L-Malic acid biosensor for field-based evaluation of apple, potato and tomato horticultural produce

Abstract

A screen-printed three-electrode amperometric biosensor incorporating malic enzyme for the measurement of L-malic acid in apple, potato and tomato horticultural samples has been developed. The working electrode contained 0.38 mU of immobilised enzyme and was fabricated using rhodinised carbon to facilitate NADPH oxidation at an operating potential of +300 mV vs. Ag/AgCl compared with >+600 mV for bare carbon. The linear range of the sensor was 0.028–0.7 mM L-malic acid with relative standard deviations of 3.3–13.3%. When testing with real apple, potato and tomato samples, the sensor accuracy was within 13.7% of a standard commercially available photometric test kit. The sensor approach is cheap, simple to perform and rapid (6 min), requiring only buffer–electrolyte and a small sample volume.

---

Introduction

The selection of a particular foodstuff by a consumer is largely based on sensory perceptions with taste, which is influenced by diverse factors including saltiness, sweetness, bitterness and acidity as perhaps the most important factors.¹ Texture is also a key parameter and is dictated by many factors including moisture content and fat, carbohydrate and protein levels.² Other important sensory factors include the aroma, shape and colour of the foodstuff.¹

L-Malic acid and citric acid are major organic acids in most fruits and vegetables. Organic acids contribute greatly to taste, particularly of fruit, with a balance of sugar and acid giving rise to the desirable taste of specific produce.³ Significant increases in L-malic acid concentration have been shown to serve as a primary indicator of fruit maturity.^4,5 Hence measurement of L-malic acid provides a more objective means of determining the ripeness and hence ‘shelf life’ of horticultural produce than simple appearance and taste.

Current methods for the determination of L-malic acid in horticultural produce include both liquid and gas chromatography, with the latter method being less widely employed owing to difficulties in isolating and derivatising fruit acids.^6,7Alternative methods include capillary isotachophoresis⁸ and sequential injection Fourier transform infrared (FTIR) spectrometry.⁹ Commercially available test kits for L-malic acid also exist, based on enzymatic assay with photometric detection.¹⁰

Biosensors, incorporating L-malic acid-specific enzymes coupled to electrochemical transducers have also been developed. Benefits include simplicity, rapidity, economy, portability and minimal sample size. Two enzyme types have been used, malic enzyme (ME) [malate dehydrogenase (decarboxylating) (NADP)] (EC 1.1.1.40) and malate dehydrogenase (MDH) (EC 1.1.1.37), the latter catalysing the reaction:

Generally, the reduced NADH co-factor is measured amperometrically and related stoichiometrically to L-malic acid levels. Owing to the low equilibrium constant (k_eq = 6.4 × 10⁻¹³ M), hydrazine, which reacts with the carbonyl group in oxaloacetate, has been used to promote the forward reaction.¹¹ A similar approach using glutamate–oxaloacetate transaminase has also been reported.¹² An alternative approach uses an NADH oxidase (diaphorase) to regenerate NAD⁺ with O₂ consumption measured with a Clark oxygen electrode.¹³ Other strategies include the use of oxaloacetate decarboxylase and pyruvate oxidase to convert oxaloacetate into acetyl phosphate via pyruvate with O₂ consumption again measured with an oxygen electrode.¹⁴ Our method uses malic enzyme (ME), which has a k_eq of 5.1 × 10⁻² M, thus negating the requirement for additional reagents to promote the forward reaction:

Gajovic et al.¹⁵ reported the coupling of ME to salicylate hydroxylase, resulting in the regeneration of NADP⁺ and L-malic acid determination via electrochemical measurement of oxygen consumption. Owing to NADP⁺ recycling, less then 0.025 mM of co-factor was required. Messia et al.¹⁶ used pyruvate oxidase in association with ME with amperometric determination of the hydrogen peroxide by-product at +650 mV vs. an Ag/AgCl reference electrode. Matsumoto et al.¹⁷ recommended ME over MDH owing to a preferred pH optimum of 7.8 compared with 9.5 for MDH.

The presence of electroactive interferents in real samples has led to the widespread coupling of mediators to enzymes in biosensor applications to allow the use of lower detection potentials.¹⁸An alternative approach uses electrocatalysts to selectively decrease the oxidation potential of the target analyte. We have evaluated a large number of commercially available metallised carbons and identified a screen-printable rhodinised carbon with excellent electrocatalytic and enzyme immobilisation properties.^19–22

The aim of this work was to develop stable, cheap, single-use (disposable) screen-printed electrochemical sensors for the rapid and reproducible measurement of L-malic acid in apple, potato and tomato horticultural preparations. Such an approach would provide a simple, rapid, field-based tool for determining the optimum time to harvest produce with respect to the key parameters of ripeness and taste. Simplicity of the measurement process was paramount, owing to the intended field-based usage of the device and target end-user. Consequently, electrocatalytic rhodinised carbon was used in this study to allow the non-mediated detection of NADPH at low potentials, an approach that is of interest for the general exploitation of dehydrogenases in non-mediated biosensor applications. The performance of the device was compared against a standard photometric procedure and tested with real horticultural samples.

Experimental

Reagents

The buffer used was NaH₂PO₄–Na₂HPO₄, pH 7.4 (100 mM unless stated otherwise). Buffer–electrolyte solutions also contained 0.1 M KCl. L-Malic acid, L-tartaric acid, L-citric acid, L-glutamic acid, NADP and NADPH were obtained from Sigma-Aldrich (Poole, Dorset, UK) and prepared in buffer–electrolyte as required. Malic enzyme (malate dehydrogenase, EC 1.1.1.40), from chicken liver (specific activity 19.6 units mg⁻¹) and hydroxyethylcellulose (HEC) were purchased from Fluka (Buchs, Switzerland). Solutions were prepared in de-ionised reverse osmosis water (Elgastat System, Elga, High Wycombe, UK).

Sensor fabrication

Three-electrode devices were mass-manufactured in-house by a multi-stage screen-printing process using a DEK 248 machine (DEK, Weymouth, UK) and screens with appropriate stencil designs (60 per screen) fabricated by DEK Precision Screen Division. The stainless-steel screen mesh was mounted at 45° to the print stroke with 77 wires cm⁻¹ and emulsion thickness of 13 and 18 μm for the solvent and water-resistant screens, respectively.

Devices were printed on to 250 μm thick polyester sheet (Cadillac Plastic, Swindon, UK). The circular electrocatalytic working electrode (planar area 0.16 cm²) was fabricated from MCA 4a (MCA Services, Cambridge, UK), a commercially available carbon powder containing 5% rhodium plus promoters, made into a screen-printable paste by mixing 1∶4 in 2.5% w/v HEC in buffer–electrolyte. The reference electrode ink contained 15% silver chloride in silver paste (MCA Services). The counter electrode and basal tracks were fabricated from 145R carbon ink (MCA Services). The basal tracks were insulated from the measurement solution using 242-SB epoxy-based protective coating ink (Agmet ESL, Reading, UK). The electrodes were then heat treated at 125 °C for 2 h, in order to cure the epoxy resin and to stabilise the electrocatalytic pad to allow prolonged use of the device in aqueous solutions.

Aliquots (10 μl) containing various amounts of ME in 10 mM buffer were pipetted on to the working electrodes, dried for >40 min at room temperature and stored wrapped in silver foil at 4 °C until required.

Electrochemical test procedure

Measurements were performed using an Autolab Bipotentiostat Electrochemical Analyser with GPES3 software (Ecochemie, Utrecht, The Netherlands). A 1.1 cm diameter Whatman 114 filter disc (Whatman, Maidstone, UK) was placed over the three-electrode assembly, which, when wetted with sample, completed the electrochemical circuit. Buffer electrolyte (40 μl) containing 10 mM of NADP, was deposited on the filter-paper and the working electrode was poised at a potential of +300 mV vs. the Ag/AgCl reference. The amperometric measurement procedure was initiated and the electrochemical response was allowed to equilibrate for 240 s, after which 20 μl of sample solution were added to the filter-paper. The change in response was recorded at 350 s (i.e., 110 s after sample addition). All test solutions were prepared in buffer–electrolyte and all tests were performed at 34 °C in accordance with the findings of Messia et al.¹⁶ Since real samples contain appreciable levels of naturally electroactive species such as organic acids, the samples were tested simultaneously on the electrochemical device in the absence of immobilised enzyme, using the Autolab in bipotentiostat mode. The difference in current between the two responses was taken as a measure of specific malic enzyme activity and provided a simple means of accounting for interference factors. All tests were performed in triplicate.

Test kit method

Sensor performance was compared against a standard L-malic acid test method based on a commercially available colourimetric malic enzyme test kit (Dyzyme, Oxford, UK). Samples were tested in accordance with the supplied protocol and the absorbance was measured at 340 nm.

Determination of optimum detection potential for NADPH

The optimum detection potential for the oxidation of NADPH was determined by step-amperometry across the potential range −600 to +600 mV using potential and time increments of 100 mV and 200 s, respectively. Tests were performed on blank electrodes in glass beakers containing 10 ml of stirred buffer–electrolyte. Current responses were recorded immediately before each potential step in the presence or absence of 10 mM NADPH.

Tests on real samples

Samples were prepared using a Tefal multi-purpose grater (Product code 6303, Lakeland, Cumbria, UK) and operated according to the manufacturer’s instructions. In outline, a manually driven screw is used to push diced sample through the grater. Therefore, the pressure applied to the system is primarily a function of the inherent resistance provided by the sample at it is forced through the grater. Juice was extracted from grated samples by wrapping 100 g of material in fine food-grade cotton mesh (plain Voile, product code K656 40921, John Lewis, Milton Keynes, UK). and applying sufficient pressure to recover produce juice. The mesh acted as a crude filter to minimise the presence of large particulates in the juice extracts. Samples were diluted in buffer as follows: Jonagold apple, ×100; Bramley apple, ×200; Russell Burbank potato, ×10; and mini plum tomato, ×10; they were then tested simultaneously using the biosensor and test kit method. The extraction method chosen was designed to be simple, rapid and easy to apply in the field. Three different potato and tomato samples and six different apple samples were analysed in triplicate.

Results and discussion

Selection of optimum measurement conditions

A pH of 7.4 and a temperature of 34 °C were used in all tests based on previous biosensor studies using ME.^15,16 The active site of malic enzyme is thought to be a sulfhydryl group, which can be stabilised using 20 μM 2-mercaptoethanol (2-MCE).¹⁷ ME also requires trace amounts of divalent cations such as Mg²⁺ or Mn²⁺. The ME stock solution used in this study contained 0.5 mM 2-MCE and 10 mM MnCl₂, which proved adequate for sensor performance even after buffer dilution (addition of 20 μM 2-MCE and 3 mM MnCl₂ to the buffer–electrolyte had no measurable effect upon sensor performance).

Choice of working electrode material and detection potential

The electrochemical behaviour of NADPH and a range of potential interferents with detection potential using both bare carbon and rhodinised carbon is shown in Fig. 1. In all cases, increases in potential resulted in increased current responses. The highest responses were observed for L-ascorbic acid on both electrode types. However, rhodinised carbon proved superior for the oxidation of NADPH at applied potentials of +200 mV or greater compared with bare carbon. Using zero current as the baseline, the response ratios for L-ascorbic acid:NADPH:glutamic acid (the most responsive of the other organic acids tested) at potentials of +200 and +300 mV were 10.6∶6.6∶1.0 and 4.3∶4.3∶1.0, respectively on rhodinised carbon. On bare carbon, corresponding response ratios of 85.2∶2.4∶1.0 and 76.4∶3.3∶1.0 were recorded at the same respective potentials.

Fig. 1 Current versus potential profile for NADPH co-factor and a range of possible organic acid interferents measured on (a) bare carbon and (b) MCA4a rhodinised carbon. 10 mM preparations were used with the mean of triplicate tests shown.

Although the rhodinised carbon gave higher response values, the signal-to-noise ratios (S/N) obtained were similar to those obtained for the bare carbon electrodes. For both carbon types, the highest S/N values recorded (∼120∶1) for NADPH were at +300 mV. A maximum S/N value was apparent at the same potential for ascorbic acid, although the other acids showed an increase in S/N at the higher potentials. The data suggest that there is no benefit in measuring at higher potentials since no improvement in the NADPH response relative to the selected interferents is apparent. Higher detection potentials increase the likelihood of the oxidation of other interferents. Rhodinised carbon, with a detection potential of +300 mV, was used in subsequent studies.

Ascorbic acid oxidase could be used to convert the highly electroactive ascorbic acid into dihydroxoascorbic acid and water but would represent an additional preparation step with the removal of only one interferent species. In this study, the approach used was to subtract the background response of each sample using electrodes without enzyme from the equivalent enzyme electrode response. A final device would therefore incorporate two working electrodes—both rhodinised carbon, but only one dosed with enzyme—and a common reference and counter electrode. The difference between the enzyme electrode and ‘compensator’ electrode responses would represent the L-malic acid-specific response.

Optimisation of enzyme loading activity

Initial tests were focused on determining the minimum enzyme activity required to generate a maximum current response from the system. In order to ensure that substrate concentration was not a limiting factor in the biosensor response, tests were performed under saturating levels of L-malic acid. A 2 mM L-malic acid solution was used, based on the criterion of saturation of 5 × K_m, where K_m = 3.9 × 10⁻⁴ M for malic enzyme from liver using L-malic acid as substrate.²³

The amperometric response of sensors containing 0–0.6 mU malic enzyme per electrode to 2 mM L-malic acid in buffer–electrolyte were determined, with a maximum sensor response recorded at enzymes loading >0.35 mU per electrode. Accordingly, enzyme loadings of 0.38 mU per electrode were subsequently used to ensure an enzyme excess to maintain maximum biosensor response using a minimum amount of enzyme.

Sensor analytical performance

A linear relationship between L-malic acid concentration (x) and current response (y) was observed up to 0.7 mM [y = 9.24 × 10⁻⁷x + 5.77 × 10⁻⁸; correlation coefficient (r²) value, 99.67%; analysis of variance (ANOVA) F significance value, 1.17 × 10⁻⁸]. The limit of detection (LOD) for L-malic acid, calculated as 2.5× the standard deviation of the zero analyte response, was 0.028 mM. Hence the linear dynamic range of the system was 0.028–0.7 mM. The relative standard deviation (RSD) values varied from 3.3 to 13.3% (six concentrations, n = 3). Across the range 0–1 mM L-malic acid, the data were best described by a polynomial relationship, y = 6 × 10⁻⁷x⁴ − 2 × 10⁻⁶x³ + 2 × 10⁻⁶x² + 6 × 10⁻⁷x + 7 × 10⁻⁸; r² = 0.9990.

The sensor was also compared against a commercially available standard colorimetric malic acid test kit (Fig. 2). A simple linear relationship (y = 3.1332x; r² = 99.23%; F significance value, 2.25 × 10⁻⁵) was observed. The concentration range over which both methods can be directly compared was 0.028–0.7 mM, dictated by the linear range of the electrochemical method.

Fig. 2 Correlation between electrochemically and photometrically determined L-malic acid concentrations. Background response values have been subtracted. Error bars = s, n = 3.

Inhibition effects

In addition to the electrochemical interferent effect [Fig. 1(b)], a number of compounds found in horticultural samples are also known to inhibit ME activity by competing with L-malic acid for enzyme binding sites. Since notable inhibitors include L-aspartic acid and L-citric acid, the same four organic acids as examined in the interference studies were evaluated with regard to their inhibitory effect on ME enzyme.

First, it was necessary to determine the extent of the sensor signal due to interference effects. The current response due to the direct electro-oxidation of 0.3 mM of the selected organic acids was determined on enzyme-free rhodinised carbon electrodes. The results are given in Table 1, column 2. As indicated previously [Fig. 1(b)], L-ascorbic acid gave the highest response (197 nA) with all other acids giving responses <75 nA.

Next, the response of the enzyme electrode to each organic acid interferent was determined. No significant changes in sensor response were noted, indicating no reaction between the enzyme and these compounds (Table 1, column 3). Inhibitory effects were then determined by preparing solutions containing 0.3 mM L-malic acid and 0.3 mM organic acid interferent and determining the biosensor response (column 4). In all cases, an increase in sensor response was obtained owing to the presence of L-malic acid substrate. The difference between the enzyme electrode and enzyme-free electrode response equated to the sensor response due to L-malic acid oxidation (column 5). Comparing the size of these responses to the L-malic acid response indicated no significant change in the enzyme electrode performance (column 6). Therfore, it was concluded that inhibition of ME by other organic acids was not significant using the enzyme electrode.

Table 1 L-Malic acid biosensor inhibition studies. Full details of the method used to calculate malic enzyme inhibition by the selected organic acids are provided in the text. Values are shown ±1s (n = 3)

Interferent/inhibitor	Rhodinised carbon electrode response/nA^a	Enzyme electrode response/nA^a	Enzyme electrode response to malic acid + interferent/nA	Malic acid-specific response/nA^c	Percentage of malic acid- specific response^d
a 0.3 mM of analyte. b 0.3 mM of analyte + 0.3 mM of L-malic acid. c Difference between columns 2 and 4. d Calculated as (column 5 response value/column 5 response value for L-malic acid) × 100%.
L-Malic acid	68 ± 1.9	337 ± 35.1	350 ± 61.8^a	282	100.0
L-Tartaric acid	45 ± 4.2	62 ± 19.1	334 ± 39.4^b	289	102.6
L-Glutamic acid	75 ± 15.2	53 ± 14.3	355 ± 71.2^b	280	99.2
L-Citric acid	73 ± 13.0	68 ± 36.2	357 ± 42.8^b	284	100.5
L-Ascorbic acid	197 ± 57.6	206 ± 6.6	474 ± 119^b	277	98.2

---

Testing real samples

A study by Hulme and Rhodes²³ indicated that L-malic acid constitutes 90% of the total organic acid content in apples. Akermann et al.²⁴ showed that an elevated amount of malic acid is present in immature fruits, including apples, which decreases as the fruit ripens. Measurement of L-malic acid therefore has possibilities as an objective indicator of fruit maturity. Correspondingly, the performance of the sensor in measuring L-malic acid levels in apple, potato and tomato preparations was determined and the results were compared against the standard photometric test kit. The selected sample preparation method was chosen for ease of performance in the field. All samples were prepared freshly and tested simultaneously using the two methods. The results are given in Table 2. Russell Burbank potatoes, mini plum tomatoes and Jonagold and Bramley apple varieties were found to have L-malic acid concentrations of 3–4, 2.0–2.7, 30–60 and 80–120 mM, respectively.

Table 2 Comparison of sensor and standard photometric test kit results for measurement of L-malic acid in potato and apple samples. Values are shown ±1s (n = 3)

Sample	Dilution	Sensor response, SR/mM	Test kit response, KR/mM	Accuracy (%)^a
a Accuracy = [(SR − KR)/KR] × 100%, where SR is the sensor response and KR is the test kit response.
Potato (Russell Burbank)	×10	0.353 ± 0.055	0.337 ± 0.019	4.8
Potato (Russell Burbank)	×10	0.380 ± 0.010	0.335 ± 0.096	13.4
Potato (Russell Burbank)	×10	0.309 ± 0.008	0.334 ± 0.09	−7.5
Apple (Jonagold)	×100	0.552 ± 0.088	0.488 ± 0.086	13.1
Apple (Jonagold)	×100	0.479 ± 0.059	0.542 ± 0.019	6.0
Apple (Jonagold)	×100	0.378 ± 0.084	0.388 ± 0.081	−2.6
Apple (Bramley)	×200	0.391 ± 0.008	0.419 ± 0.036	−6.7
Apple (Bramley)	×200	0.465 ± 0.095	0.429 ± 0.012	8.4
Apple (Bramley)	×200	0.582 ± 0.208	0.579 ± 0.124	0.5
Tomato (mini plum)	×10	0.234 ± 0.124	0.269 ± 0.020	−13.0
Tomato (mini plum)	×10	0.201 ± 0.035	0.233 ± 0.010	−13.7
Tomato (mini plum)	×10	0.270 ± 0.047	0.258 ± 0.010	4.7

---

The accuracy of the enzyme electrode response was determined against the test kit method using the equation: accuracy = [(S_R − K_R)/K_R] × 100% where S_R is sensor response and K_R is test kit response. In all cases, the sensor response was within 13.7% of the response of the standard method. Although improved accuracy would be desirable, the sensor method does provide a simple and field-based method for indicating the levels of L-malic acid in the tested horticultural samples, thus providing a more objective assessment of produce maturity and taste. The sensor method offers some advantages over the test kit method, particularly with regard to field-based measurements since the sensor merely requires the dilution and addition of prepared sample to the filter disc, whereas the test kit method requires a number of sample and reagent handling steps. Less training is required to operate the sensor method, which is more rapid (6 vs. 24 min) with lower labour, disposables and reagent costs. However, the linear range of the sensor is 0.028–0.7 mM compared with 0.0038–2.5 mM for the test kit.

In addition to horticultural produce assessment, there may also be other potential applications for the L-malic acid biosensor. According to Palleschi,²⁵ the quality of red and white wines and their organoleptic characteristics are very much dependent upon the extent of internal malo-lactic fermentation. The net result of this type of fermentation is the formation of lactic and malic acids, which will influence the taste of the wine product. Palleschi further stated that the biosensor approach matches the requirements of wine producers with regard to selective and rapid determination of lactate and malic acid in wine.

The principle organic acid in apple juice is L-malic acid (0.15–0.91% w/w) and no D-malic acid should be present. Since L-malic acid is expensive, it is not economic for use as a product adulterant. However, synthetic D,L-malic acid is cheap to produce. In unadulterated apple juice, the L-malic acid to total malic acid ratio is 1.0, compared with 0.5–1.0 for adulterated juice. The current way of assessing apple juice adulteration is to measure total malic acid by HPLC and L-malic acid using enzyme-based test kits.²⁶ The L-malic acid biosensor could have applications in place of the test kit where a limited number of simple, rapid measurements are preferred.

Conclusions

Screen-printed electrodes based on malic enzyme with amperometric measurement of NADPH oxidation have proved suitable for the simple, low-cost and rapid determination of L-malic acid in apple, potato and tomato samples. Addition of mediator is not required since the rhodinised carbon working electrode favoured the oxidation of NADPH at a lower operating potential (+300 mV vs. Ag/AgCl) compared with a number of possible organic acid interferents, except L-ascorbic acid. Residual interference effects can be accounted for through the use of a compensator electrode. The sensor performance was tested against a photometric kit using real samples, yielding accuracy values within 13.7% of the standard method. Since the measurement process is simple, it is amenable to field-based usage with a minimal training requirement. Low sensor manufacturing costs result in single-use disposable devices, thus negating problems of progressive electrode fouling.

Acknowledgement

This work was financed by the United Kingdom Ministry of Agriculture, Fisheries and Food (now DEFRA, Department of the Environment, Food and Rural Affairs).

References

1. P. J. Fellows, in Food Processing Technology—Principles and Practice, Woodhead Publishing, Cambridge, 1996, pp. 4–10.
2. A. S. Zczesniak, J. Food Sci., 1963, 28, 385.
3. Postharvest, an Introduction to the Physiology and Handling of Fruit, Vegetables and Ornamentals, ed. R. Wills, B. McGlasson, D. Graham and D. Joyce, University of New South Wales Press, Sydney, 1998, pp. 25–26.
4. J. N. Davis, J. Sci. Food Agric., 1966, 17, 396.
5. T. Wang, A. R. Gonzalez, E. E. Gbur and J. M. Aselage, J. Food Sci., 1993, 58, 631.
6. G. K. Jayaprakasha and K. K. Sakariah, J. Chromatogr., 1998, 806, 337.
7. T. J. Barden, M. Y. Croft, E. J. Murby and R. J. Wells, J. Chromatogr., 1997, 785, 251.
8. P. Blanty, E. Kvasnicka and J. Kenndler, J. Chromatogr., 1996, 737, 255.
9. R . Schindler, R. Vonanch, B. Lendl and R. Kellner, Fresenius’ J. Anal. Chem., 1998, 362, 130.
10. H. Mollering, in Methods of Enzymatic Analysis, ed. H. U. Bergmeyer, VCH, Weinheim, 1989, vol. V11, pp. 39–47.
11. A. M. Almuiabad and A. Townshend, Anal. Chim. Acta, 1989, 221, 337.
12. G. C. Chemnitius and R. D. Schmid, Anal. Lett., 1989, 22, 2987.
13. A. Maines, M. I. Prodromidis, S. M. Tzouwara-Karayanni, M. I. Karayannis, D. Ashworth and P. Vadgama, Anal. Chim. Acta, 2000, 408, 217.
14. N. Gajovic, A. Warsinke and F. W. Scheller, J. Chem. Tech. Biotechnol., 1997, 68, 31.
15. N. Gajovic, A. Warsinke and F. W. Scheller, J. Biotechnol., 1998, 61, 129.
16. M. C. Messia, D. Compagnone and M. Esti, Anal. Chem., 1996, 68, 360.
17. K. Matsumoto, S. Higuchi and T. Tsukatani, Biosci. Biotechnol. Biochem., 1996, 60, 847.
18. I. E. Tothill, Comput. Electron. Agric., 2001, 30, 205.
19. J. D. Newman, S. F. White, I. E. Tothill and A. P. F. Turner, Anal. Chem., 1995, 67, 4594.
20. I. E. Tothill, J. D. Newman, S. F. White and A. P. F. Turner, Enzyme Microb. Technol., 1997, 20, 590.
21. S. Kröger, S. J. Setford and A. P. F. Turner, Anal. Chim. Acta, 1998, 368, 219.
22. S. Kröger, S. J. Setford and A. P. F. Turner, Biotechn. Tech., 1998, 12, 123.
23. A. C. Hulme and M. J. C. Rhodes, in The Biochemistry of Fruits and Their Products, ed. A. C. Hulme, Academic Press, London, 1971, vol. 2, pp. 110–115.
24. J. Ackermann, M. Fishcher and R. Arnado Bolt, J. Agric. Food Chem., 1992, 40, 1131.
25. G. . Palleschi, in Biosensors for Food Analysis, ed. A. O. Scott, Royal Society of Chemistry, Cambridge, 1998, pp. 124–134.
26. E. R. Elkins and J. R. Heuser, J. AOAC Int., 1994, 77, 411.

---