightman, R. M.; Jankowski, J . A,; Kennedy, R. T.; Kawagoe, K. T.: Schroeder, T. J.; Leszczyszyn, D. J.;Near, J. A.; Diliberto, E. J.; Viveroa, 0. H. Proc. N a t l . Acad. Sci. U.S.A. 1991, 88, 10754. (17) Meuleumans, A,; Poulain, B.; Baux, G.; Tauc, L.; Henze, D. Anal. Chrm. 1986, 58, 2088. (181 Chow, R.; von Ruden, I,.; Neher, E. Nature 1992, 3S6, 60.
0003-2700/93/0365-1882$04.00/0 :C 1993 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 65, NO. 14, JULY 15, 1993
release of catecholamine from individual adrenal chromaffin cells in culture using a carbon fiber microelectrode placed next to the cell.15J6918 The time resolution of the method was limited only by diffusion of released catecholamine from cell to electrode. The millisecond time resolution allowed the observation that catecholamine concentration fluctuated rapidly near the cell following stimulation of secretion. The fluctuations were demonstrated to be due to exocytotic secretion of catecholamines.16 T o apply a similar approach to the measurement of insulin Secretionrequires an amperometric method to detect insulin. Although insulin has several functionalities that may be electroactive, the most studied have been its three disulfide bonds.19 One of these bonds is on the surface of the molecule and has been demonstrated to be electroactive. Another disulfide is partially shielded and reacts slower. The third bond is buried within the molecule and not readily available for reaction. Reduction of these disulfides has been studied a t mercury electrodes.19 Although potentially useful, this approach was not pursued for our application because of possible problems with interference from oxygen reduction and irreversible adsorption of insulin. Anodic detection of several disulfide compounds, including insulin, has been reported at electrodes modified with a thin film of a composite of a cyano-bridged ruthenium dimer, (CN)&unCNRum(CN)&, and a cationic, polynuclear,mixedvalent oxide of ruthenium ( R U - O / C N - R U ) . ~Although ~~ the mechanism of catalysis has not been worked out in detail, evidence from work with cystine indicates that the film catalyzes the transfer of oxygen as well as electrons.2l This is significant because oxidation of disulfides usually results in the formation of thiyl radicals, which adsorb to the electrode surface and passivate the electrode.23 The transfer of oxygen to the oxidized disulfide alters this mechanism and yields a nonpassivating product. Thus, the response of the electrode is stable. In this paper, the preparation of a microelectrode modified with the Ru-O/CN-Ru film which can detect insulin under physiological conditions is described. The electrode is used to make amperometric recordings near single 8-cells. It is demonstrated that the electrochemical technique offers the first direct detection of chemical secretions from 8-cells a t the level of single exocytosisevents. Therefore, the technique promises to allow unique insights into the secretion process. The main substance detected is tentatively identified as insulin. Thus, the electrode may also allow for the first practical use of a sensor for peptide hormone measurements in a biological system.
EXPERIMENTAL SECTION Electrode Preparation. Glass-encased carbon fiber microelectrodeswere prepared using previouslydescribed techniques.= Briefly,a glass capillary (AMsystems) containing a single carbon fiber of 9-pm diameter (P-55s from Amoco Performance Products) was pulled to a f i e tip on a commercial pipet puller (Narishige PE-2). The fiber was cut, yielding two electrodes. The carbon fibers were sealed in the tip of the glass by dipping them in epoxy. Once the epoxy was cured, the electrodes were polished at a 30-45O angle on a micropipet beveler (Sutter Instruments). Immediately after polishing, electrodes were dipped in 2-propanol for 10-15 min and then ultrasonicated in Hs0 for 5 min. An electrochemicaldeposition procedure similar to that described for macroelectrodes was used to apply the Ru(19) Stankovich, M. T.; Bard, A. J. J. Electroanal. Chem. 1977, 85, 173. (20) Cox, J. A.; Gray, T. J. Ami. Chem. 1989, 61, 2462. (21) Cox, J. A.; Gray, T. J. Anal. Chem. 1990,62,2742. (22) Kulesza, P. J. Electroanal. Chem. 1987,220, 295. (23) Pradac, J.; Koryta, J. J. Electroanal. Chem. 1968,17, 167. (24) Kelly, R. S.; Wightman, R. M. Anal. Chim. Acta 1986,187,79.
1889
O/CN-Ru film to the microelectrodes.8 Electrodes were dipped in a N2-purged solution of 0.1 M KC1,2 mM RuCls.H20, and 2 mM &Ru(CN)e adjusted to pH 2.0 with HC1. The electrode potential was cycled between +0.47 and +1.07 V (all voltages are versus a sodium-saturatedcalomel electrode (SSCE))at 50 mV/s for 25 min. Electrodes were removed from solution, rinsed with deionized water, and allowed to air-dry for 30 min at room temperature before use. Electrode Testing. Response times and calibration curves for the electrodes were obtained in a flow injection apparatus similar to that described elsewhereaMThis system consisted of a syringe pump (Harvard Apparatus 11) connected to a twoposition, six-port valve (Valco ACGUHC) equipped with a 250pL sample loop. The outlet of the valve was connected to a fused silica tube (10 cm long with 150 pm inner diameter (i.d.)) which emptied into a Lucite cell equipped with an SSCE. The working electrode was positioned in the outlet of the tubing using a micromanipulator. The flow rate was 3 pL/min. The entire flow injection system was housed in a Faraday cage. Data Collection and Analysis. Data for the single-cell measurements and electrode testing were collected using an EI400 potentiostat (Ensman Instrumentation, Bloomington, IN). The data were low pass filtered with a cutoff frequency of 20 Hz. Data were collected at 9.5 kHz, and points spanning 16.7 ms were averaged to give a single data point for single-cellmeasurements. This approach effectively reduced 60-Hz noise. The data were collected by an IBM-compatible personal computer (Gateway 2000 386-25MHz) via a TecMar multifunction board (LabMaster DMA TM-40 PGH). Data from single cells were analyzed using locally written software. The mean of measurements are presented with the standard deviation and number of samples (n). Cell Culture. Fresh human islets of Langerhans were obtained from the medical schools at Washington University or the University of Miami. (4-Cellsare dispersed throughout the pancreas in small clusters known as the islets of Langerhans.) Islets were dispersed into individual cells and maintained in culture according to the followingprocedure. Islets were washed twice with Ca2+, Mgz+-freeHank's media. Following washing, the islets were centrifuged at 1300 rpm at room temperature. After the medium was removed, 7 mL of versene was added and the islets were resuspended. After being incubated for 7 min, the isletswere centrifuged at 1300rpm for 2 min. The supernatant was removed, and 7 mg of trypsin in 7 mL of Ca2+, Mg2+-free Hank's medium at 37 O C was added. The islets were gently mixed in the trypsin solution for 3 min at 37 "C. The islets were pipeted and returned to 37 OC for 30-40 s. Culture medium (8 mL; provided with islets) was added to the vial and centrifuged at 1500 rpm for 5 rnin at 4 OC. The supernatant was removed and 3 mL of culture medium added. The cells were resuspended and counted on a hemacytometer. Cells were then plated at a density of approximately 106 cells per 35-mm petri dish. Cells were incubated at 37 O C , 100% humidity, and 5% COZ. Cells were used on days 2-6 of culturing. All chemicals for performing cell culture were from GIBCO. Single-cell Measurements. Single-cellmeasurements were performed in a manner similar to that described elsewhere.15J6 Petri dishes containingthe cells were removed from the incubator and rinsed 3 times with a Kreb's buffer. The Kreb's buffer contained (in mM) 118NaC1,5.8 KC1,2.4 CaCLl.2 MgSO,, 1.2 KH2PO,,25 NaHC03, and 3 glucose-d. The buffer was adjusted to pH 7.4 by bubbling 95% air/5% C02 through it. The dishes were filled with buffer and placed in a microincubator (Medical Systems, Inc.) on the stage of an inverted microscope equipped with phase-contrast optics (Zeiss axiovert 35). The microincubator maintained the C02 level at 5% and the temperature at 37 "C. A SSCE was fixed in the dish. To perform measurements, a cell was located with the microscope and a working microelectrode brought near the cell using a combined mechanical/ piezoelectricmicropoeitioner (BurleighPC-1OOO). Under highest magnification (640X), the electrode tip was allowed to lightly touch the cell, denoted by the cell moving, and then retracted (25) Cox, J. A,; Kulesza, P.J. Anal. Chem. 1984,56,1021. (26) Kristensen, E. W.; Kuhr, W. G.; Wightman, R. M. Anal. Chem. 1987,69,1752.
1884
ANALYTICAL CHEMISTRY, VOL. 65, NO. 14, JULY 15, 1993
C
B 1
Scan 1
!-
-
,
Scan 1 6 ’
F===
_ _ _ _ Scan 32 ~
1100
850
600
Scan 48
,
350
E (mV)vs SSCE
Flgure I.Voltammograms from electrochemical deposition of RuO/CN-RU film onto carbon fiber microelectrode. Conditions for voltammetry are described in the Experimental Section. Only four scans are shown for clarity.
the distance desired. Experiments were performed with an approximately 1-pmgap between electrode and cell unless noted otherwise. Substanceswere applied to individualcells by pressure-ejecting solutions from the tips of micropipets which were positioned approximately 30 pm from the cell. The micropipets were prepared by pulling 1 mm outer diameter (0.d.) by 0.5 mm i.d. glass capillaries (AM systems) on a Narishige PE-2 pipet puller with magnet settings of 5 and heater setting of 5 (15 A). The pipets were broken off to a tip 0.d. of about 10pm. The pressure ejection system (General Valve Picospritzer) allowed pressure pulses of variable magnitude and duration to be applied. Pressure pulses of 10 p.s.i. were used. Flow rates through the pipet tips were approximately 1 nL/s as evaluated by measuring the size of solution droplets formed after ejection under mineral oil. If glucose stimulation was performed, the solution in the pipets consisted of Kreb’s buffer with 16 mM glucose. In this case, the pressurepulses were 60 s long. If K+stimulations were performed, the pipet solution contained Kreb’s buffer with 60 mM KC1 and NaCl reduced an appropriate amount to maintain salt balance. Pressure pulses of 2 s were used in this case. The pipet tips were positioned next to the cells using a mechanicalmicromanipulator (Narishige MM-333). For recordings, the working electrode was at +0.85 V unless noted otherwise.
RESULTS AND DISCUSSION Electrode Preparation a n d Testing. Voltammograms obtained during modification of one of the microelectrodes are shown in Figure 1. The voltammograms are characteristic of surface deposition of an electroactive species with increasing surface coverage with time. The waves began with a sigmoidal shape and eventually became dominated by peak-shaped voltammograms. The peak anodic currents at +0.90 V averaged 10.1 f 0.4 nA (n = 4) following 25 min of electrochemical deposition. Microelectrodes prepared in this manner were highly active in acidic media, as reported in the literature for macroelectrodes.zOP22 Steady-statesurface waves on the microelectrodes in 0.15 M pH 2.0 sodium phosphate buffer, shown in Figure 2A, were similar to those reported previously for films deposited on other macroelectrode substrates.22~25The anodic peaks observed a t +0.78 and +0.96 V have been attributed t o the two-step oxidation of ruthenium in the ruthenium oxide portion of the film from the 3+ to the 4+ state.22 Measurements of insulin in the flow injection apparatus had detection limits of 0.1 pM and a linear range of 3 orders of magnitude in several acidic electrolytes including 0.1 M KzS04a t pH 2.0,0.15 M potassium phosphate buffer a t pH 2.0, and 0.15 M sodium phosphate buffer a t pH 2.0. The detection limit was defined as the concentration which gave a signal equal to the signal from a blank injection plus 3 times the standard deviation of the blank. These figures of merit are in agreement with those reported for macroelectrodes under similar conditions20 The electrode response to insulin
7-
I
1100
40011100
750
E (mV) vs SSCE
E
0.1 nAmp
750
400 1100
(mV) vs SSCE
I
750
400
E (mV) vs SSCE
Flgure 2. Surface voltammetric waves on Ru-OICN-Ru-modified electrodes as a function of electrolyte and resting potentlal. All voltammograms were obtained at 50 mV/s In unstirred solutions of electrolyte. Conditions were as follows: (A) after 30 min In 0.15 M sodium phosphate buffer at pH 2.0 with a resting potential of +0.85 V, ( 6 )after 30 min In 0.15 M sodium phosphate buffer at pH 7.4 with a resting potentlal of +0.65 V, and (C) after 30 mln In 0.15 M sodium phosphate buffer at pH 7.4 with B resting potential of +0.40 V.
12
.-
8 4
0
30
60
90
120
150
Time (min) Flgure 3. Stability of the peak current at Ru-O/CN-Ru-modlfied electrodes in physiological buffer resulting from injection of 10 pM insulin. The lines through the polnts are only intended to highlight the trends.
was stable for over 1h when held a t a fixed potential of +0.85 V. +0.85 V was chosen as the testing potential since a hydrodynamic voltammogram revealed that this voltage gave the peak current response. When the electrode was used in 0.15 M sodium phosphate buffer at pH 7.4, it was observed that the response to insulin deteriorated markedly. An example of the time profile of the response is shown by the open symbols in Figure 3. The peak current resulting from a 3-s-wide concentration pulse of 10 WMinsulin, applied in the flow stream, was used to obtain the points. The times indicate the time since the electrode was brought in contact with the buffer. While the decrease slows down after 30 min, the signal is low relative to the initial response. A surface voltammogram obtained in pH 7.4 sodium phosphate buffer after the electrode response decreased is shown in Figure 2B. As can be seen, the voltammetric response was drastically altered relative to the voltammograms obtained in the pH 2.0 sodium phosphate buffer. The voltammograms indicate that an electroactive film of some type still existed on the electrode; however, it was significantly different from the initial film and it did not effectively catalyze the oxidation of insulin. Similar results were obtained in pH 7.4 potassium phosphate buffer and pH 7.4 Kreb’s buffer. It was concluded that the higher pH of the buffer had a significant role in degrading the response of the electrode. The stability study was repeated, but with the electrode held at +0.4 V between injections of insulin to determine whether the deleterious effect of electrolyte pH was dependent on electrode potential. The response is shown by the solid
ANALYTICAL CHEMISTRY, VOL. 85, NO. 14, JULY 15, 1993
1885
20.020
0.015 0.010
L
0.0
2.0
4.0
6.0
8.0
10.0
Time (s) Flguro 4. Response .of Ru-O/CN-Ru-modified electrode to 20 pM insulin injected In a flowing stream. The dips in current are artifacts resulng from the actkatlon of a solenoid which turns the injection valve and are convenient markers of the injection time. The delays before Initial increase in current after injection are due to the dead volume of the flow system.
symbols ("rested" electrode) in Figure 3. The electrodes maintained catalytic activity for much longer times if the potential was not constantly a t +0.85 V. It was possible to maintain stable responses for 2 h as long as the electrode was held at +0.85 V for less than 90 s at a time and held at +0.4 V for at least 20 s at a time. A surface wave voltammogram obtained in the pH 7.4 sodium phosphate buffer after 30 min at +0.4 V is shown in Figure 2C. The irreversible voltammetry may indicate that a t neutral pH oxidation of the film results in formation of an irreversible product. This voltammetric wave under these conditions was not as stable as that obtained in pH 2.0 buffers. Voltammograms obtained every 10 min with the electrode held a t +0.4 V between scans showed progressive decreases in current levels so that after 2 h the background waves were approximately half those obtained initially. In 0.15 M sodium phosphate buffer at pH 2.0 the surface waves (shown inFigure 2A) showed no changes over a 2-h period whether the resting potential was +0.40 or +0.85 V. In physiologicalbuffer, with the electrode periodically held at +0.4 V, the detection limit, defined as described above, was approximately 0.5 pM. The poorer detection limit relative to that obtained at pH 2.0 may result from the reduced rate of oxidation of the film indicated by the drawn-out surface waves at pH 7.4 (compare parts A and C of Figure 2). A typical calibration curve obtained withinsulin concentrations of 0, 5, 10, 20, and 40 pM had a slope of 0.441 pA/pM, a y-intercept of 0.23 PA, and a linear correlation coefficient of 0.9992. The high y-intercept was due to signals obtained from blank injections which resulted from an artifact of the injector. An example of the response of the electrode to an injection of 20 pM insulin in the flow stream is shown in Figure 4. The data points are 100 ms apart. The response time was faster than could be measured by the flow injection apparatus and was C200 ms to 90% of the full height. Thus, these experiments indicated that the electrode could be used as an amperometric detector of insulin under physiological conditions as long as the electrode could be *rested" a t +0.4 V in between periods of measurement at +0.85 V. Single-cell Measurements. A typical recording taken with the electrode about 1 pm away from the cell following a glucose stimulation is shown in Figure 5. Immediately followingapplication of glucose the current was stable. After a few seconds, however, the current began to fluctuate in a series of rapid spikes. Current spikes were observed after stimulation of 11out of 30cells attempted. The time between the end of the stimulation and the beginning of spiking varied between 5 and 180 s. Spiking was generally observed to occur for over 3 min and for as long as 10 min on one occasion. Because of the need to rest the electrode, it was difficult to determine exactly the time between stimulation and beginning
5s Flguro 5. Amperometric recording from a slngle cell following a 80-5 glucose stimulation as described in the Experimental Section. The arrow indicates the end of the stlrnulatlon.
15
15 PA
30 1s
Flgure 6. Effect of distance between electrode and cell on recording from single cell followingglucose stimulation. The numbers beside the recording indicate the distance in micrometers.
of spiking and how long the spiking continued. The frequency of spikes, determined by counting the number of peak maxima that occurred during a 5-s period, varied between 6 and 10 Hz. The end of a series of current spikes was characterized by a decrease in the frequency of the spikes but not a change in their shape or height. No spikes were observed prior to application of glucose and no spikes were observed if physiological buffer was applied to the cell. Our hypothesis is that the spikes represent rapid changes in insulin concentration a t the electrode surface resulting from secretion of insulin by exocytosis. Following the lead of Wightman's group,lSJ6 a series of experiments to test this hypothesis were performed. In testing the hypothesis, the other following sources of "noise" were considered: (1) sampling artifacts or electrical noise from external sources, (2) nonfaradaic currents generated by ion fluxes or potential fluctuations at the cell, and (3) anodic detection of other compounds released from the cell. In addition, an analysis of the spike frequency and areas was performed to determine whether they were consistent with the hypothesis. The possibility of external noise or sampling artifacts was eliminated by moving the electrode away from the cell following a glucose stimulation. An example of the data is shown in Figure 6. The height and number of spikes decreased as the electrode was moved away from the electrode for all attempts of this experiment (n= 4). Spikes were still observed if the electrode was moved closer to the cell again. It was concluded that the spikes were due to the cell and not an external artifact. The average width of the spikes, calculated as the full width a t half-height from a random sampling, was 52 f 11 ms (n = 33) with a 15-pm cell-electrode spacing and 38 f 13 ms (n = 38) with a 1-pm spacing. These differences are statistically significant to a greater than 99 % confidencelevel. (The peak widths are given only for comparison to each other and should not be taken as absolutely accurate since the spikes have
Ru-O/C N-R u
i
+0.40 V -
2s 1s
Effect of electrode voltage on recording from single cell following glucose stlmulatlon.
Flgure 7.
been overfiltered, the detector time constant was 50 ms, and undersampled, the data collection rate only allowed a few points to be collected across the narrower peaks.) The trends observed in peak width and height are consistent with the notion that spikes are the detection of a packet of molecules diffusing to the electrode from the cell. The greater the diffusion distance, the more dilute and drawn out a concentration pulse will appear.27 The lack of spikes at a 30-hm distance is also consistent with the idea of detection of a diffusing packet of molecules. When the electrode is close to the cell, it effectively traps molecules which originate at the cell in a thin layer. As the electrode is pulled away, secreted molecules are more likely to diffuse away from the electrode without being detected. Two experiments were performed to test the possibility of nonfaradaic currents being generated at the electrode by the cell. In the first, measurements were taken from a glucosestimulated cell with the electrode a t +0.85 V and then at +0.4 V. An example of the two measurements is depicted in Figure 7. Decreasing the voltage to below the oxidation potential of insulin resulted in a quiet baseline compared to the signal a t +0.85 V. This result was observed for every time spiking was observed. A strong voltage dependence for faradaic current and little or no voltage dependence for nonfaradaic sources was expected. Thus, these results are consistent with the idea that the current spikes are due to faradaic current. In a second experiment, measurements a t glucose-stimulated cells using both an unmodified carbon fiber electrode and a modified electrode were performed. A modified and a bare electrode were both held at +0.85 V and mounted in a dual-electrode holder which held the electrode tips approximately 1 cm apart. The cell was stimulated with a modified electrode in place. If spiking was observed, the bare electrode was quickly positioned near the cell. A typical comparison of traces from the two electrodes is depicted in Figure 8. Current spikes were never observed for a bare electrode ( n = 4). It was expected that if the current spikes were due to a nonfaradaic source, then similar noise would be observed at both electrode types. Thus, this experiment also discounts the possibility of nonfaradaic currents being generated by the cell. A final possibility that was considered was that another chemical species was secreted by the cell, resulting in the observed currents. Fast-scan cyclicvoltammetry is often used to identify substances detected electrochemically in biological systems.gJ5 In this case, voltammetry is difficult to use and would not provide definitive evidence. The large background currents due to the electroactive film on the surface prevent the use of scan rates high enough to measure rapid concentration fluctuations. In addition, since the film works a t least (27) Engstrom, R. C.; Wightman, R. M.; Kristensen, E. W. Anal. Chem. 1988, 60, 652.
Comparison of amperometric recording from a single cell using a bare (unmodified)carbon fiber microelectrode and a Ru-O/ CN-Ru-modified electrode. Figure 8.
partially by electron m e d i a t i ~ nall , ~substances ~ ~ ~ ~ oxidized yield voltammograms similar to the surface waves. Therefore, voltammetric identification was not pursued. The first approach was to determine whether substances known to be secreted by the cells would be oxidized by the electrode. The other primary constituents of the vesicles that are released with insulin are ATP, C-peptide,and proinsulin.% Proinsulin is present in low levels in the vesicles and is not likely to be detected.l ATP and C-peptide, although present in high levels in P-cell vesicles, were not detectable by the electrode in flow injection experiments a t 1 mM and 1 pM, respectively. Norepinephrine (NE) is also suspected to be present in the vesicles and released simultaneously with insulin.28 NE is detectable at the Ru-O/CN-Ru-modified electrode; therefore, the observed spiking could be due to detection of NE. However, NE is also detectable a t a bare electrode and no significant signals have been observed at bare electrodes (see Figure 8). The detection limit for NE by modified electrodes was 50 nM compared to about 80 nM for bare electrodes. (Detection limits were calculated as for insulin in the flow stream. The slightly better detectionlimits were due tolarger signals obtained using the modified electrodes.) Given the similar sensitivities of the two electrode types, it is reasonable to conclude that the absence of spikes observed using bare electrodes indicates that NE does not significantly contribute to the spikes observed using modified electrodes. This may be due to a low level of NE present in the vesicles. It should be noted that there may be unknown constituents of the vesicles which were not tested for here and which may have contributed to the spikes. Spike Frequency and Area. Analysis of the area and frequency of the current spikes is useful in testing the hypothesis that the spikes resulted from detection of exocytosis events. If a spike represents the detection of insulin secreted by exocytosis of a single vesicle, then the area under the spikes would depend on the number of moles contained in the vesicle and the number of moles captured by the electrode. The average area from a sample of 114 spikes collected from a single electrode at a single glucose-stimulated cell was 0.23 f 0.20 pC. In obtaining the sample, obviously overlapping spikes were rejected. The average spike area corresponds to 600 zmol as calculated from Faraday’s law and using a value of 4 electrons transferred per mole. This value of electrons per mole was reported for the oxidation of cystine a t a Ru-O/CN-Ru electrode21 and was used by assuming that one cystine was oxidized in insulin by the electrode. From data on whole plates of cultured cells, it is estimated that an individual 0-cell contains between 2.5 and 10 fmol of (28) Fujita, T.; Kobayashi, S.; Serizawa, Y. In Proinsulin, Insulin, C-peptide;Baba, S., Kaneko, T., Yanaihara, N., Eds.;Excerpta Medica: Oxford, UK, 1978; pp 327-338.
ANALYTICAL CHEMISTRY, VOL. 65,
5s
Figure 0. Amperometric recording from a single cell following K+ stimulation. The bar indicatesthe application of 60 mM K+ solution.
insulin.2Qpm Ultrastructural analysis shows that the cells contain about 13 OOO v e s i c l e ~ . ~Thus, ~ J ~ if it is assumed that all insulin is contained within the vesicles, then each vesicle should contain between 200 and 800 zmol of insulin. The measured values of spike area agree well with estimates of vesicle content. Agreement would be expected if the electrode is capturing most of the insulin secreted by a given vesicle. This is reasonable given the proximity of the electrode and cell. It cannot be concluded, however, that the electrode captures all of the vesicles released by the cell since vesicles could be released on the side of the cell opposite the electrode. Another prediction of exocytosis theory is that spike frequency and number may vary with stimulus, but spike area should be independent of stimulus.16 Spike area should be unchanging since it should be determined by vesicle content, which is not affected by an acute stimulus. It is well-known that high concentrations of K+ can depolarize @-cellsand cause insulin secretion.33 Figure 9 shows a typical recording taken after application of 60 mM K+ to a single ,T-cell. The current dip during stimulation was observed in the absence of cells and is apparently due to a nonfaradaic effect. Current spikes were observed in 8 out of 24 cells attempted. Spikes began between 1 and 24 s after the beginning of stimulation. Spikes occurred for 10-60 s after the first spike was recorded at a rate of 0.4-1.3 Hz over the duration of spiking. Thus, compared to glucose-stimulated spiking, the K+-induced spiking started sooner after stimulation, was of shorter duration, and had lower frequency. The average area under a sample of 56 spikes pooled from six different K+-stimulated cells was 0.28 f 0:29 pC, which is not significantly different from the 0.23 pC for a glucose stimulation. Thus, area of spikes did not vary with stimulus as (29) Weir, G.C.; Halban, P. A.; Meda, P.; Wollheim, C. B.; Orci, L.; Renold, A. E. Metaboliem 1984,33,441. (30) Wollheim,C. B.;Meda, P.;Halban, P. A.MethodsEnzymol.1990, 192, 223. (31) Goldstein, M. B.; Davis, E. A. Acta A m t . 1968, 71, 161. (32) Dean, P. M. Diabetologia 1973, 9, 115. (33)Daweon, C. M.; Lebrun, P.; Herchuelz, W. J.; Malaisse, W. J.; Goncalvee, A. A.; Atwater, I. Horn. Metabol. Res. 1986, 18, 221.
NO. 14, JULY 15, 1993 1887
expected for an exocytosis measurement. The differences in spike frequency and duration may be related to the different mechanism of stimulation and will be explored further. Stimulation Success Rate. It is intriguing that current spikes were recorded a t only 37 % of the glucose-stimulated cells and 33% of the K+-stimulated cells. These numbers are in reasonable agreement with those obtained using the reverse hemolytic plaque assay, which have shown that only 31.5 % of isolated pancreatic &cells secrete detectable amounts of insulin in response to 30-min incubations with 16.7 mM glucose.' The reason for the low response rate of the cells is unknown; however, it underscores the value of chemical evaluation at the single-cell level.
CONCLUSIONS The combined evidence shows that the current spikes that are observed at the electrodes following stimulation are due to anodic detection of one or more substances near the cell. The frequency and area of the spikes are consistent with the hypothesis that the detected substance(s) is (are) secreted by exocytosis. Thus, the electrode has the sensitivity and time resolution to detect exocytosisevents from individual &cells based on anodic detection of the secreted substance(s). This is the first time that secretion has been observed with this resolution a t &cells,thus many new opportunities for studying secretion should be afforded by the technique. The data support, but do not prove, that the primary substance detected is insulin. The most useful experiment to determine the identity of the detected substance(s) would be an independent chemical analysis and an electrochemical measurement a t the same cell. Capillary electrophoresis and capillary chromatography techniques to provide this evidence are being developed. Confirmation that insulin is the primary detected substance would allow studies directed toward insulin-specific aspects of secretion to be performed. Other improvements will also be sought in the technique, such as improving the stability of the electrode for longer term studies and improving the stimulation conditions to allow greater control over secretion.
ACKNOWLEDGMENT We thank Dr. R. Mark Wightman and Mr. Jeffery Jankowski for helpful discussionsand the use of data collection software. We thank Dr. Paul Lacy of Washington University for recommending the cell culture protocol. We also thank Dr. David Sharp of Washington University for providing the islets. This work was supported by the University of Florida, NSF CHE-9116417 (R.T.K.), and the American Diabetes Association (M.A.A.).
RECEIVED for review October 26, 1992. Accepted April 1993.
5,
