Flow injection analysis as a diagnostic tool for development and

senslng system to monitor changes In pHand Is subjected to a thorough evaluation, using the flow Injection analysis tech- nique: sensor stability (bot...
5 downloads 15 Views 919KB Size
1250

Anal. Chem. 1988, 60,1250-1256

Flow Injection Analysis as a Diagnostic Tool for Development and Testing of a Penicillin Sensor Tracy D. Yerian, Gary D. Christian,* and Jaromir Ruzicka Center for Process Analytical Chemistry, Department of Chemistry, BG-IO, University of Washington, Seattle, Washington 98195

A penlclllln sensor is developed for contlnuous, on-llne analysis. The sensor consists of the enzyme, penlclllinase, cross-linked with bovine serum albumln Into a cellulose pad or covalently attached to the cellulose, with an acid-base Indicator dye also covalently Immoblllzed to the surface of the cellulose. The sensor is placed wlthln a flow Injection optosensing system to monitor changes In pH and Is sublected to a thorough evaluation, uslng the flow Injection analysis technlque: sensor staMllty (both dye and enzyme stabUlty), speed of sensor response, sensor sensltvlty, and sensor Metlme data are obtained.

Impulse-response flow injection analysis (FIA) is based on the repetitive action of a well-defined zone of a selected chemical specie (E) on a target, situated in an unsegmented carrier stream (1). The target may be a sensor; the carrier stream is inert-i.e., a solution to which the sensor does not respond, or to which the sensor responds at the detection limit, thus establishing a continuous base line. The injected sample is the one to which the sensor is to respond, either directly or indirectly, through a series of chemical reactions. By continuously monitoring the sensor output, while repeatedly injecting (via a valve) a well-defined zone of species at a fixed concentration, a series of peaks is generated, which reflect the response of the sensor over a test period; the response may be unchanged or show deterioration within the selected time period. The procedure is automated; hence the response is not a function of subjective judgement or manual operational errors. In addition to examination of the tendencies of a series of response peaks, changes in individual peak shapes may give information about phenomena occurring a t the sensor surface. Since many chemical sensors utilize a chain of chemical reactions in the sensing area, it should often be possible to examine each set of sensing reactions separately, to differentiate the components of the sensor response. An example of such systematic sensor testing and evaluation is the work on the penicillin sensor described below. This sensor is based on the enzymatic hydrolysis of penicillin to penicilloic acid, resulting in a p H change which is sensed optically by means of a colored acid-base indicator. An integrated microconduit (2) is designed to contain a solid support with an immobilized acid-base indicator within the flow cell. A variation in the solution pH results in a color change on the sensor surface, which is monitored via a fiber-optic bundle. The theory of this approach is well-known and is described elsewhere (3-5). EXPERIMENTAL SECTION Apparatus. All experiments were carried out with a BifokTecator FIAstar 5020 flow injection analyzer. The manifold used (Figure 1) incorporated integrated microconduits into the FIAstar system, replacing the original injection valve. The integrated microconduit was composed of a flow-through detector and miniaturized injection valve, with variable sample volume. The

flow-through cell interfaces with a bundle of plastic optical fibers, for transmission of source illumination to the flow cell and the reflected signal from the flow cell to the detector. Reflected signals were measured at 605 nm. The wavelength of maximum signal is determined by scanning through the visible reflectance spectra of the two colors of the acid-base indicator and subtracting one color spectrum from the other (base line) spectrum, a feature of the Tecator FIAstar 5023 detector. The wavelength at which the color reflectance change is maximum in going from one form to the other is selected for measurements. For Merck indicator 9582, the color change is from blue to yellow. The reflectance measurements (A,) were registered automatically on the FIAstar 5023 spectrophotometer and digitally displayed and printed on the FIAstar 5020. The results were concurrently fed to a recorder (Radiometer,Servograph REC-61, furnished with an REA-112 high-sensitivity interface). A, represents log 1 / R , where R is the reflectance (relative to the reflectance of the pad in the presence of the buffer carrier). All measurements were rate measurements, performed by stopping the flow in the sample cell for fixed time (usually 20 s), followed by flushing out by the carrier, resulting in a peak-shaped signal. The peak height was a measure of the rate of change of PH. Reagents. All chemicals used were of analytical-reagentgrade; all water was deionized. The carrier solution for the buffer pH determinations was 1 X loM3M HC1. The carrier solution for M K2HP04,0.1 M KCl, penicillin determinations was 1 X adjusted to pH 6.8 with 0.1 M NaOH, unless otherwise stated. The penicillinase (Sigma P0389, Type 1) contained 1000 units of enzyme (2.6 mg of solid). Soluble enzyme studies were performed by dissolving 1.3 mg of the solid (approximately500 units) in 10 mL of carrier solution. Penicillin standards were prepared by dilution of 50 mM stock penicillin with carrier, prepared daily (as needed), as the unbuffered solutions of penicillins are not stable (6). Stock penicillin was prepared by dissolving the penicillin salt in carrier solution. Penicillin G was purchased through Sigma; all other penicillins tested were supplied by the Eli Lilly Co. Cephalexin (Sigma C4895) and Cephalothin (Sigma C4520) solutions were prepared by dilution of 10 mM standards. The 10 mM solutions were prepared by dissolving the solid in carrier (0.105 g of Cephalothin,0.087 g of Cephalexin in 25 mL of carrier). Buffer solutions (citric acid-phosphate) were prepared according to Table 10.47 in Perrin and Dempsey (7). Procedure. Covalent Immobilization Procedure. The procedure employed was adapted from a method used to immobilize enzymes to porous glass, described by Adams and Carr (8). Silanization of the cellulose surface is a modification of the procedure described by Weibel et al. (9): the cellulose pads were stirred, at reduced pressure, in 20 mL of 10% (v/v) (3-aminopropy1)triethoxysilane (Sigma) in toluene, at 75 "C for 3 h. The cellulose pads were removed from the mixture, washed with 96% ethanol, then placed in 20 mL of 2.5% (v/v) glutaraldehyde (Sigma)in pH 7.0 phosphate buffer, and kept at room temperature and reduced pressure for 1.5 h. Finally, the activated pads were soaked in enzyme solution for 5 h, at reduced pressure to remove microbubbles from the pad and facilitate diffusion of the enzyme to the reactive CHO- group of the glutaraldehyde. Two enzyme solutions were prepared one solution contained 0.65 mg of enzyme solid in 0.5 mL of pH 7.0 phosphate buffer, and the other solution contained 0.65 mg of enzyme solid and 20 mg of bovine serum albumin (Sigma A6003) in 0.5 mL of pH 7.0 phosphate buffer.

0003-2700/88/0360-1250$01.50/00 1988 American Chemical Society

ANALYTICAL CHEMISTRY, VOL. 60, NO. 13, JULY 1, 1988

1251

0.80~

0.60

. Waste

-

I 0.20

Spectrophotometer

Figure 1. Manifold for sensor testing consisting of two peristaitlc pumps (Pl, Pa), a timer (T), and the integrated microconduit (boxed area). The flow cell communicates wlth the light source and detector of the spectrophotometer by means of optical fibers.

The pads are kept in 0.1 M pH 7.0 phosphate buffer and stored at 0-5 "C. Cross-Linked Enzyme Procedure. Bovine serum albumin (15 mg) and 1.3 mg of penicillinase (500 units) were added to 1.0 mL of pH 7.0 phosphate buffer. Approximately 50 FL of 2.5% glutaraldehyde (in pH 7.0 phosphate buffer) was added to the enzymealbumin solution, and the resulting solution was dispensed by dropper onto the cellulose pads. This is the "pad equivalent" of the cross-linking procedure used for the preparation of immobilized enzyme electrode sensors (10). RESULTS AND DISCUSSION Various components and applications of our system were subjected to systematic evaluation by repetitive measurements using injected species. These include the response of the pH indicator to solutions of different buffer concentrations, stability of the pH indicator and the immobilized enzyme, response of the enzyme to different substrates, evaluation of different methods of enzyme immobilization, evaluation of the kinetic response of the immobilized enzyme, and measurement of the effect of carrier buffer capacity and ionic strength on the dye and the enzyme response. Indicator. Type of Indicator. The solid support, or "active surface", implemented in our system is the ColorpHast (nonbleeding) indicator strips. These are cellulose fibers on which the acid-base indicators have been covalently immobilized. The acid-base indicators are azo dyes

AR-N=N-AR,

(1)

where AR is an aryl, with one or more reactive groups capable of forming a bond with cellulose; examples are -SO2CH2CH2OSO20Hand -(CH3)NS02CH2CH20S020H. The dyes are made nonbleeding by incorporating additional sulfonic acid and/or carboxylic acid groups into the aryl groups (11). The indicator strips are commercially available Merck products; the different strips available collectively span the pH range 0.0-13.0, and each individual cellulose pad will change color over a range of 2.0-3.5 pH units. Merck Indicator 9582 Response to p H . To determine the magnitude of the indicator response in the pH range 4.0-7.0, and the shape of the A, vs pH curve, a series of buffers (100 pL) was injected in duplicate into a carrier of 1 X 10-3M HCl. The response is linear with pH (as measured potentiometrically with a Radiometer pH meter and Corning glass electrode) in the pH range 4.847-6.061, with a simple linear regression correlation coefficient of 0.997. The coefficient of correlation for the best fit line drops to only 0.993 when the data are extended to include values up to pH 6.49 and down as low as 4.273.

'

0.00

4.0

5.0

6.0

7.0

PH Figure 2. Effect of carrier ionic strength on Merck indicator 9582 pH response. Data points are the average A , of duplicated injections of 50 pL of citric acid-phosphate buffer solutions, pH 4.0-7.0. Carrier solutions are 1 X M HCI and (0) no NaCI, (0)0.1 M NaCI, and (0) 1.0 M NaCI.

The average variation in duplicate injections of buffer in the linear region of dye response is 0.003 A , units (0.8% relative standard deviation at pH 5.240). Using A, values for pH 5.240, this corresponds to a change in pH of 0.005 unit. When the standard deviation of a series of buffer injections is calculated for a solution giving near maximum response (pH 6.8; corresponding to plateau of the titration curve), the relative standard deviation of 10 injections is 0.7%. This value should reflect instrumental noise, as any pH response variability is minimized in this region. Because precision in the linear region is similar to the plateau region, the major source of error is instrumental noise. Lifetime of Immobilized Indicator. The immobilized indicators have been demonstrated to be stable to at least 1200 repetitive injections of buffer solution (12, 13). To date, individual sensors have been used for over 12 months without any loss in dye activity. Ionic Strength Effect on Response. To determine the effect of carrier ionic strength on the dye response to a buffer, the peak height of duplicate injections of citric acid-phosphate buffers (1X lo4 M) ranging from pH 4.0 to pH 7.0 was plotted as a function of pH (as measured potentiometrically) at three carrier ionic strengths: 1.0 M NaC1,O.l M NaCl, and 0.01 M M HC1 (Figure 2). The effect is slight, NaCl in 1 X indicating that a low degree of interference from any variability in ionic strength can be expected. Response to Buffer Concentration. The recorded optical signal is a response to change in hydrogen ion concentration; it is not strictly which is an activity measurement (pH = -log aH+).aH+is measured on a numerical scale of potential, called the hydrogen scale, where EH = EHo R T In uH+ (2)

+

and EHo is the standard potential of the reversible hydrogen electrode (14). The dye-surface-solution measurement can be quite complicated, as the dye response is dependent on its own pK, and charge and on the ionic strength of the carrier solution (15). The buffer or sample injected will have a characteristic response peak, dependent on its molecular charge, which affects its retention time within the pad, and the ionic strength of the sample, as well as the carrier ionic strength. The response

1252

ANALYTICAL CHEMISTRY, VOL. 60, NO. 13, JULY 1, 1988

time 0



.istop I4

time

Flgwe 3. (A) Response of the Merck indicator 9582 to 100 pL of 0.1 M NBS buffer, pH 4.0, as a transient response (a) and as a 20-s stop-flow response (b). (B) Response of the sensor to 100 pL of 0.01 M NBS buffer, pH 4.0, as a transient response (a) and as a 20-s stop-flow response (b).

of the dye to a sample plug is also a strong function of the sample buffer concentration (buffer capacity), which will be to some extent a function of sample size, i.e., dispersion. To avoid sample dilution for simple pH measurements, dispersion should be minimized; this will also minimize interactions of the sample with the carrier stream. However, even with dispersion minimized, the transient signal will be a strong function of the buffer capacity of the injected solution. This can be demonstrated by two different experiments: in the first experiment, NBS pH 4.0 buffer (0.1 M) is injected into a dilute phosphate carrier stream without any stop in solution flow, and the resulting peak is compared to an injection of the NBS buffer into the manifold, where the flow is stopped for 20 s with the buffer sample in contact with the dye (Figure 3A). The extra time for the dye-buffer interaction did not result in an increase in peak signal, verifying that equilibrium was achieved-i.e., all the dye molecules responded to the buffer during the transient response. When the NBS buffer is diluted to 0.01 M, and the above experiment is repeated, there is a 12% increase in peak height following stopped-flow, when compared to the transient peak response for 0.01 M buffer (Figure 3B). This indicates that equilibrium is probably not be achieved for dilute or weakly buffered solutions; hence the transient peak response of the system to these solutions will be a function of flow conditions as well as pH of the original sample. In a second set of experiments, citric acid-phosphate buffer is injected into the dilute phosphate carrier at different buffer concentrations, with the pH of all solutions adjusted to pH 4.66 by glass electrode. When the transient responses of the sensor to injection of 100 p L of 0.1,0.01,0.001,and O.OOO1 M buffer are compared, the magnitude of the signal is clearly decreased at the low buffer concentrations (Figure 4A). When the 0.1 M signal and the 0.01 M signal are compared, peak height is not appreciably affected, but the peak width is narrowed for the more dilute buffer. This facet of the response will be a function of the carrier interaction, both washout and neutralization characteristics. Because the transient response is not at equlibrium, the sample volume for dilute solutions will have a large effect on the signal magnitude. With injection of larger sample volumes (400pL) of the dilute solutions, the equilibrium response is eventually achieved (Figure 4B). This is, in effect, a “titration” of the dye molecules on the sensor, and both sample buffer concentration and carrier neutralization capacity will be variables affecting the apparent rate a t which this titration occurs.

Flgure 4. (A) Response of the Merck indicator 9582 to 100 pL of citrate buffers, pH 4.66, at buffer concentrations of (a) 0.0001 M, (b) 0.001 M, (c) 0.01 M, and (d) 0.1 M citrate buffer. (B) Response of the sensor to 400 pL of the same citric acid buffers, (a) 0.1 M, (b) 0.01 M, (c) 0.001 M, and (d) 0.0001 M.

0

5

15

25

m M penicillin G

Flgure 5. Calibration curves for soluble penicillinase and standard

solutions of penicillin 0, at three different positions along the samplereagent interface: (a) 4, (b) 5, (c) 3,and (d) 6 s after injection. Soluble Penicillinase. Calibration Curve. To generate a calibration curve for penicillin G with soluble penicillinase, 25 p L of enzyme solution, corresponding to 1.25 units of enzyme, and 50 p L of penicillin standard were injected simultaneously via the split loop injection technique (11)into the integrated microconduit. The change in reflected signal (A,) was recorded while the flow was stopped for 20 s when the sample reached the detector. Four distinct curves were generated, by stopping the flow a t different points along the sample-reagent zone: at 3,4,5, and 6 s after injection (Figure 5). The maximum response to penicillin G standards was obtained by stopping the flow 4 s after injection. However, the usable range over which the enzyme responds to a change in concentration of substrate is the same for the 4-, 5-, and 6-s stop-time. When the flow is stopped 3 s after injection, sensitivity is low but the sensor exhibits significant change in response to 25 mM penicillin G. At 10 mM penicillin G, stopping the flow 5 s after injection for 20 s generates a change in A , of 0.429. When the flow is stopped for 30 s under the same conditions, the change in A, is 0.456. This 6.0% increase in signal indicates that the sensitivity is not significantly increased at high substrate concentrations by allowing a longer reaction time. Adsorbed Enzyme. When substrate is injected without soluble enzyme reagent, no change in A, is observed when the flow is stopped. This indicates that the soluble penicillinase is not adsorbed to the cellulose (or any other part of the system), a phenomenon that has been observed with urease on cellulose (12)and for the penicillinase enzyme on porous glass (16). The other conclusion that can be drawn is that the sensor does not give a significant blank response to dilute

ANALYTICAL CHEMISTRY, VOL. 60, NO. 13, JULY 1, 1988

Table I. Calibration Data for Soluble Penicillinase and Penicillin G Standard Solutions in Two Carriers” sample concn, mM

1 X lo-’ M KHzPOl

1 X lo4 M KH2P04

25.00 10.00 5.00 2.50 1.25 0.63 0.31

0.522 0.393 0.218 0.093 0.021

0.500 0.514 0.369 0.230 0.104 0.031 0.023

1253

Table 11. Transient Peak Response of the Immobilized Penicillinase Sensor to Different Volumes of Penicillin G Standards

i i h = 2)

“The two carriers are: 1 X M KHzP04,pH 6.8, 0.1 M KC1, and 1 X lo4 M KHzP04,pH 6.8,O.l M KC1. Stop time is 20 s, 4 EI after injection. Change in A. is recorded during stop time. penicillin samples. For unknown or variable matrix, then, any blank response could be evaluated without interference from adsorbed enzyme or the substrate. Carrier Buffer Capacity. When the carrier phosphate concentration is increased 10-fold (Table I), a loss in signal is observed at all but the highest concentration of penicillin G (25 mM), consistent with literature reports for other pHdependent measurements for immobilized penicillinase (16). As noted before, the ionic strength can also have an effect on the pH response signal. Hence, measurements should be made in solutions of fixed optimum pH and ionic strength. Immobilized Penicillinase. Enzyme Immobilization. We have tested a number of immobilization methods: inorganic bridge formation (19,cyanuric chloride activation (18), adsorption (12),cross-linking (with and without bovine serum albumin) the enzyme into the pad with glutaraldehyde (19), and covalent attachment via silanization of the cellulose, followed by glutaraldehyde coupling of the enzyme to the activated surface (8). Two immobilization methods have proved successful: glutaraldehyde cross-linking of the penicillinase (with or without bovine serum albumin) within the fibrous pad; and covalent attachment via silanization and glutaraldehyde coupling to the cellulose (without bovine serum albumin). The reagents involved in inorganic bridge formation were found to destroy the dye attachment to the cellulose. While cyanuric chloride activation conditions could be made mild enough to preserve most of the immobilized dye, very little enzyme activity was observed for previous urease immobilization attempts (13),so the resulting sensor had very low sensitivity. The two successful techniques-covalent attachment via silanization and cross-linking with BSA and glutaraldehyde-possess virtually identical response characteristics and sensor lifetimes. The main difference between the two techniques is ease of preparation: the cross-linking procedure takes less than 1h, with a 100%success rate, and the covalent immobilization takes all day, with a 25% success rate (successful in one of four attempts). Stopped Flow Response. The stopped flow approach was selected for all the enzymatic measurements, to minimize any noise due to movement of the sensor during the pH measurement and to maximize the response, as the transient response to a 50-rL sample volume is relatively small. The pH change was recorded for a fixed time interval, after which the carrier flow was initiated again. The maximum recorded signal was used for calibration (the signal resembled a peak and its maximum represented a fixed time measurement). To determine if the transient response to substrate could be improved by simply increasing the sample volume, the response of the sensor to injections of 1.0 and 10.0 mM penicillin G at 50-, loo-, ZOO-, and 400-jiL samples is compared (Table 11). By comparison, stopped flow typical response range of the sensor to 50 jiL of 10.0 mM penicillin G at a 10-s stop time is a decrease in A, of -0.23 to -0.33 and a decrease in A, of

sample vol, pL

1.0 mM pen G

10.0 m M p e a

50 100 200 400

-0.041 -0.051 -0.091 -0.105

-0.109 -0.119 -0.165

-0.30 to -0.38 for a 20-s stop of the flow (the response is negative with increasing acidity). For 1.0 mM, the typical response range to a 10-s stop flow is -0.10 to -0.22 A, and -0.15 to -0.22 for a 20-9 stop flow. For the transient signal (sample volume of 400 pL), the response to 1.0 mM substrate is approximately half that for a 20-s stop flow, and at a volume of 200 jiL the transient response to 10.0 mM substrate is approximately half of a 20-s stop-flow response. Thus, the immobilized enzyme reaction with a penicillin utilizing a 10-20-s stop flow gave measurement with good sensitivity, compared to the response to samples in which no stop of the flow was initiated. At the carrier flow rate of 1.2 mL/min, the “stop” is initiated 3-6 s after sample injection. Sample volume for all experiments is 50 jiL. Usually peak height is recorded, but if there is any variation in background (sample) pH, a kinetic measurement is performed (12,13),which is a third advantage of a stop-flow approach. Speed of Response. When stop-flow measurements are made, it is assumed that any response is due solely to the pH change from penicillin degradation. For this to be true, the dye response to any background must be fast enough to have already ocurred before products from the reaction impose a change in pH during the stop flow. This may be naive if the samples are of unknown or variable matrix, based on the results of the experiments with variation in buffer concentration (transient vs stop-flow signal). However, some evaluation of the relative speed of response can be performed. A t lower concentrations of sample buffer, the sensor shows a more gradual response to the sample; this response change reflects neutralization of the more dilute buffer with carrier but also the slower approach to equilibrium for dilute solutions. When the response to the O.OOO1 M citrate buffer is measured under the stopped-flow conditions used for the enzymatic reaction, steady state is achieved after 33 s, with the “stop” initiated at 2.5 s. Under the same conditions, steady state is achieved for the enzymatic stopped-flow reaction to 10.0 mM penicillin G after 40 s, with a “stop” initiated at 6.0 s. Clearly, the rate of the enzymatic reaction at even high concentration of substrate is slower than the slowest buffer response. This is supported by calibration curves that go through the origin and exhibit linearity. As the dye is obviously capable of reacting much faster than the pH change generated during the penicillin reaction, it can be concluded that the rate of dye response does not play a significant part in the kinetics of the enzyme system. Effect of Enzyme Immobilization on Dye Response. To determine the effect of enzyme immobilization on the dye response, a standard buffer is injected through a manifold containing a sensor with only the immobilized dye; the resulting peaks are compared to peaks resulting from the injection of the same volume of the buffer through a manifold containing the sensor with either cross-linked penicillinase or covalently attached penicillinase. Peak width is usually slightly increased for both immobilized systems, but an individual sensor from the cross-linked procedure may exhibit poor’ flow properties that result in slower washout of product-i.e., a greater increase in peak width. Both covalently attached and cross-linked enzyme to the pad results

1254

ANALYTICAL CHEMISTRY, VOL. 60,NO. 13, JULY 1, 1988

Table 111. Calibration Data for Penicillin V and Penicillin G Using Cross-Linked Penicillinase" [substrate], mM 50.00 37.50 25.00 18.75 12.50 9.37 6.25 4.69 3.13 2.35 1.51 1.17 0.78 0.59 0.39

blank

blank blank blank blank

-0.30

I

change in A. penicillin V penicillin G 0.295 0.296 0.287 0.270 0.279 0.270 0.259 0.232 0.208 0.165 0.134 0.089 0.073 0.047 0.037 0.003 0.013 0.001 0.005 0.001

0.289 0.293 0.285 0.279 0.280 0.278 0.259 0.227 0.195 0.159 0.130 0.093 0.072 0.043 0.033 0.005 0.008 0.010 0.002 0.002

"Flow is stopped for 2 s, 6 s after injection; change in A, is measured during the stop time.

J

//

-0.20

-0.101

0.

/

'

0.0

5.0 mM penicillin G

10.0

Flgure 8. Calibration curve for 0.3-10.0 mM penicillin 0 in carrier streams of different ionic strengths: (a) 0.1 M KCI, (b) 0.01 M KCI, and (c) 1.0 M KCI. All three carrier streams are 1 X M KH,P04, pH 6.8. Pen

0.20,

G

Table IV. Effect of Sample Volume on Immobilized Enzyme Response to 10 mM Penicillin G sample vol, p L

reflected absorbance (A,)

10 25 50 100

0.121 0.194 0.218 0.218

in approximately 25% reduction in peak height, all other variables being equal. A similar loss was also observed when urease was covalently immobilized or cross linked to the pad with Merck indicator 9583 bound to the cellulose (13). The loss in response to the buffer injections is most likely a result of the change in surface characteristics, as the reflected absorbance signal, as defined by the Kubelka-Munk equation (20),includes a scattering coefficient with a nonlinear dependence on path length. It is also possible that some fraction of the dye is simply made inaccessible due to the presence of the large enzyme molecule. Calibration Curve. T o determine the response to penicillin V and penicillin G, 50 r L of standard solution was injected into the microconduit, and the flow was stopped for 20 s, 6 s after injection (Table 111). The small blanks are probably due to a structural change in the pad when the flow is stopped relative to in a flowing stream. The first seven data points from Table I11 were plotted and the following regression coefficients were calculated for second-order polynomials: For penicillin G, y = -0.0073 + 0.1046~- 0.0130x2,r2 = 0.989; for penicillin V, y = -0.0005 + 0.0951~- 0.0093x2, r2 = 0.986. Detection limit, calculated from three times the standard deviation of the blank response, is 0.01 mM for both penicillin G and penicillin V. The relative standard deviation for the enzymatic response to substrate ranges from 1.1% to 2.1%, depending on instrumental performance. Sample Volume. T o determine the effect of sample volume on enzymatic response, lo-, 25-, 50-, and 1OO-wL volumes of 10 mM penicillin G were injected into the microconduit and allowed to react with the enzyme on the sensor for 20 s (flow is stopped 4 s after injection). Above 50 WL,there was no signal increase for the 10 mM sample (Table IV), indicating that dispersion is effectively zero a t the 50-wL injection volume and above.

AAr C 0.10

v

0.0 0.0

5.0

10.0

S u b s t r a t e , mM

Flgure 7. Response of the immobilized penicillinase to four sets of @-lactamstandards: penicillin G, penicillin V, hydroxypenicillin V, and 6-aminopeniciiianic acid. Concentrations of the standards are 0.3 mM to approximately 10 mM (4 mg/mL).

Carrier Ionic Strength a n d Buffer Capacity Effects on Immobilized Enzyme Response. The effect of increasing the carrier buffer capacity on the enzymatic response was determined by comparing calibration curves obtained in two carriers, each pH 6.8, one at 1X lo4 M KH2P04and the other M KH2P04. Results were much the same as for at 1 X soluble enzyme data the response is significantly suppressed in the carrier with a higher buffer capacity, and the shape of the calibration curve is otherwise unaffected. To determine the effect of ionic strength on immobilized penicillinase, calibration curves were obtained in carrier solutions of three M ionic strengths: 0.01, 0.1, and 1.0 M KCl, all in 1 X KH2P04,pH 6.8 (Figure 6). Clearly, the enzyme is much more susceptible to ionic strength effects than the dye, and there is also an optimum ionic strength for maximum substrate sensitivity. O t h e r Substrates for the Sensor. Penicillinase is not a specific enzyme; it responds to the entire class of penicillins, as well as the class of antibiotics known as cephalosporins. In Figure 7, the responses of the sensor to standards of the substrates penicillin V, penicillin G, hydroxypenicillin V (pen V-OH), and 6-aminopenicillanic acid (6-APA) are plotted. Clearly, the enzyme is sensitive to all these @-lactams,regardless of antibiotic activity. Differences in sensitivity can be mainly attributed to differences in substrate solubility, in preparing stock solutions (6-APA stock solution appeared cloudy). Each penicillin may also have different optimum pH, ionic strength, etc. for optimum response (21). One conclusion

ANALYTICAL CHEMISTRY, VOL. 60, NO. 13, JULY 1, 1988 I

1255

Table V. Long-Term Response of Immobilized Penicillinase Sensors to 10.0 mM Penicillin G" A.

covalent, BSA

0.051

0.0

/

// 5.0

and

covalent

date

penicillnase

penicillinase

2128 314 414

0.212 0.213 0.183

0.192 0.214 0.151

4/17

0.162

516

0.228

0.161 (0.182) 0.180

5/30 6123

0.235

7/20

0.069

10.0

Substrate, mM

Figure 8. Response of the immobilized penicillinase to two sets of Cephalosporin standards, 0.3-10 mM: (a)Cephalothin, 20-9 stop time: (b) Cephalexin, 20-s stop time: and (c) Cephalothin, 90-s stop time.

to be drawn from this data is that the sensor will have many potential interferenta if only one substrate is to be monitored, However, as a general sensor for antibiotic activity, or as a postcolumn HPLC detector, the enzyme sensor is a realistic choice. The sensor is also shown to be responsive to two common cephalosporins, Cephalothin and Cephalexin (Figure 8). The enzyme is significantly less sensitive to these substrates, probably due to lower solubility of the cephalosporins under these conditions (10 mM stock solutions were cloudy), and the possibility of the low pH of the standard solutions (10 mM standard solution pH is 4.7). When the flow is stopped for 90 s for the Cephalothin standards, sensitivity is greatly improved; the enzymatic rate appears to be much slower for this substrate, under these conditions. To use the sensor as a detector for cephalosporins, reoptimization of pH, ionic strength, and other variables should improve response to these substrates. Lifetime of Immobilized Enzyme Sensor. Enzyme Lifetime. The enzyme was immobilized in February, 1986, and the individual sensors still responded 10 months later with activities similar to those initially recorded. However, it is difficult to precisely characterize the degree of activity, due to variability in response from pad to pad (the same pad is not tested each time, just the same type of pad) (Table V). Much of the variability has been attributed to poor flow properties of the solid support after long-term storage in solution, but the lack of temperature control may also adversely affect the reproducibility of the kinetic measurements. The individual sensors are also somewhat variable in solid support volume (the cellulose is obtained by removal from a plastic backing before enzyme immobilization). While this does not affect the dye response, it can affect the amount of enzyme or albumin that will bind to an individual sensor, affecting speed of response as well as the pH change generated for a given substrate concentration. For these reasons, it is difficult to get an exact prediction of enzyme lifetime. Pad Lifetime. While the enzyme itself appears to offer indefinite lifetime, the solid support has limited usable lifetime. After the pad is stored in buffer solution at room temperature for 3 months, replicate injections of standard substrate solution show a gradual increase in peak width, without loss in peak height, until sampling frequency became unusably low. If the pumps are all stopped, and the sensor left to recover for a period of time (30 min), initial sampling rate is restored. This indicates that the fibrous cellulose pad has become *mushy", and compresses against the flow cell exit

0.233 0.162 (0.155) 0.201

cross-linked penicillinase 0.125 0.164 (0.177) 0.180 (0.175) 0.232 (0.170) 0.194 0.225

"Stop time is 10 s. bValuesin parentheses represent response of a second sensor of the same type, tested on the same day. port when carrier flows through the pad. During tests of the individual sensors on a monthly basis over the 10 months of storage, a low signal is often obtained for an individual pad, whether the pad has cross-linked or covalently immobilized penicillinase. This appears to be, again, a function of the relatively poor flow properties of the pad after prolonged storage, as response can be improved by as much as 100% by simply repositioning the pad in the flow cell. CONCLUSION Response of a sensor is a complex function of sensor qualities, of the monitored environment, and of the mode in which sensor and monitored species are brought together. Manual testing in a batch mode is inadequate and inefficient because it lacks the exact timing and reproducibility of events. Automated FIA provides well-reproduced conditions by allowing the sensor to be tested in a flowing stream of a carrier, into which well-defined zones of testing material are periodically injected. This allows response properties of a sensor to be exactly characterized. Such use of FIA is not limited to optical sensors. It has been used previously to test response of ion-selective electrodes (22,23), microelectrodes (24),and ion-selective field effect transistors (ISFET's) (25). Recently, the surface treatment of glassy carbon electrodes was evaluated by means of the voltammetric response observed in an FIA mode (26). The present work is an extension of our introduction of several aspects of the FIA technique for a systematic evaluation and development of a chemical sensor (13). The long-term mechanical properties of the solid support are not as good as when the enzyme is immobilized on a glass support (27). Variability in month-to-month reproducibility might be improved, possibly by a more uniform immobilization technique as well as a more stable support. The sensitivity and of the sensor is comparable to that for a pH ISFET (B), close to the sensitivity reported for potentiometric methods (27, 29) and thermistor methods (30),but low relative to colorimetric methods used in conjunction with HPLC detectors (31-34). Enzyme loading was not examined, and sensitivity could possibly be enhanced by use of a higher enzyme concentration during the immobilization procedure. Registry No. Penicillinase, 9001-74-5; penicillin G, 61-33-6; penicillin V, 87-08-1; hydroxypenicillin V, 20880-67-5; 6-aminopenicillanic acid, 551-16-6; cephalothin, 153-61-7; cephalosporin, 11111-12-9; cephalexin, 15686-71-2; penicillin, 1406-05-9. LITERATURE CITED (1) Ruzicka. J.; Hansen, E. H. Anal. Chim. Acta 1986, 179, 1. (2) Ruzicka, J. Anal. Chern. 1983, 55, 1040A

1256

Anal. Chem. 1988, 60,1256-1260

(3) Yerian, T. D.; Christian, G. D.; Ruzicka, J. Analyst (London) 1986, 1. 1. 1. , S -f-35-. (4) Ruzicka, J.; Hansen, E. H. Anal. Chim. Acta 1985, 173,3. (5) Woods, B. A.; Ruzicka, J.; Christian, G. D.; Rose, N. J.; Charlson, R. J. Analyst (London) 1988, 113,301. (6) Schwartz, M. A.; Buckwaker, F. H. J . Pharm. Sci. 1962, 5 1 , 1119. (7) Perrin. D. D.: Dempsey, B. Buffers for pH and Metal Ion Control; Chapman and Hall: London, 1974. (8) Adams, R. E.; Carr, P. W. Anal. Chem. 1978, 50,944. (9) Weibel, M. K.; Dritschilo, W.; Bright, H. J.: Humphrey, A. G. Anal. B o chem. 1973, 52, 402. (IO) Tran-Minh, C.;Broun, G. Anal. Chem. 1975, 4 7 , 1359. (11) Neisius, K.: Baumer, W. US. Patent No. 4029598, July 14, 1977. (12) Yerian, T.; Christian, G. D.; Ruzicka, J. Ana/yst (London) 1986, 7 7 7 . 865. (13) Yerian, T.; Christian, G. D.; Ruzicka, J. Anal. Chim. Acta, in press. (14) Bates, R. G. Determination of p H , Theory and Practice: Wiiey: New York, London, Sydney, 1964. (15) Bishop, E. “Indicators”, International Series of Monographs of Analytical Chemistry; Pergamon: Oxford, New York, 1972; Vol. 51. (16) Yerian. T. P h D dissertation, 1987. (17) Culien, L. F.;Rusling, J. F.: Schieifer, A.: Papariello, G. J. Anal. Chem. 1974, 4 6 , 1955. (18)Messing, R. A. Biotechnology and Bioengineering; Wiley: New York, 1974; Vol. XVI, pp 1419-1423.

(19) Saari, L. A.; Seitz, W. R. Anal. Chem. 1984, 56,810. (20) Mitsugi, K.; Mimura, N. Enzyme Engineering; Plenum: New York, London, 1978; Voi. 4. (21) Kubelka, P.; Munk, F. 2. Tech. Phys. 1931, 12, 593. (22) Hou, J. P.: Poole, J. W. J . Pharm. Sci. 1972, 6 1 , 1549. (23) Hansen, E. H.; Krug, F. J.: Ghose and, A. K.;Ruzicka. J. Ana/yst (London) 1977, 102, 705. (24) Xie, R. Y.; Gadzekpo, V. P. Y.; Kadry, A. M.; Ibrahim, Y. A.; Ruzicka, J.; Christian, G. D. Anal. Chim. Acta 1986, 184, 259. (25) Haemmeric, A.; Janata, J. Anal. Chim. Acta 1982, 144, 115. (26) Smith, R.: Huber, R. J.; Janata, J. Sens. Actuators 1984, 5 , 127. (27) Wang, J.; Tuzhi, P. Anal. Chem. 1986, 58, 1787. (28) Gnanasekaran, R.: Mottola, H. A. Anal. Chem. 1985, 57,1005. (29) Caras, S.;Janata, J. Anal. Chem. 1980, 52, 1935. (30). Rusling, J. F.; Luttrell, G. H.; Cuiien, L. F.; Papariello, G. J. Anal. Chem. 1976, 4 8 , 1211. (31) DeCristoforo, G.; Danielsson, B. Anal. Chem. 1984, 56,263. (32) Hornby, W. E.; Campbell, J.; Inman, D. J.; Morris, D. L. Enzyme Engineerlng, Plenum: New York and London, 1974; Vol. 2, p 401. (33) Haginaka. J.; Wakai, J. Anal. Chem. 1988, 58, 1896. (34) Haginaka, J.; Wakai, J. Anal. Chem. 1985. 57,1568.

RECEIVED for review July 28, 1987. Accepted February 16, 1988.

Fabrication and Characterization of a Fiber-Optic-Based Spectroelectrochemical Probe David A. Van Dyke and Hung-Yuan Cheng* Physical and Structural Chemistry, Smith Kline and French Laboratories, Philadelphia, Pennsylvania 19101

Probes were constructed from fused slllca optical fibers embedded in electrlcatly conductive graphite/epoxy material with the optical end face and the working electrode active surface In a coplanar arrangement. The electrochemlcal properties and spectroelectrochemical response of the microprobe were characterized In solutions and In gels containing UV-absorblng oxidizable compounds. The concept of eiectrochemlcal modulation of spectral signals for fiber-optlc-asslsted spectroscopy was demonstrated by using the ascorbate/dopamlne gel model. The validity of such an approach for spectroelectrochemical measurements In tissue was evaluated by using Isolated animal brain. Further development and possible applications are discussed.

The combination of optical spectroscopy and electrochemistry, often termed spectroelectrochemistry, has provided powerful tools for probing complex redox processes near the solution-electrode interface ( 1 , 2 ) . Most applications of spectroelectrochemistry have been carried out in relatively large cells and often with a rigid geometry to accommodate the conventional optical arrangements. The recent development of fiber-optic-based chemical sensors ( 3 , 4 )suggests that such capabilities as remote sensing, miniaturization, on-line monitoring, flexibility, and in situ measurement may also be exploited in spectroelectrochemical applications. Optical fibers have been used for achieving long optical path length in thin-layer electrochemistry (5), for collecting scattered light in Raman spectroelectrochemistry (6),and for near-normalincidence-reflection measurements off the reflective surface of a platinum or mercury electrode (7). In all these cases the optical fibers, although placed very close to the electrode surface, were physically separated from the electrode.

We report here the fabrication and characterization of a small spectroelectrochemical probe intended for in situ measurement in hard-breach environments or small volumes. The design strategy is to place the optical fiber end face and the working electrode active surface on the same plane by physically merging the two materials in a coaxial arrangement. In such a configuration one not only can achieve the compactness and flexibility necessary for an in situ probe but also can gain temporal and spatial advantages by placing the electrochemical diffusion layer a t the immediate vicinity of the optical fiber end face. These points will be demonstrated employing the ascorbate/dopamine model in an artificial gel matrix. Possible applications in biological systems will be discussed using spectroelectrochemical measurements in brain tissue as an example. EXPERIMENTAL SECTION Optical Equipment. The modular UV-vis spectrometer, obtained from Oriel (Stratford, CT), consisted of a 150-W Xe arc source, a stepper-motor-driven grating monochromator (200-800 nm), an UV-to-near-IR photomultiplier detection system, and a computer interface for an Apple IIe microcomputer. The spectrometer was operated in the single-beam mode. A number of bifurcated W-vis spectroelectrochemical probes were constructed from 100- or 200-pm fused silica optical fibers (Polymicro Technologies, Phoenix, AZ), as described in detail in the results section. Spectral measurements were carried out in the “Y-guide” configuration; Le., one arm transmitted light to the sample from the source, and the other arm carried light back to the detector. T h e source beam was first passed through the monochromator and then focused onto the distal end of the source fiber by means of an Oriel Model 77290 lens-filter-shutter attachment; a second Model 77290 was used to focus the light from the detection fiber onto the photomultiplier tube. (The distal ends were mounted in separate home-made aluminum cylinders,to allow them to be accurately positioned in front of the source and detector.) Most

0003-2700/88/0360-1256$01.50/00 1988 American Chemical Society