Microbial Detection by a Glucose Biosensor Coupled to a Microdialysis Fibre F. Palmisano*a, A. De Santisa, G. Tantillob, T. Volpicellaab and P. G. Zambonina a Dipartimento di Chimica, Universit`a degli Studi, Via Orabona, 4-70126 Bari, Italy b Istituto di Ispezione degli Alimenti, Universit`a degli Studi, Via per Casamassima Km, 3-70010 Valenzano, Bari, Italy The use of a glucose biosensor coupled to microdialysis sampling in a flow injection analysis system is described to follow the growth of Escherichia coli in a glucose-containing liquid culture medium.The experimental set-up permitted a throughput rate of 25 samples h21. Growth curves were modelled by a modified Gompertz equation, which permitted the determination of lag time and maximum specific growth rate. The time required to produce an appreciable variation in the biosensor response (minimum detection time, MDT) was determined. A plot of MDT versus microbial concentration was found to be linear in the range 106–1010 colony forming units (cfu) ml21.A microbial concentration of 106 cfu ml-1 can be detected after about 5 h. Keywords: Biosensor; microbial detection; Escherichia coli; glucose; microdialysis The quantification of micro-organisms plays a vital role in fermentation processes, food industry, medical practice and industrial waste water monitoring. Thus, an accurate method for the rapid (possibly real-time) determination of biomass is an important goal to be achieved.A large number of detection methods have been developed utilising the optical, electrochemical, biochemical and physical properties of microorganisms (see ref. 1 for an excellent review). An ideal microbial sensor should fulfil several requirements; it must be accurate, sensitive, easy to calibrate and robust. In addition, there should be no need for sample pre-treatment, no interference from the culture conditions, no added reagents and on-line capabilities. The analysis time is, obviously, a characteristic of paramount importance, particularly when microbial detection/quantification is required on a production line, e.g., food processing/ packaging.The term ‘rapid method’ is usually applied to any method presenting an analysis time significantly shorter (e.g., less than 24 h) than that of conventional detection procedures. Among these, viable cell counting methods2–4 (e.g., plate count) are widely used to estimate microbial populations; the main disadvantage of the plate count method is the long incubation period (24–72 h) and the high degree of operator skill required.The microbial content of a sample can be determined by monitoring the microbial metabolism instead of the biomass. Several electrochemical detection systems have been proposed: impedimetry,5 conductivity,6 potentiometry,7 voltammetry8 and amperometry.9The use of biosensors for biomass detection has been scarcely investigated. A glucose biosensor, based on the amperometric mediated enzyme electrode principle, has been adapted for the development of a biosensor ‘knife probe’ and applied to the ultra-rapid in situ assessment of meat freshness.10 Depletion of glucose at the surface relative to the bulk of the meat is indicative of microbial activity.In this paper, a different approach is explored. A glucose biosensor,11,12 based on glucose oxidase (GOD) immobilised in an electrochemically synthesised poly(pyrrole) (PPY) film, is used to monitor in near-real-time the glucose consumption in a liquid culture medium which has been inoculated with a given microbial mass.A microdialysis fibre12,13 is used for sampling purposes; hence, there is no need for biosensor sterilisation (which can denature the enzyme). At the same time on-line dilution of the sample is provided in order to obtain glucose signals in the linear range of response. The calibration status of the sensor can be conveniently checked by injecting, at regular time intervals, a glucose standard.Microbial growth causes glucose depletion in the culture medium which can be followed by the biosensor the response of which changes accordingly. The minimum detection time (MDT), i.e., the time required to detect a significant change in the sensor response, can be related (see below) to the initial microbial concentration through a suitable calibration plot. Hence, a microbial concentration of 106 colony forming units (cfu) ml21 can be detected after about 5 h.Experimental Chemicals Escherichia coli (isolated in our laboratory) was grown in two different liquid culture media: the first contained peptone, NaCl and glucose (2.0 g l21) while the second (Koser-modified medium) contained Na2NH4PO4, MgSO4·7H2O, bovine serum and glucose (2.0 g l21). Plate counts were performed using ‘Agar nutritive’ and ‘Agar nutritive with brilliant green’ (Oxoid, Basingstoke, Hampshire, UK).Glucose oxidase (EC 1.1.3.4 from Aspergillus niger, Type VII S) was obtained from Sigma (St. Louis, MO, USA). A glucose stock solution was prepared from b-D-(+)-glucose (Sigma) and allowed to mutarotate overnight; glucose standard solutions were prepared just before use by dilution of the stock with phosphate buffer. Pyrrole (Aldrich, Milwaukee, WI, USA) was purified by vacuum distillation at 62 °C. All other chemicals were of analytical-reagent grade. Biosensor Preparation Glucose enzyme electrodes (Pt–GOD–PPY) were prepared, as previously described,11 by electrochemical polymerisation at +0.7 V versus Ag/AgCl from a 10 mmol l21 KCl supporting electrolyte containing 0.4 mmol l21 pyrrole and 250 U ml21 of GOD.Pt–GOD–PPY electrodes were overoxidised at +0.7 V overnight in phosphate buffer. The detection potential was +0.7 V versus Ag/AgCl. Apparatus A PAR 273 (EG&G Princeton Applied Research, Princeton, NJ, USA) potentiostat–galvanostat was used for the elec- Analyst, October 1997, Vol. 122 (1125–1128) 1125trosynthesis of the PPY film containing immobilised GOD. A PAR Model 400 electrochemical detector coupled to a Kipp & Zonen (Delft, The Netherlands) BD112 Y–t recorder was used to monitor the response of the glucose biosensor. A Gilson (Villiers le Bel, France) Minipuls 3 peristaltic pump was used in flow experiments. Spectra/por hollow fibres (regenerated cellulose, 150 mm id, 9 mm wall thickness) having a molecular weight cut-off of 9000 Da were obtained from Spectrum Medical Industries (Los Angeles, CA, USA).A Heto (Allerød, Denmark) Type 21 DT-2 thermostatically controlled bath was used for temperature control. The set-up used to follow the microbial metabolism is shown in Fig. 1. A two-channel peristaltic pump (4 in Fig. 1) was used to pump the carrier solution (1) through the microdialysis fibrebased sampler (6), previously described in ref. 13. Unless otherwise stated, the flow rate outside/inside the microdialysis fibre was 300 ml min21.A six-way low pressure injection valve (5) equipped with a 110 ml injection loop (5A) permitted discontinuous sampling of the culture medium (3) or of glucose standards (8) required for the initial calibration of the sensor and checking of the calibration status during the experiment. The maximum throughput allowed by the above-described apparatus was 25 samples h21. Results and Discussion Microdialysis is a dynamic sampling method based on analyte diffusion across a semi-permeable membrane in the presence of a concentration gradient.The concentration ratio at the two sides of the microdialysis membrane is dependent on a number of factors, the most important being (for a given probe) the perfusion rate, temperature, analyte species and physicochemical characteristics of the external medium. The microdialysis probe can, of course, be inserted directly, once sterilised, into the growth medium and glucose in the perfusate monitored continuously.A requirement that must be fulfilled is that the fibre recovery should remain essentially constant during the experiment. This requirement might not be met since fouling of the fibre surface by the growing bacteria can occur, particularly over long periods of time. For this reason a different approach was followed based on the sampler previously described and fully characterised in ref. 13. In this approach, the fibre is continuously washed by the carrier buffer and intermittently contacted by the growth medium only during injections.Control of the calibration status of the sensor is more easily achieved owing to the relatively high sample throughput of the system depicted in Fig. 1. Fig. 2 shows a typical example of the sensor responses for different glucose standards and the relevant calibration plot. As can be seen, the system response is repeatable, linear and sufficiently fast to permit a throughput of about 25 samples h21.The sampling frequency can of course be increased, by increasing the flow rate inside/outside the fibre,13 but at the expense of fibre recovery (i.e., of sensitivity), which decreases on increasing the flow rate. When a glucose-enriched sterile liquid culture medium is inoculated with a given microbial mass, the microbial growth will cause a depletion of the glucose in the culture medium which can be followed in near-real-time by a glucose biosensor the response of which S(t) will change with time accordingly.If S(0) is the sensor response before inoculation (i.e., at t = 0) then the variable y y S t S = - 1 0 ( ) ( ) (1) which is related to the microbial growth, could be described in terms of existing models such as the Gompertz equation modified by Zwietering et al.14 y A A t = × - × - + é ë ê ù û ú ìí ï îï üý ï �ï exp ( ) exp e m m l 1 (2) where t is the time, mm is the maximum specific growth rate, l is the lag time and A is the asymptotic value reached by y for t approaching infinity.Fig. 3 shows some typical sensor outputs at different times in an experiment where the growth medium is inoculated with a known microbial mass (107 cfu ml21). The peaks marked Std1 and Std2 refer to a 10 mmol l21 glucose standard injected before and 30 h after inoculation, respectively. As can be seen, the peak height remained virtually unchanged, demonstrating that the biosensor response remained stable over the typical time frame of such an experiment, and that the fibre recovery is also constant (absence of fouling).Fig. 4 shows typical growth data obtained at different microbial concentrations; the solid lines in Fig. 4 represent the best fit of the experimental data obtained using eqn. (2). Maximum specific growth rate, mm, values of Fig. 1 Schematic diagram of the experimental set-up. 1: Carrier solution reservoir; 2: thermostatically controlled bath; 3: culture medium; 4: twochannel peristaltic pump; 5: six-way low pressure injection valve; 5A injection loop; 6: microdialysis fibre sampler; 7: three-way valve; 8: glucose solution (calibration standard); 9: syringe for manual filling of the injection loop; 10: flow cell with glucose amperometric biosensor; 11: potentiostat; and 12: recorder.Fig. 2 Glucose responses obtained at a Pt–PPYox–GOD biosensor with the experimental configuration shown in Fig. 1. Flow rate inside/outside the microdialysis fibre, 300 ml min21.Inset: calibration plot. 1126 Analyst, October 1997, Vol. 1220.093, 0.094 and 0.095 h21 were estimated at microbial concentrations of 109, 108 and 106 cfu ml21, respectively. Linear interpolation of the experimental points after the lag phase gives a straight line the intercept of which on the time axis can be assumed to be the MDT; note that the MDT values so obtained are essentially the same as the lag time, l, values obtained through eqn.(2). Modelling of the growth curves indicates that the MDT values correspond to about a (7 ± 1)% decrease in the biosensor response measured before microbial inoculation in the growth medium (hence, in practical applications the actual growth curve does not have to be followed entirely). A plot of the MDT values versus the logarithm of the initial microbial concentration (see Fig. 5) was found to be linear, giving a ‘working curve’ from which an unknown microbial concentration can be determined from the measured MDT.As can be seen, under the present experimental conditions, the MDT varies between 25 min and about 5 h for microbial concentrations varying between 1011 and 106 cfu ml21; of course, smaller microbial concentrations require longer times. In any case, it is evident that the total analysis time is significantly shorter than that required by plate count methods, and is essentially dictated by the kinetics of the growth process since the biosensor responds in near-real-time. Furthermore, the flow system described here can monitor several different microbial cultures simultaneously, potentially can be automated and does not require a skilled operator.The usefulness of such an approach in practical applications was demonstrated by preliminary experiments on a contaminated meat sample the microbial mass of which was evaluated by the proposed method and the result compared with that of the conventional plate count method.For this purpose, a naturally contaminated meat sample was washed three times with 10 ml aliquots of a sterile physiological solution to remove the surface microbial population. A 20 ml aliquot was then transferred into 100 ml of ‘modified Koser’ growth medium, and glucose depletion versus time followed. The microbial concentration was also determined, in a parallel experiment, by the plate count method. From the experimental MDT value and the working curve, a microbial concentration of (2.0 ± 1.4) 3 109 cfu g21 was calculated, which was found not to be significantly different (according to a t-test at the 95% confidence level) from the value obtained by the plate count method.It should be noted that since the growth medium may not be selective for E. coli, data obtained on the meat sample might have been influenced by the presence of micro-organisms (e.g., mesophilic bacteria) other than coliforms. The above results on real samples must therefore be considered as preliminary; further work in this direction is underway in our laboratory.The authors thank Professor G. Tiecco for helpful suggestions. Financial support from MURST and National Research Council (CNR, Rome) is gratefully acknowledged. References 1 Hobson, N. S., Tothill, I., and Turner A. P. F., Biosens. Bioelectron., 1996, 11, 455. 2 Hope, C. F. A., and Tubb, R. S., J. Inst. Brewing, 1985, 91, 12. 3 Ding, T., and Schmidt, R.D., Anal. Chim. Acta, 1990, 234, 247. 4 Ashley, N., Dairy Ind. Int., 1991, 56, 39. 5 Zafari, Y., and Martin, W. J., J. Clin. Microbiol., 1977, 5, 545. 6 Richards, J. C. S., Jason, A. C., Hobbs, G., Gibson, D. M., and Christie, R. H., J. Phys., 1978, 11, 560. 7 Wilkins, J. R., Young, R., and Boykin, E., J. Appl. Environ. Microbiol., 1978, 35, 214. 8 Matsunaga, T., and Namba, Y., Anal. Chim. Acta, 1984, 159, 87. 9 Turner, A. P. F., Ramsey, G., and Higgins, I. J. H., Biochem.Soc. Trans., 1983, 11, 445. 10 Kress-Rogers, E., D’Costa, E. J., Sollars, J. E., Gibbs, P. A., and Turner, A. P. F., J. Food Control, 1993, 4, 149. Fig. 3 Glucose responses obtained with a Pt–PPYox–GOD biosensor at different times (a = 6.5; b = 9.5; c = 13 h), during an experiment in which the microbial growth was followed with the experimental configuration shown in Fig. 1. The growth medium was inoculated with E. coli at a known concentration of 107 cfu ml21. Std1 and Std2 refer to a 10 mmol l21 glucose standard injected before inoculation and 30 h after.Fig. 4 Microbial growth curves obtained at three different E. coli concentrations: 109 (5), 108 (-) and 106 (~) cfu ml21. Solid lines are calculated as the best fit of eqn. (2). Fig. 5 MDT versus log of microbial concentration for E. coli. MDT values are obtained by extrapolation on the time axis of the linear portion of the growth curve. Analyst, October 1997, Vol. 122 112711 Centonze, D., Guerrieri, A., Malitesta, C., Palmisano, F., and Zambonin, P.G., Fresenius’ J. Anal. Chem., 1992, 342, 729. 12 Palmisano, F., Centonze, D., Guerrieri, A., and Zambonin, P. G., Biosens. Bioelectron., 1993, 8, 393. 13 Palmisano, F., Centonze, D., Quinto, M.,nd Zambonin, P. G., Biosens. Bioelectron., 1996, 11, 419. 14 Zwietering, M. H., Jongenburger, I., Rombouts, F. M., and van’t Riet, K., Appl. Environ. Microbiol., 1990, 56, 1857. Paper 7/03594B Received May 23, 1997 Accepted August 4, 1997 1128 Analyst, October 1997, Vol. 122 Microbial Detection by a Glucose Biosensor Coupled to a Microdialysis Fibre F. Palmisano*a, A. De Santisa, G. Tantillob, T. Volpicellaab and P. G. Zambonina a Dipartimento di Chimica, Universit`a degli Studi, Via Orabona, 4-70126 Bari, Italy b Istituto di Ispezione degli Alimenti, Universit`a degli Studi, Via per Casamassima Km, 3-70010 Valenzano, Bari, Italy The use of a glucose biosensor coupled to microdialysis sampling in a flow injection analysis system is described to follow the growth of Escherichia coli in a glucose-containing liquid culture medium.The experimental set-up permitted a throughput rate of 25 samples h21. Growth curves were modelled by a modified Gompertz equation, which permitted the determination of lag time and maximum specific growth rate. The time required to produce an appreciable variation in the biosensor response (minimum detection time, MDT) was determined. A plot of MDT versus microbial concentration was found to be linear in the range 106–1010 colony forming units (cfu) ml21.A microbial concentration of 106 cfu ml-1 can be detected after about 5 h. Keywords: Biosensor; microbial detection; Escherichia coli; glucose; microdialysis The quantification of micro-organisms plays a vital role in fermentation processes, food industry, medical practice and industrial waste water monitoring. Thus, an accurate method for the rapid (possibly real-time) determination of biomass is an important goal to be achieved.A large number of detection methods have been developed utilising the optical, electrochemical, biochemical and physical properties of microorganisms (see ref. 1 for an excellent review). An ideal microbial sensor should fulfil several requirements; it must be accurate, sensitive, easy to calibrate and robust. In addition, there should be no need for sample pre-treatment, no interference from the culture conditions, no added reagents and on-line capabilities.The analysis time is, obviously, a characteristic of paramount importance, particularly when microbial detection/quantification is required on a production line, e.g., food processing/ packaging. The term ‘rapid method’ is usually applied to any method presenting an analysis time significantly shorter (e.g., less than 24 h) than that of conventional detection procedures. Among these, viable cell counting methods2–4 (e.g., plate count) are widely used to estimate microbial populations; the main disadvantage of the plate count method is the long incubation period (24–72 h) and the high degree of operator skill required.The microbial content of a sample can be determined by monitoring the microbial metabolism instead of the biomass. Several electrochemical detection systems have been proposed: impedimetry,5 conductivity,6 potentiometry,7 voltammetry8 and amperometry.9The use of biosensors for biomass detection has been scarcely investigated.A glucose biosensor, based on the amperometric mediated enzyme electrode principle, has been adapted for the development of a biosensor ‘knife probe’ and applied to the ultra-rapid in situ assessment of meat freshness.10 Depletion of glucose at the surface relative to the bulk of the meat is indicative of microbial activity. In this paper, a different approach is explored. A glucose biosensor,11,12 based on glucose oxidase (GOD) immobilised in an electrochemically synthesised poly(pyrrole) (PPY) film, is used to monitor in near-real-time the glucose consumption in a liquid culture medium which has been inoculated with a given microbial mass.A microdialysis fibre12,13 is used for sampling purposes; hence, there is no need for biosensor sterilisation (which can denature the enzyme). At the same time on-line dilution of the sample is provided in order to obtain glucose signals in the linear range of response.The calibration status of the sensor can be conveniently checked by injecting, at regular time intervals, a glucose standard. Microbial growth causes glucose depletion in the culture medium which can be followed by the biosensor the response of which changes accordingly. The minimum detection time (MDT), i.e., the time required to detect a significant change in the sensor response, can be related (see below) to the initial microbial concentration through a suitable calibration plot.Hence, a microbial concentration of 106 colony forming units (cfu) ml21 can be detected after about 5 h. Experimental Chemicals Escherichia coli (isolated in our laboratory) was grown in two different liquid culture media: the first contained peptone, NaCl and glucose (2.0 g l21) while the second (Koser-modified medium) contained Na2NH4PO4, MgSO4·7H2O, bovine serum and glucose (2.0 g l21). Plate counts were performed using ‘Agar nutritive’ and ‘Agar nutritive with brilliant green’ (Oxoid, Basingstoke, Hampshire, UK).Glucose oxidase (EC 1.1.3.4 from Aspergillus niger, Type VII S) was obtained from Sigma (St. Louis, MO, USA). A glucose stock solution was prepared from b-D-(+)-glucose (Sigma) and allowed to mutarotate overnight; glucose standard solutions were prepared just before use by dilution of the stock with phosphate buffer. Pyrrole (Aldrich, Milwaukee, WI, USA) was purified by vacuum distillation at 62 °C. All other chemicals were of analytical-reagent grade.Biosensor Preparation Glucose enzyme electrodes (Pt–GOD–PPY) were prepared, as previously described,11 by electrochemical polymerisation at +0.7 V versus Ag/AgCl from a 10 mmol l21 KCl supporting electrolyte containing 0.4 mmol l21 pyrrole and 250 U ml21 of GOD. Pt–GOD–PPY electrodes were overoxidised at +0.7 V overnight in phosphate buffer. The detection potential was +0.7 V versus Ag/AgCl. Apparatus A PAR 273 (EG&G Princeton Applied Research, Princeton, NJ, USA) potentiostat–galvanostat was used for the elec- Analyst, October 1997, Vol. 122 (1125–1128) 1125trosynthesis of the PPY film containing immobilised GOD. A PAR Model 400 electrochemical detector coupled to a Kipp & Zonen (Delft, The Netherlands) BD112 Y–t recorder was used to monitor the response of the glucose biosensor. A Gilson (Villiers le Bel, France) Minipuls 3 peristaltic pump was used in flow experiments. Spectra/por hollow fibres (regenerated cellulose, 150 mm id, 9 mm wall thickness) having a molecular weight cut-off of 9000 Da were obtained from Spectrum Medical Industries (Los Angeles, CA, USA).A Heto (Allerød, Denmark) Type 21 DT-2 thermostatically controlled bath was used for temperature control. The set-up used to follow the microbial metabolism is shown in Fig. 1. A two-channel peristaltic pump (4 in Fig. 1) was used to pump the carrier solution (1) through the microdialysis fibrebased sampler (6), previously described in ref. 13. Unless otherwise stated, the flow rate outside/inside the microdialysis fibre was 300 ml min21. A six-way low pressure injection valve (5) equipped with a 110 ml injection loop (5A) permitted discontinuous sampling of the culture medium (3) or of glucose standards (8) required for the initial calibration of the sensor and checking of the calibration status during the experiment. The maximum throughput allowed by the above-described apparatus was 25 samples h21.Results and Discussion Microdialysis is a dynamic sampling method based on analyte diffusion across a semi-permeable membrane in the presence of a concentration gradient. The concentration ratio at the two sides of the microdialysis membrane is dependent on a number of factors, the most important being (for a given probe) the perfusion rate, temperature, analyte species and physicochemical characteristics of the external medium. The microdialysis probe can, of course, be inserted directly, once sterilised, into the growth medium and glucose in the perfusate monitored continuously. A requirement that must be fulfilled is that the fibre recovery should remain essentially constant during the experiment.This requirement might not be met since fouling of the fibre surface by the growing bacteria can occur, particularly over long periods of time. For this reason a different approach was followed based on the sampler previously described and fully characterised in ref. 13. In this approach, the fibre is continuously washed by the carrier buffer and intermittently contacted by the growth medium only during injections. Control of the calibration status of the sensor is more easily achieved owing to the relatively high sample throughput of the system depicted in Fig. 1. Fig. 2 shows a typical example of the sensor responses for different glucose standards and the relevant calibration plot. As can be seen, the system response is repeatable, linear and sufficiently fast to permit a throughput of about 25 samples h21.The sampling frequency can of course be increased, by increasing the flow rate inside/outside the fibre,13 but at the expense of fibre recovery (i.e., of sensitivity), which decreases on increasing the flow rate. When a glucose-enriched sterile liquid culture medium is inoculated with a given microbial mass, the microbial growth will cause a depletion of the glucose in the culture medium which can be followed in near-real-time by a glucose biosensor the response of which S(t) will change with time accordingly.If S(0) is the sensor response before inoculation (i.e., at t = 0) then the variable y y S t S = - 1 0 ( ) ( ) (1) which is related to the microbial growth, could be described in terms of existing models such as the Gompertz equation modified by Zwietering et al.14 y A A t = × - × - + é ë ê ù û ú ìí ï îï üý ï �ï exp ( ) exp e m m l 1 (2) where t is the time, mm is the maximum specific growth rate, l is the lag time and A is the asymptotic value reached by y for t approaching infinity.Fig. 3 shows some typical sensor outputs at different times in an experiment where the growth medium is inoculated with a known microbial mass (107 cfu ml21). The peaks marked Std1 and Std2 refer to a 10 mmol l21 glucose standard injected before and 30 h after inoculation, respectively. As can be seen, the peak height remained virtually unchanged, demonstrating that the biosensor response remained stable over the typical time frame of such an experiment, and that the fibre recovery is also constant (absence of fouling).Fig. 4 shows typical growth data obtained at different microbial concentrations; the solid lines in Fig. 4 represent the best fit of the experimental data obtained using eqn. (2). Maximum specific growth rate, mm, values of Fig. 1 Schematic diagram of the experimental set-up. 1: Carrier solution reservoir; 2: thermostatically controlled bath; 3: culture medium; 4: twochannel peristaltic pump; 5: six-way low pressure injection valve; 5A injection loop; 6: microdialysis fibre sampler; 7: three-way valve; 8: glucose solution (calibration standard); 9: syringe for manual filling of the injection loop; 10: flow cell with glucose amperometric biosensor; 11: potentiostat; and 12: recorder.Fig. 2 Glucose responses obtained at a Pt–PPYox–GOD biosensor with the experimental configuration shown in Fig. 1. Flow rate inside/outside the microdialysis fibre, 300 ml min21. Inset: calibration plot. 1126 Analyst, October 1997, Vol. 1220.093, 0.094 and 0.095 h21 were estimated at microbial concentrations of 109, 108 and 106 cfu ml21, respectively. Linear interpolation of the experimental points after the lag phase gives a straight line the intercept of which on the time axis can be assumed to be the MDT; note that the MDT values so obtained are essentially the same as the lag time, l, values obtained through eqn.(2). Modelling of the growth curves indicates that the MDT values correspond to about a (7 ± 1)% decrease in the biosensor response measured before microbial inoculation in the growth medium (hence, in practical applications the actual growth curve does not have to be followed entirely). A plot of the MDT values versus the logarithm of the initial microbial concentration (see Fig. 5) was found to be linear, giving a ‘working curve’ from which an unknown microbial concentration can be determined from the measured MDT.As can be seen, under the present experimental conditions, the MDT varies between 25 min and about 5 h for microbial concentrations varying between 1011 and 106 cfu ml21; of course, smaller microbial concentrations require longer times. In any case, it is evident that the total analysis time is significantly shorter than that required by plate count methods, and is essentially dictated by the kinetics of the growth process since the biosensor responds in near-real-time.Furthermore, the flow system described here can monitor several different microbial cultures simultaneously, potentially can be automated and does not require a skilled operator. The usefulness of such an approach in practical applications was demonstrated by preliminary experiments on a contaminated meat sample the microbial mass of which was evaluated by the proposed method and the result compared with that of the conventional plate count method.For this purpose, a naturally contaminated meat sample was washed three times with 10 ml aliquots of a sterile physiological solution to remove the surface microbial population. A 20 ml aliquot was then transferred into 100 ml of ‘modified Koser’ growth medium, and glucose depletion versus time followed. The microbial concentration was also determined, in a parallel experiment, by the plate count method.From the experimental MDT value and the working curve, a microbial concentration of (2.0 ± 1.4) 3 109 cfu g21 was calculated, which was found not to be significantly different (according to a t-test at the 95% confidence level) from the value obtained by the plate count method. It should be noted that since the growth medium may not be selective for E. coli, data obtained on the meat sample might have been influenced by the presence of micro-organisms (e.g., mesophilic bacteria) other than coliforms.The above results on real samples must therefore be considered as preliminary; further work in this direction is underway in our laboratory. The authors thank Professor G. Tiecco for helpful suggestions. Financial support from MURST and National Research Council (CNR, Rome) is gratefully acknowledged. References 1 Hobson, N. S., Tothill, I., and Turner A. P. F., Biosens. Bioelectron., 1996, 11, 455. 2 Hope, C. F. A., and Tubb, R. S., J. Inst. Brewing, 1985, 91, 12. 3 Ding, T., and Schmidt, R. D., Anal. Chim. Acta, 1990, 234, 247. 4 Ashley, N., Dairy Ind. Int., 1991, 56, 39. 5 Zafari, Y., and Martin, W. J., J. Clin. Microbiol., 1977, 5, 545. 6 Richards, J. C. S., Jason, A. C., Hobbs, G., Gibson, D. M., and Christie, R. H., J. Phys., 1978, 11, 560. 7 Wilkins, J. R., Young, R., and Boykin, E., J. Appl. Environ. Microbiol., 1978, 35, 214. 8 Matsunaga, T., and Namba, Y., Anal. Chim. Acta, 1984, 159, 87. 9 Turner, A. P. F., Ramsey, G., and Higgins, I. J. H., Biochem. Soc. Trans., 1983, 11, 445. 10 Kress-Rogers, E., D’Costa, E. J., Sollars, J. E., Gibbs, P. A., and Turner, A. P. F., J. Food Control, 1993, 4, 149. Fig. 3 Glucose responses obtained with a Pt–PPYox–GOD biosensor at different times (a = 6.5; b = 9.5; c = 13 h), during an experiment in which the microbial growth was followed with the experimental configuration shown in Fig. 1. The growth medium was inoculated with E. coli at a known concentration of 107 cfu ml21. Std1 and Std2 refer to a 10 mmol l21 glucose standard injected before inoculation and 30 h after. Fig. 4 Microbial growth curves obtained at three different E. coli concentrations: 109 (5), 108 (-) and 106 (~) cfu ml21. Solid lines are calculated as the best fit of eqn. (2). Fig. 5 MDT versus log of microbial concentration for E. coli. MDT values are obtained by extrapolation on the time axis of the linear portion of the growth curve. Analyst, October 1997, l. 122 112711 Centonze, D., Guerrieri, A., Malitesta, C., Palmisano, F., and Zambonin, P. G., Fresenius’ J. Anal. Chem., 1992, 342, 729. 12 Palmisano, F., Centonze, D., Guerrieri, A., and Zambonin, P. G., Biosens. Bioelectron., 1993, 8, 393. 13 Palmisano, F., Centonze, D., Quinto, M., and Zambonin, P. G., Biosens. Bioelectron., 1996, 11, 419. 14 Zwietering, M. H., Jongenburger, I., Rombouts, F. M., and van’t Riet, K., Appl. Environ. Microbiol., 1990, 56, 1857. Paper 7/03594B Received May 23, 1997 Accepted August 4, 1997 1128 Analyst, October 1997, Vol. 122