- Email: [email protected]
Adaptation of a manometric biosensor to measure glucose and lactose
Biosensors and Bioelectronics 18 (2003) 101 /107 www.elsevier.com/locate/bios
Adaptation of a manometric biosensor to measure glucose and lactose Daniel M. Jenkins, Michael J. Delwiche * Department of Biological and Agricultural Engineering, University of California, 3048 Bainer Hall, Davis, CA 95616, USA Received 25 May 2001; received in revised form 17 April 2002; accepted 18 July 2002
Abstract A manometric sensor previously developed to measure urea was modified to measure glucose and lactose through enzymatic oxidation. Change in pressure in an enclosed cavity was correlated to the depletion of oxygen resulting from the enzymatic oxidation of glucose or lactose. The response of the sensor was linear and could be made adjustable over a large range by adjusting the amount of sample loaded into the fixed volume reactor. Because of the slow mutarotation of glucose, the oxidation of glucose was not allowed to proceed to completion. Therefore, the precision of the sensor (approximately 0.2 mM in a range from 0 to 5 mM) was limited by variations in the oxidation rate of glucose by glucose oxidase. Because the assay for lactose measured glucose subsequent to the hydrolysis of lactose by b-galactosidase, the same degree of precision was observed in lactose. Milk lactose, typically at concentrations of about 150 mM, was estimated using the lactose assay after first diluting the samples. For many fluids such as milk, the use of manometric sensors for oxidizable substrates may be preferable to optical and electrochemical methods because they are robust and suffer a low degree of optical and chemical interferences. Glucose and lactose are representative of many important oxidizable substrates, which may be determined in this manner, many of which do not suffer from limitations caused by mutarotation. In theory, detection limits less than 1 mM may be achieved using these methods. # 2003 Elsevier Science B.V. All rights reserved. Keywords: Enzyme sensor; Oxidation reaction; Pressure
1. Introduction A change in dissolved gas concentration can be measured in a closed cavity as a change in partial pressure of the gas (Dixon, 1934). Using this principle, it is possible to measure biological components of a fluid such a urea, which liberates CO2 when hydrolyzed by the enzyme urease. A biosensor using this principle for the on-line analysis of urea in milk (Jenkins and Delwiche, 2002) was robust to the difficult chemical and physical properties of milk. The precision (9/0.15 mM in buffer, 9/0.25 mM in milk, in a range from 0 to 10 mM) was comparable to that of other urea sensors. In many applications, including the measurement of milk urea, instruments suffer a high degree of chemical,
* Corresponding author: Tel.: /1-530-752-7023; fax: /1-530-7522640 E-mail address: [email protected] (M.J. Delwiche).
physical, optical, or electrical interference in the biological media which requires extensive pretreatment or dialysis of the samples (Jenkins, 2001, pp. 16 /32). The manometric sensor for urea adapted for this research was relatively inexpensive to construct, did not require pretreatment, and was shown to function reproducibly in the dirty and uncontrolled environment of a typical dairy, with a reagent cost of less than US $0.03 per sample (Jenkins, 2001). Most amino acids and some b-ketoacids such as acetoacetate may also be enzymatically decarboxylated and measured through the volatilization of CO2. Similarly, reactions depleting dissolved gases such as oxygen may be measured by the change in vacuum in a closed cell. Glucose and lactose are examples of the many important metabolic carbohydrates, which may be measured through enzymatic oxidation. To illustrate the design of a gas depleting type of manometric sensor, we modified the urea sensor to measure glucose and lactose.
0956-5663/02/$ - see front matter # 2003 Elsevier Science B.V. All rights reserved. PII: S 0 9 5 6 - 5 6 6 3 ( 0 2 ) 0 0 1 4 0 - 9
102
D.M. Jenkins, M.J. Delwiche / Biosensors and Bioelectronics 18 (2003) 101 /107
2. Objectives The main objective of this research was to build an instrument to measure the concentration of an analyte in solution by measuring the vacuum created in a closed cell by the depletion of a soluble gas. The sensor was used to measure glucose and lactose because these are representative of many substrates that may selectively be measured through enzymatic oxidation. Other objectives were to demonstrate the use of the instrument in a real biological fluid by measuring lactose concentrations in the milk of dairy cows, and to illustrate how the sample volume could be adjusted in a fixed volume reaction cell to give a broad range in sensitivity.
3. Theory The sensitivity of a manometric sensor using the depletion of a dissolved gas may be derived in a manner similar to that of a sensor using the liberation of a dissolved gas (Jenkins et al., 1999). A mass balance on the dissolved gas in the system yields, Pg(i) Vg P V b(i) Vl g(f) g b(f) Vl aVl ; (1) RT RT where Pg is the partial pressure of the dissolved gas, b is the concentration of dissolved gas in the sample, a is the concentration of dissolved gas required to react with the analyte in the sample, Vl is the volume of the sample, Vg is the volume of gas adjacent to the sample, R is the universal gas constant, and T is absolute temperature. Subscripts containing (i) represent the given quantity before the reaction with the dissolved gas, and those containing (f) represent the given quantity after the reaction with the dissolved gas. Assuming that the dissolved gas is in equilibrium in the system before and after the chemical reaction is allowed to occur, one may relate b to Pg using Henry’s Law: b Pg KH ;
(2)
where KH is an empirical constant for the dissolved gas in the given sample. Substituting Eq. (2) into Eq. (1) and rearranging, the vacuum (DP ) developed in the system is: DP Pg(i) Pg(f)
aVl RT Vg KH Vl RT
:
(3)
Inspection of this equation shows that the sensitivity of the device is adjustable over a broad range by changing the volume of gas in the system relative to the sample volume. By making the volume of gas small relative to the sample volume, the theoretical detection limit can be made small. To illustrate this, consider the enzymatic oxidation of glucose where 1 molecule of O2
oxidizes each molecule of glucose. Assuming a gas volume negligible compared to the sample volume, a KH value of 1.25 /10 8 M/Pa for O2 at 25 8C (derived from Lide, 1996, solubility data for O2 in pure water), and a pressure sensor precise to 10 Pa (the measured precision of the sensors used in this research), the theoretical detection limit would be 0.125 mM, or about 22.5 ng/ml.
4. Materials and methods 4.1. Determination of glucose The urea sensor used in previous research (Jenkins and Delwiche, 2002) was used with different reagents and with minor modifications to the hardware and fluid handling sequences. The hardware for fluid handling in the sensor (Fig. 1) consisted of a bank of pinch valves (161P011, Neptune Research Inc., West Caldwell, NJ) on a common manifold from which reagents could be pumped through a positive displacement diaphragm pump with a nominal stroke volume of 50 ml (120SP 12 50-4, Bio-Chem Valve Inc., Boonton, NJ) into a reaction cell made from Delrin. By energizing either a bleed or a waste pinch valve on the reaction cell (225P011-21, Neptune Research Inc.), the cell could be filled or flushed. The tubing used was a silicone based tubing made for the pinch valves (Neptune research Inc., TBGM107, 0.8 mm ID upstream of pump, and TBGM101, 1.5 mm ID downstream of pump). When the waste and bleed valves were closed, depletion of oxygen from the gas into solution could be measured as a change in vacuum in the reaction cell using a 10 kPa piezoresistive pressure sensor (MPX 2010 DP, Motorola, Phoenix, AZ). To measure glucose, five strokes of the sample were loaded into the reaction cell with 15 strokes of an enzyme solution containing glucose oxidase (EC # 1.1.3.4). The reaction cell was then sealed and shaken with a small DC motor for 30 s during which the change in vacuum was recorded. This shaking time was taken to be sufficient to observe most of the rapid changes in vacuum in the system (Fig. 2). The stroke volume of the pump was measured as 37 ml, and the volume of the reaction cell including dead space in the adjacent tubing was measured to be about 1.2 ml. The enzyme solution was prepared with 2 mg/ml of glucose oxidase isolated from Aspergillus niger (Product # G7141, 245.9 units/ mg, Sigma Aldrich Chemical Corp., St. Louis, MO) and 1 mg/ml of peroxidase from horseradish (EC # 1.11.2.7; Product # P8125, 116 purpurogallin units/mg, Sigma Aldrich Chemical Corp.) dissolved in a citrate/ascorbate buffer (50 mM citric acid and 34 mM ascorbic acid, pH 5.4). The glucose oxidase was used to oxidize b-Dglucose to d-D-gluconolactone and peroxide, and the
D.M. Jenkins, M.J. Delwiche / Biosensors and Bioelectronics 18 (2003) 101 /107
103
Fig. 1. Schematic of fluid handling hardware for glucose sensor.
and Delwiche, 2002), except that the reaction cell was ported to the negative port of the pressure sensor in order to record vacuum. Temperature effects on gas solubility (KH) and Eq. (3) lead to a dependence of instrument sensitivity on temperature. For previous research, a combination of physical and software compensation was used to correct for these effects (Jenkins and Delwiche, 2002). For this research, standard curves were made using data at a constant temperature (9/1 8C) so that temperature could be ignored, though application of the device in a situation without temperature control would require temperature compensation such as that used in previous research. Fig. 2. Vacuum (Patm/Psample) observed in glucose sensor, at 21 8C, during agitation of 5 mM lactose sample previously hydrolyzed with bgalactosidase.
peroxidase was used to peroxidate acorbic acid, thus preventing the spontaneous decomposition of peroxide to oxygen and water (Fig. 3). To prepare standards for analysis, 50 mM citrate buffer of pH 5.4 was used to dissolve D-glucose. To allow the mutarotation of glucose to come to equilibrium, the standards were left out overnight at room temperature. The wash solution used was distilled water, which was pumped through the reaction cell and waste and bleed lines after each analysis in a wash cycle identical to that of the urea sensor used in previous research (Jenkins and Delwiche, 2002). The pressure and temperature in the reaction cell were recorded as described in previous research (Jenkins
4.2. Determination of lactose Lactose was determined by first hydrolyzing the lactose to D-glucose and D-galactose with the enzyme b-galactosidase (EC # 3.2.1.23), and subsequently measuring the glucose as described above (Fig. 3). The b-galactosidase used was an industrial preparation isolated from Aspergillus oryzae (Product # G5160, 8.7 units/mg, Sigma Aldrich Chemical Corp.) that was homogenized with dextrin. Because dextrin was shown to interfere analytically with the determination of glucose, steps were taken to separate the dextrin from the enzyme prior to its application in the lactose assay. To separate the dextrin from b-galactosidase, 1.5 g of the commercial preparation were dissolved into 20 ml of EDTA buffer (10 mM EDTA, pH 6.9) along with 150 mg of a-amylase from porcine pancreas (EC # 3.2.1.1;
104
D.M. Jenkins, M.J. Delwiche / Biosensors and Bioelectronics 18 (2003) 101 /107
Fig. 3. Chemical pathways used for the measurement of glucose and lactose. Intermediate products of the enzyme mediated reactions and of the mutarotations of glucose and lactose are not shown. Depletion of O2 during the enzymatic oxidation of b-D-glucose to d-D-gluconolactone is measured as vacuum.
Product # A3176, 12.3 units/mg a-amylase activity, 3.8 units/mg b-amylase activity, Sigma Aldrich Chemical Corp.) to hydrolyze the dextrin into maltose. The resulting enzyme stock was sealed in dialysis tubing (Spectra/Por regenerated cellulose membrane, 12,000 / 14,000 molecular weight cutoff, 16 mm diameter, Spectrum Laboratories, Rancho Dominguez, CA), and dialyzed three times for 2 h in 1 l of the EDTA buffer at room temperature to remove maltose. The solution was then saturated with ammonium sulfate to precipitate the enzyme, which was separated by centrifugation (5 min at 14,000 /g ). The resulting protein pellet was redissolved in 5 ml of a 50 mM citrate buffer of pH 4.5. Lactose standards were prepared by dissolving Dlactose into the pH 4.5 citrate buffer. To allow the mutarotation of lactose to come to equilibrium, these standards were left out overnight at room temperature. The standards were incubated at room temperature with an aliquot of 1 volume of the purified b-galactosidase stock per 10 volumes of standard for at least 30 min. These were then assayed for glucose as described in the section above. Because milk solids tended to separate spontaneously during the incubation with the enzyme, milk samples were centrifuged to remove these components prior to the assay. These clarified samples were diluted 50/ into the pH 4.5 citrate buffer, then assayed for lactose as described above. To estimate the lactose concentration of the whole milk, the lactose estimated from the calibration equation was multiplied by the dilution factor (50) and corrected for the removal of fat and protein. For comparison, milk lactose was also measured using a fourier transform infrared (FTIR) instrument (DairyLab 1, Foss Electric, Hillerød, Denmark). The same instrument was used to estimate the
milk fat and protein in order to do the lactose correction for the manometric sensor. 4.3. Adjustment of sensitivity by the gas to sample volume ratio To illustrate how the sensitivity of the sensor could be adjusted by changing the sample volume loaded into a fixed volume reactor, glucose standards were assayed with the sensor as described above except that the liquid volume was varied. The embedded controller was reprogrammed to load different volumes of sample and enzyme (always one volume of standard to three volumes of enzyme) and new standard curves were generated at each volume. The sensitivity of the sensor at each loaded volume was then compared to the sensitivity predicted by Eq. (3).
5. Results and discussion 5.1. Measurement of glucose The manometric sensor was reasonably accurate for predicting glucose (Fig. 4). Accounting for the 4/ dilution of the sample with the enzyme stock, the sensitivity of the sensor predicted by Eq. (3) at 25 8C (0.950 kPa/mM) was almost twice the observed value (0.597 kPa/mM). Also, variations in pressure change recorded for standards of the same concentration (Fig. 4) were larger than errors in pressure observed with a similar sensor for urea (Jenkins and Delwiche, 2002). These deviations from the theory and from previous research were partly due to the spontaneous mutarota-
D.M. Jenkins, M.J. Delwiche / Biosensors and Bioelectronics 18 (2003) 101 /107
Fig. 4. Standard curve of sensor with glucose at 25 8C (Vacuum/ 0.597, Glucose /0.319; R2 /0.9849; S.E./0.23 mM).
tion of glucose (Fig. 3). Because glucose oxidase is only active on the b-D-glucose isomer (Keilan and Hartree, 1952a), any glucose in the a-D-glucose conformation could not be detected. Inspection of the vacuum profile observed over a sample (Fig. 2) reveals that the pool of available b-D-glucose was exhausted shortly after introduction of glucose oxidase. When this occurred, the oxidation of glucose was limited by the rate of mutarotation of the a-D-glucose isomer to the b-D-glucose isomer. In the absense of a biological catalyst, the rate constant for the mutarotation of glucose at 20 8C has been measured as 0.015/min in pure water (Keilan and Hartree, 1952a; Livingstone et al., 1977) and has been observed to range from 0.006 to 0.186/min in various aqueous solutions (Keilan and Hartree, 1952a,b; Livingstone et al., 1977; Pigman and Isbell, 1968). The transition to this mutarotation limited reaction rate caused the abrupt change in the rate of change of vacuum over the sample (Fig. 2). The ratio of observed to predicted sensitivity (62.8%) was similar to the amount of D-glucose measured to be in the b configuration at equilibrium (62.6%) by Pigman and Isbell (1968), suggesting that little of the a-D-glucose was converted to b-D-glucose during the 30 s reaction period in the sensor. Because the reaction was not allowed to proceed to completion, errors were introduced by variations in the rates of mutarotation, mass transfer of gas into solution, and enzymatic oxidation of glucose. 5.2. Measurement of lactose The 30 min incubation time of lactose standards with b-galactosidase was shown to be adequate to hydrolyze all of the lactose to glucose and galactose, and in fact the incubation period could have been made as short as 5 min without any loss of sensitivity (Fig. 5). The sensitivity (0.572 kPa/mM) and precision (9/0.19 mM) for the assay of lactose standards at 25 8C (Fig. 6) were
105
Fig. 5. Observed vacuum change in assay of 5 mM lactose standard against incubation time with b-galactosidase, 23 8C.
Fig. 6. Standard curve of sensor for lactose at 25 8C (Vacuum/ 0.572, Lactose /0.169; R2 /0.9892; S.E./0.19 mM).
similar to the respective values in glucose (0.597 kPa/ mM and 0.23 mM in Fig. 4). The slightly smaller sensitivity for lactose may have been partially attributable to the slight dilution of the lactose standards with b-galactosidase solution prior to assay. Because the vaporization of water leads to pressurization in the sensor, the differences in the intercepts of the calibration equations may have been due to differences in atmospheric humidity on the days that the standards were assayed. The slopes of the calibration equations, however, were reproducible, so that the equations could be estimated by assuming a slope and assaying a single standard. Alternatively, effects of background humidity and other dissolved gases could be compensated for by taking a background reading of pressure change in a sample without enzyme treatment (Jenkins and Delwiche, 2002). The estimate of lactose in milk was shown to be reasonably accurate compared to the reference FTIR instrument (Fig. 7). The correlation did not appear very
106
D.M. Jenkins, M.J. Delwiche / Biosensors and Bioelectronics 18 (2003) 101 /107
Fig. 8. Observed (m) and predicted ( */) sensitivity of glucose sensor at 25 8C for different ratios of gas to sample volume in fixed volume reactor.
Fig. 7. Comparison of lactose determination in milk by manometric sensor with values determined by FTIR spectroscopy (R2 /0.45; root mean squared deviation of sensor prediction from FTIR instrument is 8.1 mM, or about 5.8% of the average observation).
strong (R2 /0.45) because the milk samples collected had similar lactose concentrations. However, the root mean squared deviation of the two methods (8.1 mM) was only 5.8% of the average milk lactose value, showing that the two methods agreed well. Some of the observed error may have been attributable to the FTIR instrument. According to the distributor, the accuracy of the Dairy Lab 2 FTIR instrument is 9/1.2 mM for lactose in milk (calculated from Foss North America, 2001).
5.3. Adjustment of sensitivity by adjusting the loaded volume The observed sensitivities for glucose in the sensor for different volumes of sample and enzyme loaded followed the pattern predicted by Eq. (3) (Fig. 8). Higher sensitivity was observed when smaller volumes of gas were available to provide oxygen for a given volume of sample. Because the slow mutarotation of glucose prevented the endpoint of oxidation from being reached, the observed sensitivities at most loaded volumes were lower than predicted. Still, the observations corroborated the theory. Based on Eq. (3), the sensitivity of an oxygen depleting manometric sensor could be as high as 80 kPa/mM at 25 8C when the liquid volume is in excess of the available gas, the sample is undiluted, and the solubility of O2 in pure water is assumed. In the future, it may be possible to microfabricate sensors to achieve this level of sensitivity. Also, because the presence of
most electrolytes tends to decrease the solubility of O2, it is feasible that systems may achieve even greater sensitivities through the use of different buffers.
6. Conclusions A manometric sensor to measure milk urea through the enzymatic liberation of CO2 gas was successfully modified to measure glucose and lactose through the consumption of oxygen during the enzymatic oxidation of those substrates. The average error of the modified sensor was about 0.2 mM for both substrates in the range of 0 /5 mM. Much of the error was attributed to the slow mutarotation of D-glucose between the a (enzymatically inactive) and b (enzymatically active) forms which prevented the reaction from proceeding to completion within a reasonable period of time, and made the assay dependent on the rate of enzymatic oxidation and oxygen mass transfer. Many other oxidizable substrates, for example L-lactic acid through the enzyme lactate oxidase (EC # 1.1.3.2), could be more easily and accurately measured by allowing the reaction to proceed to completion because they do not spontaneously change conformation. Furthermore, the detection limit may be made much smaller (down to less than 1 mM) by a small adjustment in the amount of sample loaded into the sensor.
Acknowledgements This work was partially supported by USDA/BARD Research Project US-2638-95 and grants from DEC International NZ, Hamilton, New Zealand.
D.M. Jenkins, M.J. Delwiche / Biosensors and Bioelectronics 18 (2003) 101 /107
References Dixon, M., 1934. Manometric Methods. Cambridge University Press, London, UK, p. 1. Foss North America, 2001. DairyLab 2 (Specifications). Eden Prairie, MN, USA. Jenkins, D.M., 2001. Manometric Sensor to Measure Urea in Milk for Improvement of Dairy Cow Nutritional Management. Ph.D. Dissertation, University of California, Davis, CA, USA. Jenkins, D.M., Delwiche, M.J., 2002. Manometric biosensor for online measurement of milk urea. Biosensors and Bioelectronics 17, 557 /563.
107
Jenkins, D.M., Delwiche, M.J., DePeters, E.J., BonDurant, R.H., 1999. Chemical assay of urea for automated sensing in milk. Journal of Dairy Science 82, 1999 /2004. Keilan, D., Hartree, E.F., 1952. Specificity of glucose oxidase (notatin). Biochemical Journal 50, 331 /341. Keilan, D., Hartree, E.F., 1952. Biological catalysis of mutarotation of glucose. Biochemical Journal 50, 341 /348. Lide, D.R. (Ed.), CRC Handbook of Chemistry and Physics, 77th ed. CRC Press, New York, NY, USA 1996, pp. 4 /6. Livingstone, G., Franks, F., Aspinall, L.J., 1977. The effects of aqueous solvent structure on the mutarotation kinetics of glucose. Journal of Solution Chemistry 6, 203 /216. Pigman, W., Isbell, H.S., 1968. Mutarotation of sugars in solution: part I. Advances in Carbohydrate Chemistry 23, 11 /57.
Adaptation of a manometric biosensor to measure glucose and lactose Daniel M. Jenkins, Michael J. Delwiche * Department of Biological and Agricultural Engineering, University of California, 3048 Bainer Hall, Davis, CA 95616, USA Received 25 May 2001; received in revised form 17 April 2002; accepted 18 July 2002
Abstract A manometric sensor previously developed to measure urea was modified to measure glucose and lactose through enzymatic oxidation. Change in pressure in an enclosed cavity was correlated to the depletion of oxygen resulting from the enzymatic oxidation of glucose or lactose. The response of the sensor was linear and could be made adjustable over a large range by adjusting the amount of sample loaded into the fixed volume reactor. Because of the slow mutarotation of glucose, the oxidation of glucose was not allowed to proceed to completion. Therefore, the precision of the sensor (approximately 0.2 mM in a range from 0 to 5 mM) was limited by variations in the oxidation rate of glucose by glucose oxidase. Because the assay for lactose measured glucose subsequent to the hydrolysis of lactose by b-galactosidase, the same degree of precision was observed in lactose. Milk lactose, typically at concentrations of about 150 mM, was estimated using the lactose assay after first diluting the samples. For many fluids such as milk, the use of manometric sensors for oxidizable substrates may be preferable to optical and electrochemical methods because they are robust and suffer a low degree of optical and chemical interferences. Glucose and lactose are representative of many important oxidizable substrates, which may be determined in this manner, many of which do not suffer from limitations caused by mutarotation. In theory, detection limits less than 1 mM may be achieved using these methods. # 2003 Elsevier Science B.V. All rights reserved. Keywords: Enzyme sensor; Oxidation reaction; Pressure
1. Introduction A change in dissolved gas concentration can be measured in a closed cavity as a change in partial pressure of the gas (Dixon, 1934). Using this principle, it is possible to measure biological components of a fluid such a urea, which liberates CO2 when hydrolyzed by the enzyme urease. A biosensor using this principle for the on-line analysis of urea in milk (Jenkins and Delwiche, 2002) was robust to the difficult chemical and physical properties of milk. The precision (9/0.15 mM in buffer, 9/0.25 mM in milk, in a range from 0 to 10 mM) was comparable to that of other urea sensors. In many applications, including the measurement of milk urea, instruments suffer a high degree of chemical,
* Corresponding author: Tel.: /1-530-752-7023; fax: /1-530-7522640 E-mail address: [email protected] (M.J. Delwiche).
physical, optical, or electrical interference in the biological media which requires extensive pretreatment or dialysis of the samples (Jenkins, 2001, pp. 16 /32). The manometric sensor for urea adapted for this research was relatively inexpensive to construct, did not require pretreatment, and was shown to function reproducibly in the dirty and uncontrolled environment of a typical dairy, with a reagent cost of less than US $0.03 per sample (Jenkins, 2001). Most amino acids and some b-ketoacids such as acetoacetate may also be enzymatically decarboxylated and measured through the volatilization of CO2. Similarly, reactions depleting dissolved gases such as oxygen may be measured by the change in vacuum in a closed cell. Glucose and lactose are examples of the many important metabolic carbohydrates, which may be measured through enzymatic oxidation. To illustrate the design of a gas depleting type of manometric sensor, we modified the urea sensor to measure glucose and lactose.
0956-5663/02/$ - see front matter # 2003 Elsevier Science B.V. All rights reserved. PII: S 0 9 5 6 - 5 6 6 3 ( 0 2 ) 0 0 1 4 0 - 9
102
D.M. Jenkins, M.J. Delwiche / Biosensors and Bioelectronics 18 (2003) 101 /107
2. Objectives The main objective of this research was to build an instrument to measure the concentration of an analyte in solution by measuring the vacuum created in a closed cell by the depletion of a soluble gas. The sensor was used to measure glucose and lactose because these are representative of many substrates that may selectively be measured through enzymatic oxidation. Other objectives were to demonstrate the use of the instrument in a real biological fluid by measuring lactose concentrations in the milk of dairy cows, and to illustrate how the sample volume could be adjusted in a fixed volume reaction cell to give a broad range in sensitivity.
3. Theory The sensitivity of a manometric sensor using the depletion of a dissolved gas may be derived in a manner similar to that of a sensor using the liberation of a dissolved gas (Jenkins et al., 1999). A mass balance on the dissolved gas in the system yields, Pg(i) Vg P V b(i) Vl g(f) g b(f) Vl aVl ; (1) RT RT where Pg is the partial pressure of the dissolved gas, b is the concentration of dissolved gas in the sample, a is the concentration of dissolved gas required to react with the analyte in the sample, Vl is the volume of the sample, Vg is the volume of gas adjacent to the sample, R is the universal gas constant, and T is absolute temperature. Subscripts containing (i) represent the given quantity before the reaction with the dissolved gas, and those containing (f) represent the given quantity after the reaction with the dissolved gas. Assuming that the dissolved gas is in equilibrium in the system before and after the chemical reaction is allowed to occur, one may relate b to Pg using Henry’s Law: b Pg KH ;
(2)
where KH is an empirical constant for the dissolved gas in the given sample. Substituting Eq. (2) into Eq. (1) and rearranging, the vacuum (DP ) developed in the system is: DP Pg(i) Pg(f)
aVl RT Vg KH Vl RT
:
(3)
Inspection of this equation shows that the sensitivity of the device is adjustable over a broad range by changing the volume of gas in the system relative to the sample volume. By making the volume of gas small relative to the sample volume, the theoretical detection limit can be made small. To illustrate this, consider the enzymatic oxidation of glucose where 1 molecule of O2
oxidizes each molecule of glucose. Assuming a gas volume negligible compared to the sample volume, a KH value of 1.25 /10 8 M/Pa for O2 at 25 8C (derived from Lide, 1996, solubility data for O2 in pure water), and a pressure sensor precise to 10 Pa (the measured precision of the sensors used in this research), the theoretical detection limit would be 0.125 mM, or about 22.5 ng/ml.
4. Materials and methods 4.1. Determination of glucose The urea sensor used in previous research (Jenkins and Delwiche, 2002) was used with different reagents and with minor modifications to the hardware and fluid handling sequences. The hardware for fluid handling in the sensor (Fig. 1) consisted of a bank of pinch valves (161P011, Neptune Research Inc., West Caldwell, NJ) on a common manifold from which reagents could be pumped through a positive displacement diaphragm pump with a nominal stroke volume of 50 ml (120SP 12 50-4, Bio-Chem Valve Inc., Boonton, NJ) into a reaction cell made from Delrin. By energizing either a bleed or a waste pinch valve on the reaction cell (225P011-21, Neptune Research Inc.), the cell could be filled or flushed. The tubing used was a silicone based tubing made for the pinch valves (Neptune research Inc., TBGM107, 0.8 mm ID upstream of pump, and TBGM101, 1.5 mm ID downstream of pump). When the waste and bleed valves were closed, depletion of oxygen from the gas into solution could be measured as a change in vacuum in the reaction cell using a 10 kPa piezoresistive pressure sensor (MPX 2010 DP, Motorola, Phoenix, AZ). To measure glucose, five strokes of the sample were loaded into the reaction cell with 15 strokes of an enzyme solution containing glucose oxidase (EC # 1.1.3.4). The reaction cell was then sealed and shaken with a small DC motor for 30 s during which the change in vacuum was recorded. This shaking time was taken to be sufficient to observe most of the rapid changes in vacuum in the system (Fig. 2). The stroke volume of the pump was measured as 37 ml, and the volume of the reaction cell including dead space in the adjacent tubing was measured to be about 1.2 ml. The enzyme solution was prepared with 2 mg/ml of glucose oxidase isolated from Aspergillus niger (Product # G7141, 245.9 units/ mg, Sigma Aldrich Chemical Corp., St. Louis, MO) and 1 mg/ml of peroxidase from horseradish (EC # 1.11.2.7; Product # P8125, 116 purpurogallin units/mg, Sigma Aldrich Chemical Corp.) dissolved in a citrate/ascorbate buffer (50 mM citric acid and 34 mM ascorbic acid, pH 5.4). The glucose oxidase was used to oxidize b-Dglucose to d-D-gluconolactone and peroxide, and the
D.M. Jenkins, M.J. Delwiche / Biosensors and Bioelectronics 18 (2003) 101 /107
103
Fig. 1. Schematic of fluid handling hardware for glucose sensor.
and Delwiche, 2002), except that the reaction cell was ported to the negative port of the pressure sensor in order to record vacuum. Temperature effects on gas solubility (KH) and Eq. (3) lead to a dependence of instrument sensitivity on temperature. For previous research, a combination of physical and software compensation was used to correct for these effects (Jenkins and Delwiche, 2002). For this research, standard curves were made using data at a constant temperature (9/1 8C) so that temperature could be ignored, though application of the device in a situation without temperature control would require temperature compensation such as that used in previous research. Fig. 2. Vacuum (Patm/Psample) observed in glucose sensor, at 21 8C, during agitation of 5 mM lactose sample previously hydrolyzed with bgalactosidase.
peroxidase was used to peroxidate acorbic acid, thus preventing the spontaneous decomposition of peroxide to oxygen and water (Fig. 3). To prepare standards for analysis, 50 mM citrate buffer of pH 5.4 was used to dissolve D-glucose. To allow the mutarotation of glucose to come to equilibrium, the standards were left out overnight at room temperature. The wash solution used was distilled water, which was pumped through the reaction cell and waste and bleed lines after each analysis in a wash cycle identical to that of the urea sensor used in previous research (Jenkins and Delwiche, 2002). The pressure and temperature in the reaction cell were recorded as described in previous research (Jenkins
4.2. Determination of lactose Lactose was determined by first hydrolyzing the lactose to D-glucose and D-galactose with the enzyme b-galactosidase (EC # 3.2.1.23), and subsequently measuring the glucose as described above (Fig. 3). The b-galactosidase used was an industrial preparation isolated from Aspergillus oryzae (Product # G5160, 8.7 units/mg, Sigma Aldrich Chemical Corp.) that was homogenized with dextrin. Because dextrin was shown to interfere analytically with the determination of glucose, steps were taken to separate the dextrin from the enzyme prior to its application in the lactose assay. To separate the dextrin from b-galactosidase, 1.5 g of the commercial preparation were dissolved into 20 ml of EDTA buffer (10 mM EDTA, pH 6.9) along with 150 mg of a-amylase from porcine pancreas (EC # 3.2.1.1;
104
D.M. Jenkins, M.J. Delwiche / Biosensors and Bioelectronics 18 (2003) 101 /107
Fig. 3. Chemical pathways used for the measurement of glucose and lactose. Intermediate products of the enzyme mediated reactions and of the mutarotations of glucose and lactose are not shown. Depletion of O2 during the enzymatic oxidation of b-D-glucose to d-D-gluconolactone is measured as vacuum.
Product # A3176, 12.3 units/mg a-amylase activity, 3.8 units/mg b-amylase activity, Sigma Aldrich Chemical Corp.) to hydrolyze the dextrin into maltose. The resulting enzyme stock was sealed in dialysis tubing (Spectra/Por regenerated cellulose membrane, 12,000 / 14,000 molecular weight cutoff, 16 mm diameter, Spectrum Laboratories, Rancho Dominguez, CA), and dialyzed three times for 2 h in 1 l of the EDTA buffer at room temperature to remove maltose. The solution was then saturated with ammonium sulfate to precipitate the enzyme, which was separated by centrifugation (5 min at 14,000 /g ). The resulting protein pellet was redissolved in 5 ml of a 50 mM citrate buffer of pH 4.5. Lactose standards were prepared by dissolving Dlactose into the pH 4.5 citrate buffer. To allow the mutarotation of lactose to come to equilibrium, these standards were left out overnight at room temperature. The standards were incubated at room temperature with an aliquot of 1 volume of the purified b-galactosidase stock per 10 volumes of standard for at least 30 min. These were then assayed for glucose as described in the section above. Because milk solids tended to separate spontaneously during the incubation with the enzyme, milk samples were centrifuged to remove these components prior to the assay. These clarified samples were diluted 50/ into the pH 4.5 citrate buffer, then assayed for lactose as described above. To estimate the lactose concentration of the whole milk, the lactose estimated from the calibration equation was multiplied by the dilution factor (50) and corrected for the removal of fat and protein. For comparison, milk lactose was also measured using a fourier transform infrared (FTIR) instrument (DairyLab 1, Foss Electric, Hillerød, Denmark). The same instrument was used to estimate the
milk fat and protein in order to do the lactose correction for the manometric sensor. 4.3. Adjustment of sensitivity by the gas to sample volume ratio To illustrate how the sensitivity of the sensor could be adjusted by changing the sample volume loaded into a fixed volume reactor, glucose standards were assayed with the sensor as described above except that the liquid volume was varied. The embedded controller was reprogrammed to load different volumes of sample and enzyme (always one volume of standard to three volumes of enzyme) and new standard curves were generated at each volume. The sensitivity of the sensor at each loaded volume was then compared to the sensitivity predicted by Eq. (3).
5. Results and discussion 5.1. Measurement of glucose The manometric sensor was reasonably accurate for predicting glucose (Fig. 4). Accounting for the 4/ dilution of the sample with the enzyme stock, the sensitivity of the sensor predicted by Eq. (3) at 25 8C (0.950 kPa/mM) was almost twice the observed value (0.597 kPa/mM). Also, variations in pressure change recorded for standards of the same concentration (Fig. 4) were larger than errors in pressure observed with a similar sensor for urea (Jenkins and Delwiche, 2002). These deviations from the theory and from previous research were partly due to the spontaneous mutarota-
D.M. Jenkins, M.J. Delwiche / Biosensors and Bioelectronics 18 (2003) 101 /107
Fig. 4. Standard curve of sensor with glucose at 25 8C (Vacuum/ 0.597, Glucose /0.319; R2 /0.9849; S.E./0.23 mM).
tion of glucose (Fig. 3). Because glucose oxidase is only active on the b-D-glucose isomer (Keilan and Hartree, 1952a), any glucose in the a-D-glucose conformation could not be detected. Inspection of the vacuum profile observed over a sample (Fig. 2) reveals that the pool of available b-D-glucose was exhausted shortly after introduction of glucose oxidase. When this occurred, the oxidation of glucose was limited by the rate of mutarotation of the a-D-glucose isomer to the b-D-glucose isomer. In the absense of a biological catalyst, the rate constant for the mutarotation of glucose at 20 8C has been measured as 0.015/min in pure water (Keilan and Hartree, 1952a; Livingstone et al., 1977) and has been observed to range from 0.006 to 0.186/min in various aqueous solutions (Keilan and Hartree, 1952a,b; Livingstone et al., 1977; Pigman and Isbell, 1968). The transition to this mutarotation limited reaction rate caused the abrupt change in the rate of change of vacuum over the sample (Fig. 2). The ratio of observed to predicted sensitivity (62.8%) was similar to the amount of D-glucose measured to be in the b configuration at equilibrium (62.6%) by Pigman and Isbell (1968), suggesting that little of the a-D-glucose was converted to b-D-glucose during the 30 s reaction period in the sensor. Because the reaction was not allowed to proceed to completion, errors were introduced by variations in the rates of mutarotation, mass transfer of gas into solution, and enzymatic oxidation of glucose. 5.2. Measurement of lactose The 30 min incubation time of lactose standards with b-galactosidase was shown to be adequate to hydrolyze all of the lactose to glucose and galactose, and in fact the incubation period could have been made as short as 5 min without any loss of sensitivity (Fig. 5). The sensitivity (0.572 kPa/mM) and precision (9/0.19 mM) for the assay of lactose standards at 25 8C (Fig. 6) were
105
Fig. 5. Observed vacuum change in assay of 5 mM lactose standard against incubation time with b-galactosidase, 23 8C.
Fig. 6. Standard curve of sensor for lactose at 25 8C (Vacuum/ 0.572, Lactose /0.169; R2 /0.9892; S.E./0.19 mM).
similar to the respective values in glucose (0.597 kPa/ mM and 0.23 mM in Fig. 4). The slightly smaller sensitivity for lactose may have been partially attributable to the slight dilution of the lactose standards with b-galactosidase solution prior to assay. Because the vaporization of water leads to pressurization in the sensor, the differences in the intercepts of the calibration equations may have been due to differences in atmospheric humidity on the days that the standards were assayed. The slopes of the calibration equations, however, were reproducible, so that the equations could be estimated by assuming a slope and assaying a single standard. Alternatively, effects of background humidity and other dissolved gases could be compensated for by taking a background reading of pressure change in a sample without enzyme treatment (Jenkins and Delwiche, 2002). The estimate of lactose in milk was shown to be reasonably accurate compared to the reference FTIR instrument (Fig. 7). The correlation did not appear very
106
D.M. Jenkins, M.J. Delwiche / Biosensors and Bioelectronics 18 (2003) 101 /107
Fig. 8. Observed (m) and predicted ( */) sensitivity of glucose sensor at 25 8C for different ratios of gas to sample volume in fixed volume reactor.
Fig. 7. Comparison of lactose determination in milk by manometric sensor with values determined by FTIR spectroscopy (R2 /0.45; root mean squared deviation of sensor prediction from FTIR instrument is 8.1 mM, or about 5.8% of the average observation).
strong (R2 /0.45) because the milk samples collected had similar lactose concentrations. However, the root mean squared deviation of the two methods (8.1 mM) was only 5.8% of the average milk lactose value, showing that the two methods agreed well. Some of the observed error may have been attributable to the FTIR instrument. According to the distributor, the accuracy of the Dairy Lab 2 FTIR instrument is 9/1.2 mM for lactose in milk (calculated from Foss North America, 2001).
5.3. Adjustment of sensitivity by adjusting the loaded volume The observed sensitivities for glucose in the sensor for different volumes of sample and enzyme loaded followed the pattern predicted by Eq. (3) (Fig. 8). Higher sensitivity was observed when smaller volumes of gas were available to provide oxygen for a given volume of sample. Because the slow mutarotation of glucose prevented the endpoint of oxidation from being reached, the observed sensitivities at most loaded volumes were lower than predicted. Still, the observations corroborated the theory. Based on Eq. (3), the sensitivity of an oxygen depleting manometric sensor could be as high as 80 kPa/mM at 25 8C when the liquid volume is in excess of the available gas, the sample is undiluted, and the solubility of O2 in pure water is assumed. In the future, it may be possible to microfabricate sensors to achieve this level of sensitivity. Also, because the presence of
most electrolytes tends to decrease the solubility of O2, it is feasible that systems may achieve even greater sensitivities through the use of different buffers.
6. Conclusions A manometric sensor to measure milk urea through the enzymatic liberation of CO2 gas was successfully modified to measure glucose and lactose through the consumption of oxygen during the enzymatic oxidation of those substrates. The average error of the modified sensor was about 0.2 mM for both substrates in the range of 0 /5 mM. Much of the error was attributed to the slow mutarotation of D-glucose between the a (enzymatically inactive) and b (enzymatically active) forms which prevented the reaction from proceeding to completion within a reasonable period of time, and made the assay dependent on the rate of enzymatic oxidation and oxygen mass transfer. Many other oxidizable substrates, for example L-lactic acid through the enzyme lactate oxidase (EC # 1.1.3.2), could be more easily and accurately measured by allowing the reaction to proceed to completion because they do not spontaneously change conformation. Furthermore, the detection limit may be made much smaller (down to less than 1 mM) by a small adjustment in the amount of sample loaded into the sensor.
Acknowledgements This work was partially supported by USDA/BARD Research Project US-2638-95 and grants from DEC International NZ, Hamilton, New Zealand.
D.M. Jenkins, M.J. Delwiche / Biosensors and Bioelectronics 18 (2003) 101 /107
References Dixon, M., 1934. Manometric Methods. Cambridge University Press, London, UK, p. 1. Foss North America, 2001. DairyLab 2 (Specifications). Eden Prairie, MN, USA. Jenkins, D.M., 2001. Manometric Sensor to Measure Urea in Milk for Improvement of Dairy Cow Nutritional Management. Ph.D. Dissertation, University of California, Davis, CA, USA. Jenkins, D.M., Delwiche, M.J., 2002. Manometric biosensor for online measurement of milk urea. Biosensors and Bioelectronics 17, 557 /563.
107
Jenkins, D.M., Delwiche, M.J., DePeters, E.J., BonDurant, R.H., 1999. Chemical assay of urea for automated sensing in milk. Journal of Dairy Science 82, 1999 /2004. Keilan, D., Hartree, E.F., 1952. Specificity of glucose oxidase (notatin). Biochemical Journal 50, 331 /341. Keilan, D., Hartree, E.F., 1952. Biological catalysis of mutarotation of glucose. Biochemical Journal 50, 341 /348. Lide, D.R. (Ed.), CRC Handbook of Chemistry and Physics, 77th ed. CRC Press, New York, NY, USA 1996, pp. 4 /6. Livingstone, G., Franks, F., Aspinall, L.J., 1977. The effects of aqueous solvent structure on the mutarotation kinetics of glucose. Journal of Solution Chemistry 6, 203 /216. Pigman, W., Isbell, H.S., 1968. Mutarotation of sugars in solution: part I. Advances in Carbohydrate Chemistry 23, 11 /57.