ANALYST, JUNE 1991, VOL. 116 569 Automated Enzyme Packed-bed System for the Determination of Vitamin C in Foodstuffs Simon Daily, Susan J. Armfield,* Barry G. D. Haggettt and Mark E. A. Downs Laboratory of the Government Chemist, Queens Road, Teddington, Middlesex TWI I OLY, UK A microprocessor controlled flow injection system is described for the determination of vitamin C in foodstuffs. The system is based on amperometric detection at a wall-jet electrode coupled with an ascorbate oxidase packed bed. A commercially available Cartesian robotic auto-sampler-dilutor is used as a means of fully automating the sample handling and dilution. Dithiothreitol (DTT) is used to reduce dehydroascorbic acid t o ascorbic acid and t o stabilize ascorbic acid standard solutions. Initially, the system was connected in series with a high-performance liquid chromatography column and ultraviolet (UV) detector t o allow identification of possible interferents and to allow comparative evaluation of results.The system showed a linear response t o the concentration of L-ascorbic acid in the range 1-200 pg ml-1 and was capable of detecting total vitamin C in a range of foodstuffs at a sample throughput of 15 samples h-1. Correlations t o existing methods of 0.98 were obtained. Keywords: ,Vitamin C determination; enzyme packed bed; ascorbate oxidase; amperometric detection; automated food analysis The physiological roles of vitamin C have been widely reported. 1 The biologically active compounds are L-ascorbic acid (AA) and L-dehydroascorbic acid (DHAA), while vitamin C refers to the sum of these two forms.In addition to occurring naturally in a wide range of foods, vitamin C is commonly added to foodstuffs as an anti-oxidant. The most commonly used chemical methods for the determination of vitamin C have been based on one of three classical methods. The first method relies on the oxidation of AA to DHAA which is derivatized with 2,4-dinitrophenyl- hydrazine and then determined spectrophotometrically.2 The second method is based on the colorimetric titration of AA with 2,6-dichloroindophenol [4-(2,6-dichloro-4-hydroxy- phenylimino)cyclohexa-2,5-dienone].3-~ The third method involves the measurement of a fluorophor produced from the treatment of DHAA with o-phenylenediamine . 3 5 These methods, although they have been adapted to semi-automatic flow injection analysis,3.6 have proved to be complicated, time consuming and non-specific.More recently methods for the determination of AA and DHAA based on HPLC have been developed .7 However, problems with non-specific absorption have limited the applications of the ultraviolet (UV) detec- torss-9 originally employed. These problems have led to the development of dttection by electrochemical methods which offer substantial increases in both sensitivity and selectivity when compared with UV detection.l0-11 In this work an attempt has been made to develop the enzyme reactor bed method reported by Bradberry and Adams.12 The method relies on the sensitivity and selectivity of electrochemical detection coupled with the biological elimination of AA by the enzyme, ascorbate oxidase (AO, E.C.1.10.3.3).Initially, an aliquot of each food sample was taken and divided into two. The first aliquot was passed through an enzyme packed bed that had been previously heat- denatured. An amperometric signal, proportional to the amount of AA plus other electro-oxidizable species (interfer- ents) present in the sample, was produced at the detector. The second aliquot was then passed through a similar bed containing active enzyme where AA was selectively removed giving a second signal due only to the interfering species. The difference between the two signals generated at the detector * Present address: DTI, Environment Unit, 151 Buckingham Palace Road. London SWlW 9SS, UK. t Present address: Applied Research and Consultancy Centre, Putteridge Bury, Hitchin Road, Luton, Bedfordshire LU2 SLE, UK.electrode can be related to the concentration of ascorbic acid. This approach is summarized as follows: At the electrode Glassy carbon AA - + 850 mV versus Interferentsred - Ag- AgC1 In the packed bed A 0 2 A A + 0 2 ---+ Then at the electrode Glassy carbon Interferentsred ---+ 850 mV versus Ag-AgCI DHAA+2H++2e- + Charge = Qtotal [nterferents,,+ ne- 2 DHAA + 2Hz0 Interferents,, + ne- Charge = Qi The charge due to AA can then be estimated from the difference between the two charges measured QAA = Qtotal - Qi In the work reported here, vitamin C is determined after the DHAA present in the sample is converted to AA by extraction of the original samples with a solution of dithio- threitoll3, which also serves to stabilize the AA.l4,1S During the preliminary stages of system development, a high-performance liquid chromatography (HPLC) column and UV detector were connected in series with the enzyme bed and electrochemical detector in order to facilitate identification of possible interferents and to generate com- parative results.Experimental Reagents Mobile phase. Potassium dihydrogen orthophosphate (15 mmol dm-3, pH 5.0) was prepared by dissolving 2.04 g of H2P04 (Fisons, Loughborough, UK) in 1000 ml of purified water. The solution was then filtered and de-gassed with helium.570 ANALYST, JUNE 1991. VOL. 116 Extracting solution. Dithiothreitol (DTT) (1 mmol dm-3, Sigma, Poole, Dorset, UK) was prepared daily by dissolving 0.154 g of DTT in 1000 ml of the mobile phase. Ascorbic acid standard solution.A 200 pg ml-1 (1.136 mmol dm-3) stock solution was prepared by dissolving 20 mg of AA (Fisons) in 100 ml of the extracting solution. A stock solution was prepared daily and automatically diluted to the range 1-20 pg ml-1 when required. Ascorbate oxidase (250 U mg-1). This enzyme from Curcurbita sp. was obtained from Borhringer (Lewes, Sussex, UK). (1U = 16.67 nkat). Procedures Preparation of immobilized ascorbate oxidase Aminopropyl controlled-pore glass beads (600 mg, 125-177 pm, pore diameter 500A, Pierce, Chester, UK) were added to 6 ml of a 2.5% (v/v) aqueous solution of glutaraldehyde (Sigma) and mixed on a roller mixer for 10 min. The slurry was de-gassed in a freeze-drying chamber at -40 "C for 30 min then replaced on the roller mixer for a further 1 h at 25 "C.After washing with distilled water, the slurry was again de-gassed in the freeze-drier for 30 min. The excess water was removed and A 0 (10 mg) dissolved in 10 ml of 0.1 mol dm-3 tris(hydroxymethy1)aminomethane hydrochloride buffer (pH 8) was added. The mixture was kept on the roller mixer at 4 "C over-night after which the non-immobilized enzyme was removed by washing with distilled water. The immobilized enzyme was packed into a flow-through Kel-F bed (2 mm i.d. X 40 mm, Oxfq-d Electrodes, Abingdon, UK) held by 20 pm stainless-steel mesh frits. When not in use the bed was stored refrigerated at 4 "C. A second bed, used as a control, was packed with controlled-pore glass prepared as above with heat-denatured AO.Sample extraction For this study, four foodstuffs were analysed extensively. Samples containing an estimated 50 pg of vitamin C were weighed, i.e., orange juice (1.0 g), grapefruit juice (1.0 g), instant mash potato powder (0.5 g) and freeze-dried brussel sprouts (0.1 g). The samples were shaken with extracting solution (30 ml), filtered (grade 541, Whatman, Maidstone, Kent, UK) and bulked to 50 mi. Each extract was then re-filtered through a syringe filter (0.2 pm Acrodisk, Gelman, Northampton, UK) directly into a 2 ml amber glass auto- sampler vial (Chromacol, London, UK) which was then sealed. Analysis was initiated immediately in order to minimize sample decay. Instrumentation The apparatus is shovw in Fig. 1. The experiment was controlled from a personal computer (Olivetti M24SP).Experimental parameters were entered at the computer and passed to an interface rack (Imperial College Chemistry Microprocesor Unit, London, UK) via an RS232 serial link. The rack was of a modular design, based on a Z80 micro- processor controlling an internal potentiostat, timer, digital to analogue and analogue to digital converters for the monitoring of the electrochemical detector. It also had control over sample handling via a second RS232 port to an autosampler (Gilson 222, Anachem, Luton, UK) and dilutor (Gilson 401, Anachem), and two six-port pneumatically operated injection valves (Rheodyne 7010P, Anachem). The autosampler was used for automatic dilution and for filling of the injection loop with calibrant and sample solutions, and operated indepen- dently of the rack. This arrangement allowed the dilution of one sample while the rack monitored another.One six-port valve was set up as a conventional injection valve with a 20 PI loop and the other as a two-column selector for switching the Fig. 1 Experimental apparatus for the determination of vitamin C. Chromatographic equipment in the shaded area was used initially to identify interferents and to obtain comparative results. WE, Working electrode; RE, reference electrode; and CE. counter electrode Fig. 2 Wall-jet electrode sample stream either through the enzyme packed bed or through the dummy bed containing inactive enzyme. The controlling software for both the microcomputer and the interface rack was written in Pascal (Prosper0 Software, London, UK), the rack program being cross-compiled for the 280 processor and loaded serially to the rack.The rack program controlled the systems hardware, monitored the electrochemical detector and passed data to the microcom- puter for data handling (smoothing and integration) , storage and presentation. An HPLC pump (LKB 2150, Pharmacia, Milton Keynes, UK) delivered the mobile phase at a constant flow rate of 1 ml min-1 to the detector, which was a glassy carbon wall-jet electrode as shown in Fig. 2. The potential of the electrode was 850 mV versus Ag-AgC1 (saturated KCl). A platinum counter electrode was situated in the outlet stream. The electrodes and cell block were purchased from Oxford Electrodes, Abingdon, UK. Initially the electrochemical cell w$s connected in series with an HPLC column (PL-SAX, 1000A, 8 pm, 150 mm X 4.6 mm, Polymer Laboratories, Church Stretton, Shropshire, UK) and a UV/VIS detector (Spectra-Physics, Model SP8450, St Albans, UK) at 251 nm.The outlet from the UV detector was connected directly to the electrochemical cell. Tubing with & in 0.d. x 0.03 in i.d. was used. Results and Discussion Stability of AA The accurate measurement of vitamin C has always been hampered by its instability, as it is readily oxidized and is photosensitive, especially in dilute solutions. Therefore , a variety of physical methods as well as antioxidants and chelating agents have been used to increase its ~ t a b i l i t y . ~ ~ ? ' ~ The use of DTT has been reported not only as a reducing agent of DHAA but also as a stabilizer of AA.Much of theANALYST, JUNE 1991, VOL. 116 571 work on the stabilization of AA was concerned with preserv- ing standard solutions or reference samples for days to weeks. For the present study only stabilization over the few hours required by long automated runs was of concern. Repeated over-night sampling of standard solutions of AA with and without DTT at various concentrations showed that a 0.1 mol dm-3 solution of DTT would stabilize a 100 pg ml-1 solution of AA for more than 10 h (Fig. 3) without appreciable deterioration in the concentration of AA. A similar sample with no DTT decayed to about 30% of its original value of AA over the same period. This stabilization should provide more than adequate time in which to carry out most analyses without significant loss of AA in standard solutions. Determinations of Vitamin C Values for the concentration of vitamin C found in the four food samples studied were initially determined with the UV and electrochemical detectors simultaneously by using chro- matographic separation.The chromatographic apparatus (shown in the dotted area in Fig. 2) was then removed and identical samples were determined by use of the switched enzyme bed technique with electrochemical detection. These results were compared with values obtained using two methods given by the Association of Official Analytical Chemists (AOAC) (titration with 2,6-dichloroindophen013.~ and a semi-automated method based on o-phenylenedi- amine3-6). The results obtained are summarized in Table 1.The electrochemical detector displayed higher specificity towards AA than the UV detector but also showed a higher sensitivity to DTT. Fig. 4 shows outputs from the UV and electrochemical detectors for a sample containing 14 pg ml-1 of AA and 1 mmol dm-3 of DTT. In addition to the difference in the ratio of AA to DTT, it is also interesting to note the reduction in interfering peaks obtained using the electrochem- ical method. Removal of the chromatographic system considerably decreased the sample analysis time from 10-12 min to 3 min. A linear calibration graph for L-AA without HPLC separation was produced in the range 1-200 pg ml-1 with a correlation 70 I 1 A 30 - 20 - 10 0 1 2 3 4 5 6 7 8 9 10 1 1 Time/h Fig. 3 0.1 mmol dm--7 DTT; and B, no DTT Decay of L-ascorbic acid (100 pg ml-*) in the presence of: A.coefficient of 0.995. Fig. 5 shows a typical output from a sample of instant mash potato after passing through the inactive and active bed using this approach. Enzyme Packed-bed Efiiciency and Stability The efficiency of the packed bed was estimated by determin- ing the AA from a calibration series with both the UV and electrochemical (EC) detectors. The samples were first passed through a dummy bed containing no A 0 and then through the active enzyme (AO) packed bed to remove AA. Plotting the responses of the two detectors against each other for both calibration series and comparing gradients gave an estimation of the efficiency of the immobilized enzyme of 99.9% (Fig. 6). This efficiency for the conversion of AA was reproduced with all the batches of immobilized enzyme that were prepared and occurred at up to 200 pg ml-1 of AA, far greater than that likely to be found in any foodstuffs.An excellent correlation between the two methods of detection (Y = 0.994) was observed. The immobilized enzyme remained fully active for 6 , 0.12 f O.I0 2 2 5 0.08 0.06 0 200 400 600 80010001200 Ti me/s Fig. 4 Chromatograms of instant mash potato using ( a ) UV and (b) electrochemical detection showing peaks for 1, D I T and 2, AA. Sample of 0.5 g potato in 50 ml extracting solution was used containing an expected 14 pg ml-1 of vitamin C Table 1 Vitamin C in sampled foodstuffs by five methods. Results are based on averages of at least 15 assays of each sample.Each analysis was carried out in duplicate. The standard deviation for each method is shown in parentheses. OPD = o-phenylenediamine method; 2,6-DCIP = 2.6-dichloroindophenol method; HPLC-UV = HPLC method with UV detection; HPLC-EC = HPLC method with electrochemical detection; and EPB-EC = enzyme packed bed method with electrochemical detection Vitamin C content/mg per 100 g Sample OPD 2,6-DCIP HPLC-UV HPLC-EC EPB-EC Orange juice 38.0( 3.1) 34.6( 3 -3) 35.4(2.7) 37.3(2.6) 36.5( 1.4) Grapefruit juice 33.1 (1.9) 29.6( 2.2) 32.1(3.1) 32.9( 2.8) 34.8( 0.8) Mash potato powder 152.4(7.5) 148.7( 9.5) 148.1(6.2) 160.2( 7.6) 142.7( 7.9) Brussel sprouts 558.6( 17.1) 552.1(18.2) 596.5(20.5) 547.7(17.2) 561.4(19.0)572 ANALYST, JUNE 1991, VOL. 116 0.20 0.18 %0.16 % 0.14 a 0 0.12 0.10 0.08 I I I I I I I I I 0 20 40 60 80 100 120 140 160 180 Ti me/s Fig.5 Electrochemical detector outputs for a sample of instant mash potato (0.5 g in 50 ml of extracting solution) through 1, inactive bed and 2, active bed months when stored at 4 "C between analyses and for at least 200 assays when in use. The efficiency of the enzyme bed decreased only slightly (99.7%) when the procedure described above was carried out using the stereoisomer of L-AA, D-isoascorbic acid (erythorbic acid). This meant that although it is physiologically inactive , D-isoascorbic acid was measured as vitamin C using this method. This should not, however, be seen as a major problem, as the standard methods currently used are also unable to distinguish between the two isomers.Previous work carried out on the identification of possible contaminants of AO17 showed that there are none commonly found in foodstuffs that are likely to affect seriously the performance of the enzyme. Conclusion This method of analysis appeared to work well for more than 30 different foods and beverages that have been considered briefly (unpublished work), as well as the major four reported here. The method provides a rapid sample analysis of <3 rnin (1.5 min per injection compared with 5-10 min for many of the HPLC methods that have been published) and when used in conjunction with the automated instrumentation it provides high sample throughput (15 samples h-1). Sample and mobile phase cleanliness and quality, vital for HPLC methods, were less important for this method.Using the method described, the wall-jet electrode was operated for at least 500 analyses without the need for cleaning. The use of DITT without HPLC separation has proved to be capable, on occasions, of producing spurious results due to a decrease in the concentration of DTT of up to 8% when passed through the active enzyme bed. This appears to be caused by either nonspecific absorption of DTT by the packed bed, or the active breakdown of DTT by AO. As might be anticipated, initial results indicate that the former is more likely. This problem has been addressed by saturating the packed bed with DTT (0.1 mol dm-3 DTT was passed through the bed for 30 min followed by distilled water for 60 min). This work along with further optimization of the enzyme-electrode buffering system is continuing in order to determine conditions under which the enzyme will operate at high efficiency while allowing optimum selectivity of the electrochemical detector towards AA.280 1 ++ 40 0, n 0 10 20 30 40 50 60 Peak area found by UV detection (arbitrary units) Fig. 6 Comparison of EC and UV detectors with + , inactive and x , active A 0 beds placed between them. Efficiency of bed (99.9%) calculated as the ratio of the two gradients The authors thank K. Thurlow and colleagues from the Nutrition and Microbiology Division of The Laboratory of the Government Chemist (LGC) for their help with vitamin analyses, and the Imperial College Chemistry Microcomputer Unit for their technical assistance with the computer hardware and software.This work was undertaken as part of the Validation of Analytical Measurement (VAM) programme of the LGC funded by the Department of Trade and Industry. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 References Vitamin C (Ascorbic Acid), eds., Counsell, J. N., Hornig, D. H., Applied Science Publishers, London and New Jersey, 1981. Roe, J. H., Mills, M. B., Oesterling, M. J., andDamson, C. M., J. Biol. Chem., 1948, 174, 201. Official Methods of Analysis of the Association of Official Analytical Chemists, ed. Honvitz, W., Association of Official Analytical Chemists, Arlington, VA, 5th edn., 1990. Tillmans, J., Hirsh, P., and Siebert, F. Z., 2. Lebensm. Unters. Forsch., 1932, 63, 21. Deutsch, M. J., and Weeks, C. E., J. Assoc. Off Anal. Chem.. 1965,48, 1248. Egburg, D. C., Potter, R. H., and Heroff, J. C., J . Assoc. Off Anal. Chem., 1977, 60, 126. Polesello, A., and Rizzolo, A., J. Micronutrient Anal., 1986,2, 153. Rouseff, R., Liquid Chromatography of Food & Beverage, Academic Press, New York, 1979, vol. 1, pp. 161-177. Wills, R. B. H., Wimalasiri, P., and Greenfield, H., J. Agric. Food Chem., 1984,32, 836. Pachla, L. A., and Kissinger, P. T., Anal. Chem., 1976,48.364. Wilson, C. W. 111, and Shaw, P. E., J. Agric. Food Chem., 1987, 35, 329. Bradberry, C. W., and Adams, R. N., Anal. Chem., 1983, 55, 2439. Okamura, M., Clin. Chim. Acta, 1980, 103,259. Doner, L. W., and Hicks, K. B., Anal. Biochem., 1981, 115, 225. Margolis, S. A. and Black, I., J. Assoc. Off Anal. Chem., 1987, 70, 806. Maeda, E. E., and Mussa, D. M. D. N., Food Chemistry, 1986, 22, 51. Greenway, G. M., and Ongomo, P., Analyst, 1990, 115, 1297. Paper Of05781 J Received December 27th, 1990 Accepted February 18th, 1991