BIOMEDICAL CHROMATOGRAPHY, VOL. 5, 180-183 (1991)

High Performance Liquid Chromatography of Glucose using a Post-column Reactor of Immobilized Enzyme followed by Electrochemical Detection (Glucose-LCEC)

Noriyuki Watanabe Department of Industrial Chemistry, Faculty of Engineering, University of Tokyo, Hongo 7-3-1, Bunkyo-ku, Tokyo 113, Japan

A sensitive and selective, reasonably fast method for the determination of glucose content has been developed. A glucose oxidase immobilized column was coupled to a small-size anion exchange columnlborate buffer chromatograph. The hydrogen peroxide produced in the enzyme reaction was detected directly by an amperometric detector using a platinum working electrode. The detection Limit was 0.03 ppm (1.5 X lo-' M, 3 pmollinjection). The linear dynamic range was three orders of magnitude at least. The system was stable and reproducible both in short- and long-term operation. The proposed method is suitable for analysis of complicated matrices of biological samples because of its good selectivity and sensitivity.

INTRODUCTION

Glucose is an important analyte in biological and indus- trial fields. Most of the chemical methods for the determination of glucose content are based on their reducing properties. The detection of glucose (or generally carbohydrates) by high performance liquid chromatography (HPLC) has mainly relied upon these methods (Mooper and Degens, 1972; Mooper and Gindler, 1973; Honda et al., 1980; Watanabe and Inoue, 1983). However, the selectivity of these meth- ods was not necessarily high enough because of inter- ference due to other reducing substances present in the sample. To enhance the selectivity, utilization of enzyme-coupled reactions is a possible alternative. Enzymes (e.g., glucose oxidase or glucose dehydroge- nase) have been used in the form of either an enzyme electrode (Lobe1 and Rishpon, 1981; Lange and Chambers, 1985; Gregg and Heller, 1990) or an immo- bilized enzyme reactor (Emneus and Gorton, 1990). However, their application has been limited to flow- injection analysis, not to HPLC. Amperometric detec- tion is so highly sensitive and selective that it has been successfully used in HPLC for the analysis of biological samples. In spite of enzyme specificity, amperometric detection with an enzyme electrode in the flow- injection mode is prone to the influences of bulk changes in effluent composition due to sample injec- tion. In addition, glucose is not the only substrate for glucose oxidase. For example, 2-deoxyglucose is also quite a good substrate. These points leave some ambi- guity in the flow-injection analysis results. Hence, sepa- ration by HPLC is desirable for glucose analysis of biological samples.

In this study an immobilized glucose oxidase column was used as the post-column reactor between an analy- tical column and an amperometric detector. From the enzyme reaction with glucose, the reactor produced hydrogen peroxide which was directly detected on a platinum electrode placed downstream. This scheme

0 1991 by John Wiley & Sons, Ltd. 0269-3879/91/040180-04 $05.00

was recently utilized by the author in polyamine analy- sis (Watanabe et al., 1989). The combination of an anion exchange column and borate buffer as eluent provides the most effective separation of glucose from other components. However, it takes a long time for glucose to elute in this mode with the conventional (commercially available) size of analytical column- typically 60 min for 4.6 mm i.d., 15 cm column length, 0.5 M borate buffer as eluent. To shorten the analysis time, a small-size column was prepared.

EXPERIMENTAL

Materials. Glucose oxidase (A . niger) was supplied from Biozyme Laboratories (Blaenavon, South Wales, UK) and used as received. Preactivated gels, Eupergit C and Reacti-gel were purchased from Rhom Pharma GrnbH (Darmstadt, Germany) and Pierce Chemical Co. (Rockford, IL, USA), respectively. Anion exchange gel Sugar A X I was supplied from TOSOH Co. (Tokyo, Japan). All other chemi- cals were analytical grade. Water was deionized and then single-distilled in presence of alkaline potassium permanga- nate. The standard solution of glucose was prepared each day from successive dilution of the stock solution which was made up to 10 mM once a week and stored in a refrigerator.

Apparatus. The apparatus used in this work is shown in Fig. 1. A computer-controlled pump CCPM (TOSOH Co.), an amperometric detector LC4B with a platinum working elec- trode (Bioanalytical Systems, West Lafayette, IN, USA) and a sample injector, model 7125 with a 20 pl loop (Rheodyne, Berkefey, CA, USA), were used. The applied potential was fixed at +SO0 mV throughout this study. Analytical columns were prepared with anion exchange TSK gel Sugar AX1 (8 p, TOSOH Co.) packed into a stainless steel column (2 rnm or 1 rnm i d . , 5 crn length). The analytical column was installed in an oven RE-8000 (TOSOH Co.) with the temperature controlled at 55 "C. The enzyme column (1 mm i.d., Teflon

Receiued 9 July 1990 Accepted 11 October 1990

DETERMINATION OF GLUCOSE USING GLUCOSE-LCEC 181

injector, 20p1 rpZ-l-9

column

oxidase glucose eluent

0.2M borate buffer PH 8.5

Figure 1. Diagram of the apparatus used for glucose-LCEC.

tubing) was placed between the analytical column and the amperometric detector. Borate buffer (0.2 M, pH 8.5) was used as eluent. The eluent was filtered with a membrane filter (0.45pm) before use. The flow rate of eluent was fixed at 0.2 mL/min except for the study of flow rate dependence. The enzyme columns were stored in a refrigerator every working day after the experiments had been carried out.

Enzyme immobilization. Two kinds of preactivated gel (Eupergit C and Reacti-gel) were examined. (i) Eupergit C is a polyacrylamide gel with an epoxide group as the active site. 4 mL Eupergit C was successively washed with 50 mL cooled water and 50 mL cooled 0.2 M phosphate buffer (pH 6.1) on a glass filter. The gel was then dispersed in 10mL 0 . 0 5 ~ phosphate buffer (pH 6.1) containing about 160 mg glucose oxidase for 120 h at 5 "C. After the enzyme immobilization, the gel was washed with 0.2 M phosphate buffer (pH 6.1) and dispersed in 10mL of the same buffer containing 500mg tris(hydroxymethy1)aminomethane for 48 h at 5 "C. Finally the enzyme-bonded gel was thoroughly washed with 0.2 M

phosphate buffer and water and then stored in water contain- ing 0.05% sodium a i d e in a refrigerator before use. The enzyme-immobilized gel was packed into Teflon tubing (1 mm i.d., 1.5 mm o.d., various lengths 3.5-28 cm), stopped with small balls of Teflon fibre and short pieces of Teflon tubing of smaller outer diameter at both ends as shown in Fig. 2. (ii) Reacti-gel (HW-65) is preactivated agarose gel derivatized with 1,l'-carbonyldiimidaole. 3 mL gel was washed with 50 mL cooled water and 10 mL cooled 0.2 M carbonate buffer (Na,C03/NaHC03 = 3 : 7, pH 10.2) on a glass filter, and then suspended in 10 mL 0.2 M carbonate buffer containing 160 mg glucose oxidase for 50h at 5°C. After washing with the

enzyme immobilized gel

U fiber 1 mm i.d. PTFE tube

Figure 2. Diagram of the enzyme-immobilized gel column.

buffer, the gel was dispersed in 10mL of the same buffer containing 500 mg tris(hydroxymethy1)aminomethane for 48 h at 5 "C. Finally, the gel was washed thoroughly with the carbonate buffer and water and then stored in water contain- ing 0.05% sodium azide in a refrigerator before use. The gel was packed in the same way as Eupergit C.

Sample preparation. (i) Human urine. Immediately after collection, 1 mL urine was mixed with 1 mL acetonitrile followed by centrifugation for 10 min at 4500 g. A portion of the supernatant was diluted 20-fold with the eluent and injected into the high performance liquid chromatograph. (ii) Fruits (strawberry, Japanese summer orange and kiwi). After homogenization and centrifugation (for 10 min at 4500 g), the supernatant was mixed with an equivolume of acetonitrile. The mixture was centrifuged for 10min at 4500g and a portion of the supernatant diluted to 1/1OOO with water and to 1/10 with the eluent, successively, and then the diluted sample (corresponding to 1/20,000 of the original homoge- nized fruit) was injected into the high performance liquid chromatograph.

Dependence of activity on pH and buffer concentration. The activity of the enzyme immobilized on gel was examined at various pH values by the flow injection mode (without analy- tical column). The mobile phase used was 0.05 M phosphate buffer with changing pH at a flow rate of OSmL/min. The effect of buffer concentration was also examined by using phosphate buffer (pH 8.0, K2HP04/KH2P04 = 95 : 5 , 0.025- 0.4 M). A one ppm (20 ng/20 pL injection) solution of glucose prepared with each mobile phase was injected.

RESULTS AND DISCUSSION ~

The two kinds of preactivated gel examined provided almost the same results. Hence, only the results obtained with Eupergit C are described in the rest of this report.

The first concern was how the activity of the immobi- lized glucose oxidase depended on pH. It is well known that the native enzyme has optimum activity at pH 5.6 (Nakamura and Ogura, 1962; Gibson et al., 1964), whereas the eluent in the anion exchange column/ borate buffer mode has a certain fixed pH range, that is pH8.0-9.0. Figure 3 shows the pH dependence obtained in the flow-injection mode. The optimal pH of the immobilized enzyme shifted to 6.8 from the 5.6 of the native enzyme. The shifts of optimal pH because of immobilization have been recognized in many cases so far (Levin et al., 1964; Boppana et al., 1986). Fortunately, the immobilized enzyme still retains much of the activity appropriate for chromatographic sepa- ration of glucose even at pH 8.5.

The influence of the eluent buffer concentration on enzyme activity was examined with phosphate and borate buffer in the flow-injection mode. A slight decline at higher phosphate content was observed, as shown in Fig. 4. The tendency was almost the same for borate buffer.

Various parameters such as borate concentration, pH and column temperature interact as determining factors for the elution of glucose in the anion exchange/ borate buffer mode. The retention time of glucose became longer and the elution profile sharper with increasing column temperature from room temperature

182 N . WATANABE

PH

Figure 3. Peak response against pH of .the mobile phase in flow-injection mode (0.1 M phosphate buffer, 0.2 mllmin).

to 80 "C. Higher borate concentration made the peak sharper and the elution faster. However, higher borate concentration often caused problems in the operation of the pump due to solidification of the solute. An increase in eluent pH played a parallel role in terms of elution power to an increase in borate concentration. However, the use of higher pH was limited on account of the decrease in enzyme activity at higher pH. A mutual compromise among these factors made us choose 0.2 M borate buffer at pH 8.5 as the eluent and 55 "C as the column temperature.

Figure 5 shows chromatograms of the standard sam- ple obtained with and without the enzyme column. No peaks corresponding to glucose were observed without the enzyme column. Comparison of Figs 5(b) and 5(c) gives a quick indication of the detection limits of the met hod.

As an example of a substrate other than glucose, Fig. 6 shows a chromatogram obtained for 2-deoxyglucose. The relative response of 2-deoxyglucose was nearly equivalent to glucose in the flow-injection mode.

nA

30

20

10

, 0.1 0.2 0.3 0.4

M

Figure 4. Peak response against buffer concentration of the mobile phase in flow-injection mode (phosphate buffer, pH 8.0, 0.2 mL/min).

Figure 5. Effectiveness of the enzyme column. (a) Without enzyme column, (b) and (c) with enzyme column; (a) and (b) 1 ppm glucose, (c) 0.1 ppm glucose. Analytical column: Sugar AXI, 2 mm i.d. x 5 cm, 55 "C; 0.2 M phosphate buffer, pH 8.5, 0.2 mllmin. Enzyme column: 1 mrn i.d. x 14 cm.

However, 2-deoxyglucose was chromatographically completely separated from glucose. The enzyme col- umn did not react with the other mono- and di- saccharides examined (fructose, mannose, galactose, xylose, sucrose, maltose and lactose).

The peak response at different flow rates was plotted against enzyme column length as in Fig. 7. From the comparison of the detector response for glucose with that for a known amount of hydrogen peroxide exa- mined in flow-injection mode, it was estimated that the enzyme column (28 cm length) converted approxima- tely 45% of glucose to hydrogen peroxide at a flow rate of 0.1 mL/min. The enzyme reaction did not complete even at a flow rate of 0.05 mL/min with a column length of 28 cm. However, this was not a problem in practice.

The detection limit was 0.03 ppm (1.5 x M, 0.6ng or 3pmol injection) at S/N=3. The response depended linearly on glucose concentration from 0.1 to 100 ppm (5 x to 5 x M). Reproducibility, both in shofl- and long-term operation, was quite good. The coefficient of variance for repeated injections of 1 ppm glucose (n = 10) intraday was 1.7%. Interday variation over five days (n = 10) was 4.5%. The standard sample was prepared fresh each day. Considering the uncer-

Figure 6. Example of a substrate other than glucose which reacts with the enzyme column. (a) Blank, (b) 10 ppm glucose, (c) 10 ppm 2-deoxyglucose. The chromatographic conditions were as in Fig. 5. Enzyme column: 1 mm i.d. x3.5 cm.

DETERMINATION OF GLUCOSE USING GLUCOSE-LCEC 183

nA 1 I 2nA I

7 14 28 column length, cm

Figure 7. Peak response against length of the enzyme column at various flow rates. The chromatographic condition were as in Fig. 5; 1 ppm (20 ng/injection) glucose.

tainty involved in the sample preparation, the value for interday variation was satisfactory. Durability of the enzyme column was excellent. Loss of enzyme activity was not noticeable after continuous usage for three months and even after intermittent work over a year.

Figure 8. Chromatograms for real samples obtained with glucose-LCEC. The chromatographic conditions were as in Fig. 5. (a) Standard (1 ppm glucose), (b) strawberry, (c) Japanese summer orange, (d) kiwi, (e) human urine (1/40 dilution); (b)- (d), 1/20,000 dilution. Enzyme column: 1 mm i.d. x 14 cm.

Figure 8 shows an example of chromatograms obtained for homogenized fruits and human urine. One sample can be analysed within 15 min. Recently, micro- dialysis has been emerging as a new bioanalytical sam- pling technique with remarkable potential (Sharp et al . , 1987). The typical sampling time is 10-20 min to which the method described here is quite compatible. We are planning a direct coupling of glucose-LCEC to micro- dialysis as an attractive application.

REFERENCES

Boppana, V. K., Fong, K. L., Ziemrniak, J. A. and Lynn, R. K.

Emneus, J. and Gorton, L. (1990). Anal. Chem. 62, 263. Gibson, 0. H., Swoboda, B. E. P. and Massey, V. (1964). J. Bid.

Gregg, B. A. and Heller, A. (1990). Anal. Chem. 62, 258. Honda, S., Matsuda, Y.,Takahashi, M., Kakehi, K. and Ganno,S.

Lange, M. A. and Chambers, J. Q. (1985). Anal. Chim. Acta 175,

Levin, Y., Pecht, M., Goldstein, L. and Katchalski, E. (1964).

(1986). J. Chromatogr. 353, 231.

Chem. 239,3927.

(1980). Anal. Chem. 52, 1079.

89.

Biochemistry 3, 1905.

Lobel, E. and Rishpon, J. (1981). Anal. Chem. 53,51. Mooper, K. and Degens, E. T. (1972). Anal. Biochem. 45, 147. Mooper, K. and Gindler, E. M. (1973). Anal. Biochem. 56, 440. Nakamura, T. and Ogura, Y. (1962). J. Biochem. (Tokyo) 52,214. Sharp, T., Zetterstrom. T., Series, H. G., Carlsson, A.,

Graharne-Smith, 0. G. and Ungerstedt, U. (1987). Life Sci. 41,869.

Watanabe, N. and Inoue, M. (1983). Anal. Chem. 55, 1016. Watanabe. N., Asano, M., Yamamoto, K., Nagatsu, T.,

Matsumoto, T. and Fujita, K. (1989). Biomed. Chromatogr. 3, 187.