- Email: [email protected]
Manometric transduction in enzyme biosensors
Biosensors and Bioelectronics 22 (2006) 94–101
Manometric transduction in enzyme biosensors Beatriz Fernandez Serra 1 , Achilles Tzoris, Elizabeth A.H. Hall ∗ Institute of Biotechnology, University of Cambridge, Tennis Court Road, Cambridge CB2 1QT, UK Received 19 September 2005; received in revised form 28 November 2005; accepted 13 December 2005 Available online 23 January 2006
Abstract The combination of enzymatic recognition and manometric transduction is explored, using enzymes that consume or evolve a gas with low solubility in aqueous media. A design is discussed whereby change in partial pressure of a gas in the headspace is related to the turnover of analyte by the enzyme. Headspace and sample volume dimensions are considered, demonstrating the influence of flux at the air–water interface. The relative importance of diffusion and reaction for the enzyme solution is shown. When enzyme kinetics dominate, the concentration gradient is low and the overall kinetics are determined by the total amount of active enzyme, reducing either enzyme concentration or enzyme layer thickness will reduce the diffusion limitation. A Teflon–enzyme composite is presented to allow a reuseable immobilised enzyme preparation and a disc with stirring magnet identified as an efficient configuration. A glucose oxidase system was tested in the monitoring of glucose consumption during fermentation. Application to other enzyme systems is discussed. © 2006 Elsevier B.V. All rights reserved. Keywords: Pressure sensor; Manometric transduction; Enzyme biosensors
1. Introduction A focus for enzyme biosensors has been driven by “home” and “field” use diagnostics (Ramsay, 1998), but process control management systems have also been cited (Rogers et al., 1995). Biosensors still mostly fall into categories of electrochemical, optical and acoustic: techniques that lend themselves to further innovation, but may not be the only or ‘best’ solutions for achieving an “in situ” type process control measurement, whether it be on- or off-line. Possible new designs of the classical biosensor may be limited but for enzymes such as oxidases, dehydrogenases, ammonia-liases or decarboxylases, oxygen is consumed or ammonia or carbon dioxide is formed, respectively. The possibility to base an enzyme linked assay on determination of the consumption or production of a gas is enormous. There are at least 50 known oxidases acting on fatty acids, sugars, amino acids, aldehydes, etc., around 80 decarboxylases, and dehydrogenases and ammonia-liases for every amino acid (Dixon and Webb, 1958). ∗
Corresponding author. Tel.: +44 1223 334149; fax: +44 1223 334162. E-mail address: [email protected] (E.A.H. Hall). 1 Current address: Faculty of Chemistry, University Complutense, Avda. Complutense s/n, 28040 Madrid, Spain. 0956-5663/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.bios.2005.12.009
Oxygen was also measured in the first glucose oxidase-linked biosensor (Clark and Lyons, 1962), but in that instance, was linked to an oxygen electrode. However, it is well established that a change in dissolved gas concentration can be measured in a closed cavity as a change in partial pressure of the gas (Dixon, 1934; Dixon and Webb, 1958). This simple concept is widely used in respirometry (Ros, 1993; Rozich and Gaudy, 1992; Spanjers, 1998; Copp et al., 2002), but deployment has been mainly confined to rather large (circa >250 mL) and animate (mainly microorganisms) samples. The fundamental principles of respirometry were recently revisited (Tzoris et al., 2002a,b) and it was shown that highly sensitive measurements can be made even on samples <1 mL. In comparison, pressure has scarcely been explored as a signal transduction mechanism for use in biosensors. Only Jenkins et al. (1999) and Jenkins and Delwiche (2002, 2003a,b) have reported enzyme biosensors based on manometric transduction for urea and glucose determination, while Sand et al. (2003) studied the possibility of using a pressure sensor for detecting immunorecognition through enzyme-labelling. In parallel with a design maximising sensitivity by minimising the headspace volume, VG (Tzoris et al., 2002a), Jenkins and Delwiche (2003a,b) have considered VG relative to the sample volume, VL , and proposed a minimum-headspace design. In both studies the nature of the interface between liquid and gas phases was important,
B.F. Serra et al. / Biosensors and Bioelectronics 22 (2006) 94–101
requiring rapid and efficient gas exchange and minimisation of concentration gradients leading to oxygen depletion in the solution phase. Jenkins and Delwiche for example, noted that sensitivity was limited by slow mass transfer across the membrane separating the two phases. However, based on our research in respirometry, and earlier work looking at diffusion through enzyme layers (Martens and Hall, 1994; Gooding and Hall, 1996) the diffusion path in the sample phase is also a key factor in finding lower sensitivities. In comparison with the classical biosensor, a manometric approach is interesting since there is no intimate contact between the pressure transducer and the enzyme. Turbidity, colour or sample matrix has little unforeseen effect on the measurement, which remains highly specific, since the substance to be monitored results from an enzyme catalysis. Only changes in the gas phase are measured without the need of mediation, labelling or other reagent in the system. In the work reported herein we explore the manometric enzyme linked assay further and test how this can be exploited for the determination of compounds that play an important role in the fermentation processes. 2. Materials and methods Warburg manometry was performed as described previously for respirometric measurements (Tzoris et al., 2002a). In short, the Manometer, a PVC pressure chamber (2.8 cm long, 2.2, 3.0 or 4.0 cm internal diameter (i.d.)) accommodated the enzyme pellet in the sample solution, leaving a defined volume headspace. A port interconnected the headspace with a Sensor Technics precision differential pressure transducer and with the atmosphere or reference chamber. The pressure range of the transducer was ±100 Pa and its response in Volts, with 1 V change corresponding to a pressure change of 100 Pa. The pressure chamber could be “closed” or “open” to external pressure. The sample was introduced into the chamber with a 1- or 2-ml syringe via a sample port in the lid. The syringe remained in place during measurement to ensure air tightness. Enzyme pellets were fabricated using glucose oxidase, GOx (EC 1.1.3.4, activity 121 U mg−1 of solid), lactate oxidase, LOD (EC.1.1.3.2, activity 35 U mg−1 of solid), glutamate oxidase, GluOD (EC.1.4.3.7, activity 10.8 U mg−1 of solid), glutamic dehydrogenase, GluDH (EC.1.4.1.4, activity 80 U mg−1 of solid), amino acid oxidase, AAOD (EC.1.4.3.2, activity 0.2 U mg−1 of solid) all from Sigma-Aldrich, Gillingham, Dorset, UK, and were accurately weighed and thoroughly mixed by hand with 0.3 g of Teflon powder (Sigma–Aldrich). The mixture was pressed into pellets by means of a Carver pellet press (Perkin-Elmer) at 10 000 kg cm−2 for 10 min. These pellets were of 1.3 cm diameter and approximately 0.1 cm thick. Pellets containing a magnet were fabricated in the same way, but placing the magnet in the hydraulic press, together with the powder mixture. Stock standard solutions (0.1 M) of d-glucose, l-lactate, l-glutamate, l-lysine, and NAD+ (0.01 M) (all from Sigma– Aldrich), were prepared in a 0.05 M phosphate buffer solution of appropriate pH. Other concentrations were prepared by suitable dilution with phosphate buffer solution.
95
Glucose monitoring in bacterial culture media: P. putida KT2440 (ATCC 47054), a gift from Dr. A. Tunnacliffe, of the Institute of Biotechnology, University of Cambridge, UK, was grown in 100 mL M9 minimal salts medium (0.5% (w/v)) glucose, M9 minimal salts (Sigma) 11.28 g/L, 1 mM MgSO4 ·7H2 O, 0.1 mM CaCl2 ·2H2 O, and 1 mL trace mineral solution (0.5 g/L of each FeSO4 , ZnSO4 , MnSO4 and 1 ml conc. H2 SO4 ) in 500mL Erlenmeyer flasks at 30 ◦ C under shaking (200 rpm). At fixed intervals following bacterial incubation, aliquots (1 mL) were taken and filtered through a 0.2 m filter to remove the bacteria. 0.1 mL of the aliquot was added to 0.9 mL of 0.05 M phosphate buffer solution (pH 7.2) and placed in the reference and sample chambers of the Manometer (1 mL total volume for each chamber). The differential measurement was started by adding glucose oxidase to the sensor chamber only. Pressure drop was measured for 5 min, and the response was compared with that of glucose standard solutions. Colorimetric assay for glucose (Barham and Trinder, 1972; Thomas, 1992) was based on hydrogen peroxide assay (formed when glucose is oxidised in the presence of GOx). A sample was taken after 8 min of reaction with GOx, and was added to 2 mL of 0.05 M phosphate buffer, pH 7.2, 0.45 units of HRP, 2 mL of 16.2 mg/mL phenol and 0.5 mg/mL 4-aminoantipyrine. The mixture was allowed to react at room temperature for 10 min, and its absorbance was measured at 510 nm. Absorbance values of the samples were compared to absorbance values of glucose standard solutions under the same conditions. 3. Results and discussion 3.1. Theoretical background The pressure drop (p) measured in the chamber is given by the Gas Laws: p =
nRT VG
(1)
At the end of the enzyme reaction with the analyte/substrate in solution, assumming no mass transfer limitation, p can be related to the oxygen consumed by the enzyme in solution: p =
RT (CLi − CLf ) VG VL
+ KH RT
(2)
where CLi and CLf represent the equilibrated initial and final dissolved oxygen concentrations respectively, KH the Henry’s solubility constant (typically, under standard conditions, oxygen solubility in pure water is ∼0.3 mM and alters by 1–2% for ∼200 Pa change in pressure, KH ∼ 1.25 × 10−8 M Pa−1 at 25 ◦ C), T the temperature, and R is the universal gas constant. If VG /VL KH RT then because the mole fraction of O2 dissolved in the liquid is small, the concentration of oxygen in the gas phase, CG , is approximately two orders of magnitude greater than that in the liquid phase, CL . Conversely if VG /VL KH RT then p ≈
CLi − CLf KH
(3)
96
B.F. Serra et al. / Biosensors and Bioelectronics 22 (2006) 94–101
Several issues have to be considered however, since p is obtained once all analyte has been consumed from the sample. Prior to this, the rate of mass transfer across a gas–liquid interface, the diffusion in the liquid phase and the enzyme kinetics all influence the equilibrium state of the system. When diffusion is slower than reaction, an oxygen depletion zone will develop. The concentration gradient depends on the diffusion/reaction balance. Oxygen uptake by a biomass layer has generally been expressed by a Michaelis Menten type equation (Ngian and Lin, 1976) with constant, Km . Even when CL Km , and diffusion in the liquid and biomass exceeds reaction by the biomass, the concentration of oxygen across the liquid layer may not be constant. For Km CL , the reaction will no longer be zero order. Under these conditions the rate of oxygen uptake will also depend on CL . From the work of Spanjers et al. (Spanjers, 1998) on respirometry in a flow system, dissolved oxygen mass balance has been given as follows: d(VL CL ) = Flowin CL,in − Flowout CL dt + VL KL a(CL∗ − CL ) − VL r
(4)
where t is the time, the subscript “in” for CL indicates the concentration in the entering “Flowin ” stream (subscript “out” indicates outflow) and CL∗ is the saturating concentration, r the respiration rate, subscript L refers to the liquid phase and. KL is the mass transfer coefficient based on the liquid phase and a is the specific surface area (a = A/VL , where A is the total interface area). In the gas phase oxygen mass balance can be presented as follows: d(VG CG ) VL = Flowin CG,in − Flowout CG − KL a(CL∗ − CL ) dt VG (5) In the work presented here, the “biomass” is an enzyme and the change in pressure in the headspace of the cell is measured at constant VG , with no flow. As a result of O2 consumption by enzyme in solution, O2 is transferred from the gas to the liquid controlled by interfacial flux: VL p = − RTKL a(CL∗ − CL ) t VG
(6)
In the liquid, the concentration gradient depends on the diffusion/reaction balance. The coefficient KL , the volumetric mass transfer coefficient for oxygen, describes mass transfer limitation in the liquid phase, kl (including the enzyme) and gas phase, kg : 1 RT 1 = + KL kl KH k g
(7)
Since kg kl and KH /RT is large (∼0.2) then 1/KL ∼ 1/kl . 3.2. Manometer design for glucose measurement For oxidase enzymes, with oxygen as cosubstrate, a direct relationship is expected between decrease in pressure measured
Fig. 1. Differential response for sample consisting of 1 mL of 1.25 mM dglucose in 0.05 M phosphate buffer, pH 7.2. Headspace: 1.14 cm3 . Cell diameter: 2 cm. Background and sensor chambers contain the same solution; sensor chamber also contains 140 units (nominally) of glucose oxidase. O2 ), and oxygen consumed by the enzymatic reaction (CL Km so pressure change will be related to the substrate of interest consumed via the enzyme reaction. If (i) the gas behaves ideally, (ii) mass transfer is not limiting and (iii) CL∗ = CL , the measurement of pressure will give an absolute value for O2 consumed; this is in contrast to amperometry or optical methods where measurement is proportional to O2 . Fig. 1 shows p measured in a closed pressure cell, for an unstirred sample containing d-glucose, when glucose oxidase (GOx) is added to the assay chamber. For [S] Km , it can be estimated (Eq. (2)) that 140 units of enzyme should produce a maximum pressure change rate of 291 kPa min−1 (VG /VL = 1.14). From the Michaelis Menten equation, O2 uptake rate at 1.25 mM glucose is 9.5 kPa min−1 ([S] < Km , taking Km = 37 mM reported in the literature for glucose oxidase, Takegawa et al., 1989). From the data reported here we obtain only 120 Pa min−1 or 5.8 × 10−5 mol L−1 min−1 O2 uptake. The cell is unstirred, so hydrodynamic effects on mass transfer should be negligible and the pressure change measured describes the flux at the air–water interface boundary (Eq. (6)). Without taking into account the enzyme reaction in solution, the transfer velocity kl is given by D (8) kl = πt
where D is the diffusion coefficient. However, the concentration gradient will also depend on the diffusion/reaction balance and the model must describe the enzyme kinetics and the effect of limitation of flux of oxygen (Martens and Hall, 1994; Gooding and Hall, 1996). This non-linear multi-boundary value problem does not have a general analytical solution, but in this instance, some insight into the limiting parameters can be gained just by substitution in Eq. (6). This gives an idea of the apparent oxygen depletion (CL∗ − CL )app . For the data from Fig. 1 the rate of pressure change is driven by (CL∗ − CL )app ∼ 8.8 × 10−8 mol cm−3 (taking CL∗ ∼ 2.7 × 10−7 mol cm−3 at the start of the assay). As shown previously (Martens and Hall, 1994), for the enzyme (E) reaction: k
k3
k4
1 Eox + S ⇔ ES −→Ered + P||Ered + O2 −→Eox + H2 O2
k2
(9)
B.F. Serra et al. / Biosensors and Bioelectronics 22 (2006) 94–101
where Eox and Ered are the oxidised and reduced forms of the enzyme, S the substrate and P is the product. ES is the enzyme–substrate complex. Et is the total enzyme. At a given substrate concentration, the concentration in the enzyme layer of thickness h (the depth of the 1 mL sample) varies according to the Thiele modulus (Φ2 ): Φ2 =
h2 k3 [Et ] DO2 [O2 ]sat
(10)
where DO2 is the diffusion coefficient for oxygen. This parameter describes the relative importance of diffusion and reaction in the enzyme layer. When Φ2 is small, enzyme kinetics dominate, the concentration gradient is low and the overall kinetics are determined by total amount of active enzyme. In contrast, when the Thiele modulus is large, diffusion limitations dominate and (CL∗ − CL ) increases. It can be seen that reducing either [Et ] or h will reduce Φ2 , thereby reducing the diffusion limitation. Reducing enzyme concentration in the cell to 20 units (predicted Vmax = 41 kPa min−1 ) gave an O2 consumption rate of 60 Pa min−1 at 1 mM glucose (expected rate when CL∗ = CL is 1 kPa min−1 at 1 mM glucose). These data correspond to (CL∗ − CL )app ∼3.3 × 10−8 mol cm3 , consistent with less oxygen limitation at lower enzyme concentrations. However, reducing enzyme concentration also reduces signal magnitude, but a higher rate of change of pressure requires higher flux across the interface and thus potentially a steeper concentration gradient (CL∗ − CL ) in the diffusion layer, δ, in the solution. For effective mass transfer between the two phases, a large surface area (see role of a in Eq. (6)) and hence, a large cell diameter, and thin VL is desired, so the system has more scope for optimisation than just maximising p through altering VG . It is necessary therefore, to reach a compromise between all these factors.
97
In unstirred solution the surface area at the gas liquid interface (or depth/thickness of the liquid phase) influences the response. Fig. 2a shows 20% increase is obtained by reducing sample depth from 3.1 to 1.4 mm. (CL∗ − CL )app , calculated using Eq. (6) also reduces (3.3 × 10−8 mol cm−3 at h = 3.1 mm; 2.3 × 10−8 mol cm−3 at h = 1.4 mm; 1.3 × 10−8 mol cm−3 at h = 0.8 mm) and confirms that oxygen diffusion through the liquid phase is probably important in maximising the response. When the solution is stirred, the effective surface area increases and the effective diffusion layer thickness, δ, decreases. This increases the flux at the interface, but diffusion limitation is minimised because the thickness of the depletion zone is minimised. Under these conditions kl = DO2 /δ, with kl(stirred) > kl(stagnant) and thus exchange between the gas and liquid phases and enzyme turnover should be accelerated. At 1 mM glucose with 20 units glucose oxidase (Fig. 2b), an order of magnitude increase in relative rate is obtained with stirring compared with Fig. 1 where 140 units were used without stirring (Table 1). Under these conditions, cell diameter became less influential. Jenkins and Delwiche (2003a,b) have proposed that reducing the VG /VL ratio gives better sensitivity; scrutiny of Eqs. (2) and (7) suggest that it is possible to adjust this. According to Eq. (2), the smaller the VG /VL ratio, the bigger the response for the same analyte concentration, and thus, the greater the sensitivity of the device. It would be attractive therefore to reduce VG . Jenkins and Delwiche (2003b), using >1000 units GOx/mL, proposed an immersible manometric sensor, protecting the gas/solution interface with a membrane but, remembering that kl will then be decreased since diffusion of oxygen through the membrane is lower than exchange across the air–water interface, this introduces greater diffusion limitation. The data above also suggest that lowering the VG /VL ratio alone will not achieve maximum response if mass transfer is limiting the response.
Fig. 2. Comparison of cell geometry. Response for of 1 mM d-glucose in 0.05 M phosphate buffer (pH 7.2) containing 20 units of GOx in (a) unstirred solutions and (b) stirred solutions. VL = 1 mL; VG = 1.14 cm3 ; cell diameter:liquid depth ratio: 2.2 cm:3.1 mm, 3 cm:1.4 mm, 4 cm:0.8 mm. Variation in response with (c) VG (VL = 1 mL) and (d) VL (VG = 0.76 mL) for 1.25 mM d-glucose in 0.05 M phosphate buffer (pH 7.2) containing 16.25 units of GOx, 2.2 cm cell diameter.
98
B.F. Serra et al. / Biosensors and Bioelectronics 22 (2006) 94–101
Table 1 Efficiency of the pressure change rate for different experimental conditions Units GOx
VG (cm3 )
VL (cm3 )
VG /VL
A (cm2 )
Vmax (kPa) (theory)
Glucose (mM)
p/t (kPa/min) (theory)
p/t (Pa/min) (exp) (R.S.D., %)
16.25 16.25 16.25 16.25 16.25 16.25 16.25 16.25 16.25 16.25 20 20 20 20 20 20 140
0.38 0.76 1.14 1.52 1.90 2.28 0.76 0.76 0.76 0.76 1.14 1.14 1.14 1.14 1.14 1.14 1.14
1.00 1.00 1.00 1.00 1.00 1.00 0.50 1.00 1.50 2.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
0.38 0.76 1.14 1.52 1.90 2.28 1.52 0.76 0.50 0.38 1.14 1.14 1.14 1.14 1.14 1.14 1.14
3.80 3.80 3.80 3.80 3.80 3.80 3.80 3.80 3.80 3.80 3.80 7.07 12.56 3.80 7.07 12.56 3.80
96.5 50.1 33.8 25.5 18.0 17.1 25.5 50.1 74.7 96.5 41.6 41.6 41.6 41.6 41.6 41.6 291
1.25 1.25 1.25 1.25 1.25 1.25 1.25 1.25 1.25 1.25 1.00 1.00 1.00 1.00 1.00 1.00 1.25
3.15 1.64 1.10 0.83 0.59 0.56 1.64 1.64 1.64 1.64 1.09 1.09 1.09 1.09 1.09 1.09 9.5
101.2 (9.9) 95.3 (11.5) 100.6 (11.9) 82.6 (16.9) 42.8 (23.4) 44.0 (18.2) 121.0 (8.3) 120.0 (7.5) 100.3 (8.0) 76.0 (11.8) 40 (13.9) 51 (17.3) 54 (8.9) 175 (4.2) 178 (13.5) 115 (14.7) 120 (7.6)
1.2 1.2 1.6 1.6 16.25 16.25 16.25 140 140
1.14 1.14 1.14 1.14 1.14 1.14 1.14 1.14 1.14
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
1.14 1.14 1.14 1.14 1.14 1.14 1.14 1.14 1.14
3.80 3.80 3.80 3.80 3.80 3.80 3.80 3.80 3.80
2.54 2.54 3.38 3.38 33.8 34.4 34.4 291 291
1.25 0.125 1.25 0.125 1.25 1.25 0.125 1.25 0.125
0.082 0.0085 0.109 0.0114 1.1 1.109 0.116 9.50 0.98
70 (13.4) 12 (29.7) 86 (11.6) 15 (24.5) 100.6 (8.6) 105 (10.0) 20 (18.4) 127 (9.7) 18 (25.7)
Efficiency (%) 3.2 (s) 5.8 (s) 9.1 (s) 10.0 (s) 7.3 (s) 7.9 (s) 7.4 (s) 7.3 (s) 6.1 (s) 4.6 (s) 3.6 (ns) 4.7 (ns) 5.0 (ns) 16.1 (s) 16.3 (s) 10.6 (s) 1.3 (ns) 85 (s) 100 (s) 79 (s) 100 (s) 9.1 (s) 9.5 (s) 17.2 (s) 1.3 (s) 1.8 (s)
s: stirred solution; ns: not stirred solution.
As oxygen is the second substrate, too small a headspace may result in overall oxygen limitation. So, it is necessary to take enzyme kinetics into account as well as designing the system to maximise observed pressure changes. Furthermore, reducing VG is geometricially limited by the diameter of the cell, so that from this aspect small diameters may be preferred, but as noted above, larger surface areas allow better gas-exchange so there is a peak in efficiency (Table 1; Fig. 2c). Eq. (2) predicts that increase in VL should increase sensitivity, whereas Eq. (6) suggests that sample depth (h = VL a) is more critical to the flux. Fig. 2d shows that a decrease in sample volume for the same interface area produces higher pressure drops, so that sample depth appears to be more important than sample volume. In comparing these results with those obtained using the data from Fig. 1, the importance of enzyme concentration is evident (Table 1). CL∗ is not generally achieved ‘instantaneously’, even when the gas/liquid phases are not divided by a diffusion limiting membrane so, low concentrations of enzyme and low glucose concentrations show pressure drops concomitant with a small (CL∗ − CL ), whereas at high enzyme and/or glucose concentrations (CL∗ − CL ) is larger and the oxygen exchange with the gas phase under diffusion control. Thus the magnitude of p becomes a versatile analytical signal. To obtain accurate kinetic data, low concentrations of enzyme are required and low glucose concentrations, whereas higher enzyme concentrations (>20 units) will allow measurement of glucose with little correc-
tion for enzyme concentration (data not shown due to restrictions imposed by editor). Based on these expectations, Fig. 3a shows the manometric response profile for glucose when GOx is in solution in excess (140 units). The initial rate of pressure drop (Fig. 3b) changes with glucose concentration for [glucose] < 1 mM, whereas for [glucose] > 2 mM, the initial rate remains the same. However, rates vary with concentration for times >150 s (Fig. 3a). If the enzyme substrate [S] < KM and [O2 ]L is not limiting, then dp/dt will give a rate dependent on [S]. At very long t, S will become consumed. From Fig. 3a maximum initial pressure change rate of ∼230 Pa min−1 is observed for [S] > 2 mM, corresponding to 9.4 × 10−8 mol min−1 . If the reaction proceeds to completion, p will give an absolute estimate of the total S present. For example, for glucose concentration of 2.5 × 10−7 mol mL−1 the rate begins to decline after ∼ 150 s corresponding to 2.1 × 10−7 mol consumed (see Fig. 3a). In principle the latter measurement requires no calibration and, depending on the sampling time chosen, gives a greater linear range for the calibration curve. For example, data taken after 15 min (Fig. 3c) gives a large pressure drop and the response up to 5 mM glucose is very reproducible. Shorter sampling times, (e.g. 5 min, Fig. 3d) do not consume all the substrate, so reduces the linear response range and requires separate calibration. Similarly, the initial pressure change rate also requires calibration (as seen above in Fig. 3b), and only reflects substrate concentration at low glucose concentrations, but reduces the time of
B.F. Serra et al. / Biosensors and Bioelectronics 22 (2006) 94–101
99
Fig. 3. (a). Differential pressure drop for different glucose concentrations in 1 mL 0.05 M phosphate buffer solution, pH 7.2. Headspace: 1.14 mL. Stirred solutions. 140 GOx units in solution. 1 cm cell diameter. Calibration of pressure measurement against glucose concentration using: (b) initial rate of pressure change; (c) differential pressure drop measured after 15 min; (d) differential pressure drop measured after 5 min.
analysis. Although this yields a relatively narrow linear range, it is a robust measurement with calibration plots obtained on different days (using VG /VL = 1.14) yielding initial slopes of 158 ± 0.3 Pa min−1 mM−1 . 3.3. Immobilization of enzyme
Fig. 4. (a) Calibration plots for glucose using differential pressure drop rate vs. glucose concentration. Background and sensor chamber contain 1 mL of glucose solution in 0.05 M phosphate buffer, pH 7.2. Stirred solutions. Headspace: 1.14 mL (sensor chamber also contains 565 units GOx–0.3 g Teflon pellet or 1815 units GOx–0.3 g Teflon pellet). (b) Storage stability of the Teflon–GOx pellet as measured with differential pressure drop rate measurements (n = 5) relating to 1.25 mM glucose (1 mL solution in 0.05 M phosphate buffer, pH 7.2) in contact with the Teflon–GOx pellet (2000 GOx units–0.3 g Teflon). Relative standard deviations (n = 5) for 0.125 and 1.25 mM glucose were 10 and 9%, respectively.
Classical gels and entrapment methods (Cosnier, 1999; Gerard et al., 2002; Mousty, 2004; Ferrer et al., 2004) would introduce mass transport limitation and thus reduce the signal measured. Del Cerro et al. (1997) and Pe˜na et al. (2001) have reported graphite–Teflon electrode biosensors, creating a rigid disc of physically entrapped enzyme. Enzyme–Teflon discs supply an amount of enzyme proportional to their surface area and can be retained in the cell for subsequent measurements. Fig. 4a shows the results obtained over a period of 20 days using the same disc. However, as predicted from the foregoing discussion, even in a stirred sample, the mass transfer between the gas phase and the enzyme on the surface of the disc is not instantaneous. This is highlighted by changing VL (and thus sample depth) while keeping VG constant as shown in Fig. 5a. According to Eq. (2), when VG /VL KH RT increasing VG should increase the change in pressure. On the contrary, Fig. 6 shows that (a) the response is delayed for 150 s for the greater VG and (b) the rate of pressure change for t > 150 s is lower for VG = 2 mL than for VG = 1 mL. To overcome these limitations the Teflon–enzyme disc was constructed with an embedded stirring magnet (Fig. 5b). This kept the enzyme suspended and moving in the sample and led to faster/higher responses with improved stability. For 1815 units GOx/pellet, sensitivity 84.48 mM min−1
100
B.F. Serra et al. / Biosensors and Bioelectronics 22 (2006) 94–101
which required the addition of reagents, and a time of analysis of 20 min. 4. Conclusions Manometric transduction for enzymatic reactions is reported with good sensitivity and great robustness although it is clear that full efficiency is not attainted, probably due to mass transport limitations. If this is the limiting factor then Eq. (4), predicting a simple relationship between VG and VL in tuning the sensitivity cannot be applied without a full understanding of the dimensions of the pressure cell and the effects on mass transport. Nevertheless, the signal obtained can be tuned using VG and VL to give a robust measurement with good sensitivity. The system is also appropriate for other substrates of interest in fermentation processes: l-lactate oxidase (140 units/pellet, sensitivity 67.4 mM min−1 ; detection limit 0.02 mM), l-glutamate oxidase (5 units/pellet, sensitivity 60 mM min−1 ; detection limit 0.1 mM) and l-amino acid oxidase (10 units/pellet, sensitivity 18.43 mM min−1 ; detection limit 0.1 mM), immobilized in Teflon pellets have been used for the measurements of l-lactate, l-glutamate and l-cysteine stock solutions, respectively. For glutamate determination, glutamic dehydrogenase results in ammonia evolution: l-glutamate + H2 O + NAD(P)+ glutamic-dehydrogenase
−→
Fig. 5. (a) Differential pressure drop for VL = 1 and 2 mL; 1.25 mM glucose in 0.05 M phosphate buffer solution, pH 7.2. Sensor chamber contains GOx (2000 units)–Teflon (0.3 g) pellet. (b) 1 mL 0.5 mM glucose in 0.05 M phosphate, pH 7.2. 1 mm headspace. Sensor chamber contains either GOx (2000 units)–Teflon (0.3 g) pellet or GOx (2000 units)–Teflon (0.1 g) pellet + embedded magnet bar. Headspace: 1.14 mL. Stirred solution (R.S.D. = 6%, n = 10 for 0.5 mM glucose solutions). (c) Glucose estimations during the growth of a bacterial culture in a minimal medium. obtained with the manometric biosensor and the colorimetric method.
was found using the initial rate method, with a detection limit of 0.1 mM. 3.4. Monitoring of glucose in bacterial culture media To study the possibility of applying this system in the monitoring of a fermentation process, Pseudomonas putida was grown in a M9 minimal salts medium, using glucose as sole carbon source. Glucose concentration was measured in aliquots from the medium taken at different times. Results are compared in Fig. 5c. The initial concentration of glucose in the medium was approximately 27 mM, and both manometric and colorimetric methods gave a close estimate of this value. Both methods yielded similar responses, the manometric one having the advantage of not requiring any additional reagent, and a much shorter time for analysis (5 min), in contrast to the colorimetric assay,
2-oxoglutarate + NH3 + NAD(P)H
An increase in the pressure in the cavity due to ammonia formation might be predicted, but the solubility of ammonia in aqueous solution is high, even for warm alkaline solutions, so no pressure change is detected. In contrast, for glutamate oxidase: l-glutamate + O2 + H2 O glutamate-oxidase
−→
2-oxoglutarate + NH3 + H2 O2
Both O2 uptake and NH3 evolution is involved, but since ammonia is not detected, it is possible to follow the reaction via oxygen consumption. Future work includes the design and fabrication of an automated sensor battery of manometric biosensors for on line measurement of glucose, lactate, ethanol, amino acids and glutamate during fermentation processes with automated action control protocol. Acknowledgements BS acknowledges the financial support (fellowship) from the Ministry of Science and Technology (Ministerio de Ciencia y Tecnolog´ıa), Spain. References Barham, D., Trinder, P., 1972. Analyst 97, 142–145. Clark, L.C., Lyons, C., 1962. Ann. N. Y. Acad. Sci. 105, 20–45.
B.F. Serra et al. / Biosensors and Bioelectronics 22 (2006) 94–101 Copp, J.B., Spanjers, H., et al., 2002. Respirometry in control of the activated sludge process: benchmarkingcontrol strategies. Iowa, London. Cosnier, S., 1999. Biosens. Bioelectr. 14, 443–456. Del Cerro, M.A., Cayuela, G., Reviejo, A.J., Pingaron, J.M., Wang, J., 1997. Electroanalysis 9, 1113–1119. Dixon, M., 1934. Manometric Methods. Cambridge University Press, London, UK. Dixon, M., Webb, E.C., 1958. Enzymes. Longman Group Limited, London, UK. Ferrer, M.L., Levy, D., Gomez-Lor, B., Iglesias, M., 2004. J. Mol. Catal. B: Enzym. 27, 107–111. Gerard, M., Chaubey, A., Malhotra, B.D., 2002. Biosens. Bioelectr. 17, 345–359. Gooding, J.J., Hall, E.A.H., 1996. J. Electroanal. Chem. 417, 25–33. Jenkins, D.M., Delwiche, M.J., Depeters, E.J., Bondurant, R.H., 1999. J. Dairy Sci. 82, 1999–2004. Jenkins, D.M., Delwiche, M.J., 2002. Biosens. Bioelectr. 17, 557–563. Jenkins, D.M., Delwiche, M.J., 2003a. Biosens. Bioelectr. 18, 101–107. Jenkins, D.M., Delwiche, M.J., 2003b. Biosens. Bioelectr. 18, 1085–1093. Martens, N., Hall, E.A.H., 1994. Anal. Chem. 66 (17), 2763–2770. Mousty, C., 2004. Appl. Clay Sci., 159–177. Ngian, K.F., Lin, S.H., 1976. Biotech. Bioeng. 28, 1623–1627. Pe˜na, N., Ruiz, G., Reviejo, A.J., Pingaron, J.M., 2001. Anal. Chem. 73, 1190–1195.
101
Ramsay, G., 1998. Commercial Biosensors: Applications to Clinical, Bioprocess, and Environmental Samples. John Wiley & Sons Cop., New York. Rogers, K.R., Mulchandani, A., et al., 1995. Biosensors and Chemical Sensor Technology: Process Monitoring and Control. American Chemical Society, Washington, DC. Ros, M., 1993. Respirometry of Activated Sludge. Technomic Publishing Co., Lancaster, PA. Rozich, A.F., Gaudy, A.F., 1992. Design and Operation of Activated Sludge Processes Using Respirometry. Lewis Publishers, Boca Raton. Sand, T.T., Zielinski, J.E., Arthur, C., Bradley, D., Wie, S., 2003. Biosens. Bioelectr. 18, 797–804. Spanjers, H., Q. International Association on Water (1998). Respirometry in Control of the Activated Sludge Process: Principles. International Association on Water Quality, London. Takegawa, K., Fujiwara, K., Iwahara, S., Yamamoto, K., Tochikura, T., 1989. Biochem. Cell Biol. 67, 460–464. Thomas, L., 1992. Labor and Diagnose, fourth ed., Die Medizinische Verlagsgesellschaft, Marburg, 169. Tzoris, A., Cane, D., Maynardc, P., Hall, E.A.H., 2002a. Anal. Chim. Acta 460 (2), 257–270. Tzoris, A., Fearnside, D., Lovell, M., Hall, E.A.H., 2002b. Environ. Toxicol. 17 (3), 284–290.
Manometric transduction in enzyme biosensors Beatriz Fernandez Serra 1 , Achilles Tzoris, Elizabeth A.H. Hall ∗ Institute of Biotechnology, University of Cambridge, Tennis Court Road, Cambridge CB2 1QT, UK Received 19 September 2005; received in revised form 28 November 2005; accepted 13 December 2005 Available online 23 January 2006
Abstract The combination of enzymatic recognition and manometric transduction is explored, using enzymes that consume or evolve a gas with low solubility in aqueous media. A design is discussed whereby change in partial pressure of a gas in the headspace is related to the turnover of analyte by the enzyme. Headspace and sample volume dimensions are considered, demonstrating the influence of flux at the air–water interface. The relative importance of diffusion and reaction for the enzyme solution is shown. When enzyme kinetics dominate, the concentration gradient is low and the overall kinetics are determined by the total amount of active enzyme, reducing either enzyme concentration or enzyme layer thickness will reduce the diffusion limitation. A Teflon–enzyme composite is presented to allow a reuseable immobilised enzyme preparation and a disc with stirring magnet identified as an efficient configuration. A glucose oxidase system was tested in the monitoring of glucose consumption during fermentation. Application to other enzyme systems is discussed. © 2006 Elsevier B.V. All rights reserved. Keywords: Pressure sensor; Manometric transduction; Enzyme biosensors
1. Introduction A focus for enzyme biosensors has been driven by “home” and “field” use diagnostics (Ramsay, 1998), but process control management systems have also been cited (Rogers et al., 1995). Biosensors still mostly fall into categories of electrochemical, optical and acoustic: techniques that lend themselves to further innovation, but may not be the only or ‘best’ solutions for achieving an “in situ” type process control measurement, whether it be on- or off-line. Possible new designs of the classical biosensor may be limited but for enzymes such as oxidases, dehydrogenases, ammonia-liases or decarboxylases, oxygen is consumed or ammonia or carbon dioxide is formed, respectively. The possibility to base an enzyme linked assay on determination of the consumption or production of a gas is enormous. There are at least 50 known oxidases acting on fatty acids, sugars, amino acids, aldehydes, etc., around 80 decarboxylases, and dehydrogenases and ammonia-liases for every amino acid (Dixon and Webb, 1958). ∗
Corresponding author. Tel.: +44 1223 334149; fax: +44 1223 334162. E-mail address: [email protected] (E.A.H. Hall). 1 Current address: Faculty of Chemistry, University Complutense, Avda. Complutense s/n, 28040 Madrid, Spain. 0956-5663/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.bios.2005.12.009
Oxygen was also measured in the first glucose oxidase-linked biosensor (Clark and Lyons, 1962), but in that instance, was linked to an oxygen electrode. However, it is well established that a change in dissolved gas concentration can be measured in a closed cavity as a change in partial pressure of the gas (Dixon, 1934; Dixon and Webb, 1958). This simple concept is widely used in respirometry (Ros, 1993; Rozich and Gaudy, 1992; Spanjers, 1998; Copp et al., 2002), but deployment has been mainly confined to rather large (circa >250 mL) and animate (mainly microorganisms) samples. The fundamental principles of respirometry were recently revisited (Tzoris et al., 2002a,b) and it was shown that highly sensitive measurements can be made even on samples <1 mL. In comparison, pressure has scarcely been explored as a signal transduction mechanism for use in biosensors. Only Jenkins et al. (1999) and Jenkins and Delwiche (2002, 2003a,b) have reported enzyme biosensors based on manometric transduction for urea and glucose determination, while Sand et al. (2003) studied the possibility of using a pressure sensor for detecting immunorecognition through enzyme-labelling. In parallel with a design maximising sensitivity by minimising the headspace volume, VG (Tzoris et al., 2002a), Jenkins and Delwiche (2003a,b) have considered VG relative to the sample volume, VL , and proposed a minimum-headspace design. In both studies the nature of the interface between liquid and gas phases was important,
B.F. Serra et al. / Biosensors and Bioelectronics 22 (2006) 94–101
requiring rapid and efficient gas exchange and minimisation of concentration gradients leading to oxygen depletion in the solution phase. Jenkins and Delwiche for example, noted that sensitivity was limited by slow mass transfer across the membrane separating the two phases. However, based on our research in respirometry, and earlier work looking at diffusion through enzyme layers (Martens and Hall, 1994; Gooding and Hall, 1996) the diffusion path in the sample phase is also a key factor in finding lower sensitivities. In comparison with the classical biosensor, a manometric approach is interesting since there is no intimate contact between the pressure transducer and the enzyme. Turbidity, colour or sample matrix has little unforeseen effect on the measurement, which remains highly specific, since the substance to be monitored results from an enzyme catalysis. Only changes in the gas phase are measured without the need of mediation, labelling or other reagent in the system. In the work reported herein we explore the manometric enzyme linked assay further and test how this can be exploited for the determination of compounds that play an important role in the fermentation processes. 2. Materials and methods Warburg manometry was performed as described previously for respirometric measurements (Tzoris et al., 2002a). In short, the Manometer, a PVC pressure chamber (2.8 cm long, 2.2, 3.0 or 4.0 cm internal diameter (i.d.)) accommodated the enzyme pellet in the sample solution, leaving a defined volume headspace. A port interconnected the headspace with a Sensor Technics precision differential pressure transducer and with the atmosphere or reference chamber. The pressure range of the transducer was ±100 Pa and its response in Volts, with 1 V change corresponding to a pressure change of 100 Pa. The pressure chamber could be “closed” or “open” to external pressure. The sample was introduced into the chamber with a 1- or 2-ml syringe via a sample port in the lid. The syringe remained in place during measurement to ensure air tightness. Enzyme pellets were fabricated using glucose oxidase, GOx (EC 1.1.3.4, activity 121 U mg−1 of solid), lactate oxidase, LOD (EC.1.1.3.2, activity 35 U mg−1 of solid), glutamate oxidase, GluOD (EC.1.4.3.7, activity 10.8 U mg−1 of solid), glutamic dehydrogenase, GluDH (EC.1.4.1.4, activity 80 U mg−1 of solid), amino acid oxidase, AAOD (EC.1.4.3.2, activity 0.2 U mg−1 of solid) all from Sigma-Aldrich, Gillingham, Dorset, UK, and were accurately weighed and thoroughly mixed by hand with 0.3 g of Teflon powder (Sigma–Aldrich). The mixture was pressed into pellets by means of a Carver pellet press (Perkin-Elmer) at 10 000 kg cm−2 for 10 min. These pellets were of 1.3 cm diameter and approximately 0.1 cm thick. Pellets containing a magnet were fabricated in the same way, but placing the magnet in the hydraulic press, together with the powder mixture. Stock standard solutions (0.1 M) of d-glucose, l-lactate, l-glutamate, l-lysine, and NAD+ (0.01 M) (all from Sigma– Aldrich), were prepared in a 0.05 M phosphate buffer solution of appropriate pH. Other concentrations were prepared by suitable dilution with phosphate buffer solution.
95
Glucose monitoring in bacterial culture media: P. putida KT2440 (ATCC 47054), a gift from Dr. A. Tunnacliffe, of the Institute of Biotechnology, University of Cambridge, UK, was grown in 100 mL M9 minimal salts medium (0.5% (w/v)) glucose, M9 minimal salts (Sigma) 11.28 g/L, 1 mM MgSO4 ·7H2 O, 0.1 mM CaCl2 ·2H2 O, and 1 mL trace mineral solution (0.5 g/L of each FeSO4 , ZnSO4 , MnSO4 and 1 ml conc. H2 SO4 ) in 500mL Erlenmeyer flasks at 30 ◦ C under shaking (200 rpm). At fixed intervals following bacterial incubation, aliquots (1 mL) were taken and filtered through a 0.2 m filter to remove the bacteria. 0.1 mL of the aliquot was added to 0.9 mL of 0.05 M phosphate buffer solution (pH 7.2) and placed in the reference and sample chambers of the Manometer (1 mL total volume for each chamber). The differential measurement was started by adding glucose oxidase to the sensor chamber only. Pressure drop was measured for 5 min, and the response was compared with that of glucose standard solutions. Colorimetric assay for glucose (Barham and Trinder, 1972; Thomas, 1992) was based on hydrogen peroxide assay (formed when glucose is oxidised in the presence of GOx). A sample was taken after 8 min of reaction with GOx, and was added to 2 mL of 0.05 M phosphate buffer, pH 7.2, 0.45 units of HRP, 2 mL of 16.2 mg/mL phenol and 0.5 mg/mL 4-aminoantipyrine. The mixture was allowed to react at room temperature for 10 min, and its absorbance was measured at 510 nm. Absorbance values of the samples were compared to absorbance values of glucose standard solutions under the same conditions. 3. Results and discussion 3.1. Theoretical background The pressure drop (p) measured in the chamber is given by the Gas Laws: p =
nRT VG
(1)
At the end of the enzyme reaction with the analyte/substrate in solution, assumming no mass transfer limitation, p can be related to the oxygen consumed by the enzyme in solution: p =
RT (CLi − CLf ) VG VL
+ KH RT
(2)
where CLi and CLf represent the equilibrated initial and final dissolved oxygen concentrations respectively, KH the Henry’s solubility constant (typically, under standard conditions, oxygen solubility in pure water is ∼0.3 mM and alters by 1–2% for ∼200 Pa change in pressure, KH ∼ 1.25 × 10−8 M Pa−1 at 25 ◦ C), T the temperature, and R is the universal gas constant. If VG /VL KH RT then because the mole fraction of O2 dissolved in the liquid is small, the concentration of oxygen in the gas phase, CG , is approximately two orders of magnitude greater than that in the liquid phase, CL . Conversely if VG /VL KH RT then p ≈
CLi − CLf KH
(3)
96
B.F. Serra et al. / Biosensors and Bioelectronics 22 (2006) 94–101
Several issues have to be considered however, since p is obtained once all analyte has been consumed from the sample. Prior to this, the rate of mass transfer across a gas–liquid interface, the diffusion in the liquid phase and the enzyme kinetics all influence the equilibrium state of the system. When diffusion is slower than reaction, an oxygen depletion zone will develop. The concentration gradient depends on the diffusion/reaction balance. Oxygen uptake by a biomass layer has generally been expressed by a Michaelis Menten type equation (Ngian and Lin, 1976) with constant, Km . Even when CL Km , and diffusion in the liquid and biomass exceeds reaction by the biomass, the concentration of oxygen across the liquid layer may not be constant. For Km CL , the reaction will no longer be zero order. Under these conditions the rate of oxygen uptake will also depend on CL . From the work of Spanjers et al. (Spanjers, 1998) on respirometry in a flow system, dissolved oxygen mass balance has been given as follows: d(VL CL ) = Flowin CL,in − Flowout CL dt + VL KL a(CL∗ − CL ) − VL r
(4)
where t is the time, the subscript “in” for CL indicates the concentration in the entering “Flowin ” stream (subscript “out” indicates outflow) and CL∗ is the saturating concentration, r the respiration rate, subscript L refers to the liquid phase and. KL is the mass transfer coefficient based on the liquid phase and a is the specific surface area (a = A/VL , where A is the total interface area). In the gas phase oxygen mass balance can be presented as follows: d(VG CG ) VL = Flowin CG,in − Flowout CG − KL a(CL∗ − CL ) dt VG (5) In the work presented here, the “biomass” is an enzyme and the change in pressure in the headspace of the cell is measured at constant VG , with no flow. As a result of O2 consumption by enzyme in solution, O2 is transferred from the gas to the liquid controlled by interfacial flux: VL p = − RTKL a(CL∗ − CL ) t VG
(6)
In the liquid, the concentration gradient depends on the diffusion/reaction balance. The coefficient KL , the volumetric mass transfer coefficient for oxygen, describes mass transfer limitation in the liquid phase, kl (including the enzyme) and gas phase, kg : 1 RT 1 = + KL kl KH k g
(7)
Since kg kl and KH /RT is large (∼0.2) then 1/KL ∼ 1/kl . 3.2. Manometer design for glucose measurement For oxidase enzymes, with oxygen as cosubstrate, a direct relationship is expected between decrease in pressure measured
Fig. 1. Differential response for sample consisting of 1 mL of 1.25 mM dglucose in 0.05 M phosphate buffer, pH 7.2. Headspace: 1.14 cm3 . Cell diameter: 2 cm. Background and sensor chambers contain the same solution; sensor chamber also contains 140 units (nominally) of glucose oxidase. O2 ), and oxygen consumed by the enzymatic reaction (CL Km so pressure change will be related to the substrate of interest consumed via the enzyme reaction. If (i) the gas behaves ideally, (ii) mass transfer is not limiting and (iii) CL∗ = CL , the measurement of pressure will give an absolute value for O2 consumed; this is in contrast to amperometry or optical methods where measurement is proportional to O2 . Fig. 1 shows p measured in a closed pressure cell, for an unstirred sample containing d-glucose, when glucose oxidase (GOx) is added to the assay chamber. For [S] Km , it can be estimated (Eq. (2)) that 140 units of enzyme should produce a maximum pressure change rate of 291 kPa min−1 (VG /VL = 1.14). From the Michaelis Menten equation, O2 uptake rate at 1.25 mM glucose is 9.5 kPa min−1 ([S] < Km , taking Km = 37 mM reported in the literature for glucose oxidase, Takegawa et al., 1989). From the data reported here we obtain only 120 Pa min−1 or 5.8 × 10−5 mol L−1 min−1 O2 uptake. The cell is unstirred, so hydrodynamic effects on mass transfer should be negligible and the pressure change measured describes the flux at the air–water interface boundary (Eq. (6)). Without taking into account the enzyme reaction in solution, the transfer velocity kl is given by D (8) kl = πt
where D is the diffusion coefficient. However, the concentration gradient will also depend on the diffusion/reaction balance and the model must describe the enzyme kinetics and the effect of limitation of flux of oxygen (Martens and Hall, 1994; Gooding and Hall, 1996). This non-linear multi-boundary value problem does not have a general analytical solution, but in this instance, some insight into the limiting parameters can be gained just by substitution in Eq. (6). This gives an idea of the apparent oxygen depletion (CL∗ − CL )app . For the data from Fig. 1 the rate of pressure change is driven by (CL∗ − CL )app ∼ 8.8 × 10−8 mol cm−3 (taking CL∗ ∼ 2.7 × 10−7 mol cm−3 at the start of the assay). As shown previously (Martens and Hall, 1994), for the enzyme (E) reaction: k
k3
k4
1 Eox + S ⇔ ES −→Ered + P||Ered + O2 −→Eox + H2 O2
k2
(9)
B.F. Serra et al. / Biosensors and Bioelectronics 22 (2006) 94–101
where Eox and Ered are the oxidised and reduced forms of the enzyme, S the substrate and P is the product. ES is the enzyme–substrate complex. Et is the total enzyme. At a given substrate concentration, the concentration in the enzyme layer of thickness h (the depth of the 1 mL sample) varies according to the Thiele modulus (Φ2 ): Φ2 =
h2 k3 [Et ] DO2 [O2 ]sat
(10)
where DO2 is the diffusion coefficient for oxygen. This parameter describes the relative importance of diffusion and reaction in the enzyme layer. When Φ2 is small, enzyme kinetics dominate, the concentration gradient is low and the overall kinetics are determined by total amount of active enzyme. In contrast, when the Thiele modulus is large, diffusion limitations dominate and (CL∗ − CL ) increases. It can be seen that reducing either [Et ] or h will reduce Φ2 , thereby reducing the diffusion limitation. Reducing enzyme concentration in the cell to 20 units (predicted Vmax = 41 kPa min−1 ) gave an O2 consumption rate of 60 Pa min−1 at 1 mM glucose (expected rate when CL∗ = CL is 1 kPa min−1 at 1 mM glucose). These data correspond to (CL∗ − CL )app ∼3.3 × 10−8 mol cm3 , consistent with less oxygen limitation at lower enzyme concentrations. However, reducing enzyme concentration also reduces signal magnitude, but a higher rate of change of pressure requires higher flux across the interface and thus potentially a steeper concentration gradient (CL∗ − CL ) in the diffusion layer, δ, in the solution. For effective mass transfer between the two phases, a large surface area (see role of a in Eq. (6)) and hence, a large cell diameter, and thin VL is desired, so the system has more scope for optimisation than just maximising p through altering VG . It is necessary therefore, to reach a compromise between all these factors.
97
In unstirred solution the surface area at the gas liquid interface (or depth/thickness of the liquid phase) influences the response. Fig. 2a shows 20% increase is obtained by reducing sample depth from 3.1 to 1.4 mm. (CL∗ − CL )app , calculated using Eq. (6) also reduces (3.3 × 10−8 mol cm−3 at h = 3.1 mm; 2.3 × 10−8 mol cm−3 at h = 1.4 mm; 1.3 × 10−8 mol cm−3 at h = 0.8 mm) and confirms that oxygen diffusion through the liquid phase is probably important in maximising the response. When the solution is stirred, the effective surface area increases and the effective diffusion layer thickness, δ, decreases. This increases the flux at the interface, but diffusion limitation is minimised because the thickness of the depletion zone is minimised. Under these conditions kl = DO2 /δ, with kl(stirred) > kl(stagnant) and thus exchange between the gas and liquid phases and enzyme turnover should be accelerated. At 1 mM glucose with 20 units glucose oxidase (Fig. 2b), an order of magnitude increase in relative rate is obtained with stirring compared with Fig. 1 where 140 units were used without stirring (Table 1). Under these conditions, cell diameter became less influential. Jenkins and Delwiche (2003a,b) have proposed that reducing the VG /VL ratio gives better sensitivity; scrutiny of Eqs. (2) and (7) suggest that it is possible to adjust this. According to Eq. (2), the smaller the VG /VL ratio, the bigger the response for the same analyte concentration, and thus, the greater the sensitivity of the device. It would be attractive therefore to reduce VG . Jenkins and Delwiche (2003b), using >1000 units GOx/mL, proposed an immersible manometric sensor, protecting the gas/solution interface with a membrane but, remembering that kl will then be decreased since diffusion of oxygen through the membrane is lower than exchange across the air–water interface, this introduces greater diffusion limitation. The data above also suggest that lowering the VG /VL ratio alone will not achieve maximum response if mass transfer is limiting the response.
Fig. 2. Comparison of cell geometry. Response for of 1 mM d-glucose in 0.05 M phosphate buffer (pH 7.2) containing 20 units of GOx in (a) unstirred solutions and (b) stirred solutions. VL = 1 mL; VG = 1.14 cm3 ; cell diameter:liquid depth ratio: 2.2 cm:3.1 mm, 3 cm:1.4 mm, 4 cm:0.8 mm. Variation in response with (c) VG (VL = 1 mL) and (d) VL (VG = 0.76 mL) for 1.25 mM d-glucose in 0.05 M phosphate buffer (pH 7.2) containing 16.25 units of GOx, 2.2 cm cell diameter.
98
B.F. Serra et al. / Biosensors and Bioelectronics 22 (2006) 94–101
Table 1 Efficiency of the pressure change rate for different experimental conditions Units GOx
VG (cm3 )
VL (cm3 )
VG /VL
A (cm2 )
Vmax (kPa) (theory)
Glucose (mM)
p/t (kPa/min) (theory)
p/t (Pa/min) (exp) (R.S.D., %)
16.25 16.25 16.25 16.25 16.25 16.25 16.25 16.25 16.25 16.25 20 20 20 20 20 20 140
0.38 0.76 1.14 1.52 1.90 2.28 0.76 0.76 0.76 0.76 1.14 1.14 1.14 1.14 1.14 1.14 1.14
1.00 1.00 1.00 1.00 1.00 1.00 0.50 1.00 1.50 2.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
0.38 0.76 1.14 1.52 1.90 2.28 1.52 0.76 0.50 0.38 1.14 1.14 1.14 1.14 1.14 1.14 1.14
3.80 3.80 3.80 3.80 3.80 3.80 3.80 3.80 3.80 3.80 3.80 7.07 12.56 3.80 7.07 12.56 3.80
96.5 50.1 33.8 25.5 18.0 17.1 25.5 50.1 74.7 96.5 41.6 41.6 41.6 41.6 41.6 41.6 291
1.25 1.25 1.25 1.25 1.25 1.25 1.25 1.25 1.25 1.25 1.00 1.00 1.00 1.00 1.00 1.00 1.25
3.15 1.64 1.10 0.83 0.59 0.56 1.64 1.64 1.64 1.64 1.09 1.09 1.09 1.09 1.09 1.09 9.5
101.2 (9.9) 95.3 (11.5) 100.6 (11.9) 82.6 (16.9) 42.8 (23.4) 44.0 (18.2) 121.0 (8.3) 120.0 (7.5) 100.3 (8.0) 76.0 (11.8) 40 (13.9) 51 (17.3) 54 (8.9) 175 (4.2) 178 (13.5) 115 (14.7) 120 (7.6)
1.2 1.2 1.6 1.6 16.25 16.25 16.25 140 140
1.14 1.14 1.14 1.14 1.14 1.14 1.14 1.14 1.14
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
1.14 1.14 1.14 1.14 1.14 1.14 1.14 1.14 1.14
3.80 3.80 3.80 3.80 3.80 3.80 3.80 3.80 3.80
2.54 2.54 3.38 3.38 33.8 34.4 34.4 291 291
1.25 0.125 1.25 0.125 1.25 1.25 0.125 1.25 0.125
0.082 0.0085 0.109 0.0114 1.1 1.109 0.116 9.50 0.98
70 (13.4) 12 (29.7) 86 (11.6) 15 (24.5) 100.6 (8.6) 105 (10.0) 20 (18.4) 127 (9.7) 18 (25.7)
Efficiency (%) 3.2 (s) 5.8 (s) 9.1 (s) 10.0 (s) 7.3 (s) 7.9 (s) 7.4 (s) 7.3 (s) 6.1 (s) 4.6 (s) 3.6 (ns) 4.7 (ns) 5.0 (ns) 16.1 (s) 16.3 (s) 10.6 (s) 1.3 (ns) 85 (s) 100 (s) 79 (s) 100 (s) 9.1 (s) 9.5 (s) 17.2 (s) 1.3 (s) 1.8 (s)
s: stirred solution; ns: not stirred solution.
As oxygen is the second substrate, too small a headspace may result in overall oxygen limitation. So, it is necessary to take enzyme kinetics into account as well as designing the system to maximise observed pressure changes. Furthermore, reducing VG is geometricially limited by the diameter of the cell, so that from this aspect small diameters may be preferred, but as noted above, larger surface areas allow better gas-exchange so there is a peak in efficiency (Table 1; Fig. 2c). Eq. (2) predicts that increase in VL should increase sensitivity, whereas Eq. (6) suggests that sample depth (h = VL a) is more critical to the flux. Fig. 2d shows that a decrease in sample volume for the same interface area produces higher pressure drops, so that sample depth appears to be more important than sample volume. In comparing these results with those obtained using the data from Fig. 1, the importance of enzyme concentration is evident (Table 1). CL∗ is not generally achieved ‘instantaneously’, even when the gas/liquid phases are not divided by a diffusion limiting membrane so, low concentrations of enzyme and low glucose concentrations show pressure drops concomitant with a small (CL∗ − CL ), whereas at high enzyme and/or glucose concentrations (CL∗ − CL ) is larger and the oxygen exchange with the gas phase under diffusion control. Thus the magnitude of p becomes a versatile analytical signal. To obtain accurate kinetic data, low concentrations of enzyme are required and low glucose concentrations, whereas higher enzyme concentrations (>20 units) will allow measurement of glucose with little correc-
tion for enzyme concentration (data not shown due to restrictions imposed by editor). Based on these expectations, Fig. 3a shows the manometric response profile for glucose when GOx is in solution in excess (140 units). The initial rate of pressure drop (Fig. 3b) changes with glucose concentration for [glucose] < 1 mM, whereas for [glucose] > 2 mM, the initial rate remains the same. However, rates vary with concentration for times >150 s (Fig. 3a). If the enzyme substrate [S] < KM and [O2 ]L is not limiting, then dp/dt will give a rate dependent on [S]. At very long t, S will become consumed. From Fig. 3a maximum initial pressure change rate of ∼230 Pa min−1 is observed for [S] > 2 mM, corresponding to 9.4 × 10−8 mol min−1 . If the reaction proceeds to completion, p will give an absolute estimate of the total S present. For example, for glucose concentration of 2.5 × 10−7 mol mL−1 the rate begins to decline after ∼ 150 s corresponding to 2.1 × 10−7 mol consumed (see Fig. 3a). In principle the latter measurement requires no calibration and, depending on the sampling time chosen, gives a greater linear range for the calibration curve. For example, data taken after 15 min (Fig. 3c) gives a large pressure drop and the response up to 5 mM glucose is very reproducible. Shorter sampling times, (e.g. 5 min, Fig. 3d) do not consume all the substrate, so reduces the linear response range and requires separate calibration. Similarly, the initial pressure change rate also requires calibration (as seen above in Fig. 3b), and only reflects substrate concentration at low glucose concentrations, but reduces the time of
B.F. Serra et al. / Biosensors and Bioelectronics 22 (2006) 94–101
99
Fig. 3. (a). Differential pressure drop for different glucose concentrations in 1 mL 0.05 M phosphate buffer solution, pH 7.2. Headspace: 1.14 mL. Stirred solutions. 140 GOx units in solution. 1 cm cell diameter. Calibration of pressure measurement against glucose concentration using: (b) initial rate of pressure change; (c) differential pressure drop measured after 15 min; (d) differential pressure drop measured after 5 min.
analysis. Although this yields a relatively narrow linear range, it is a robust measurement with calibration plots obtained on different days (using VG /VL = 1.14) yielding initial slopes of 158 ± 0.3 Pa min−1 mM−1 . 3.3. Immobilization of enzyme
Fig. 4. (a) Calibration plots for glucose using differential pressure drop rate vs. glucose concentration. Background and sensor chamber contain 1 mL of glucose solution in 0.05 M phosphate buffer, pH 7.2. Stirred solutions. Headspace: 1.14 mL (sensor chamber also contains 565 units GOx–0.3 g Teflon pellet or 1815 units GOx–0.3 g Teflon pellet). (b) Storage stability of the Teflon–GOx pellet as measured with differential pressure drop rate measurements (n = 5) relating to 1.25 mM glucose (1 mL solution in 0.05 M phosphate buffer, pH 7.2) in contact with the Teflon–GOx pellet (2000 GOx units–0.3 g Teflon). Relative standard deviations (n = 5) for 0.125 and 1.25 mM glucose were 10 and 9%, respectively.
Classical gels and entrapment methods (Cosnier, 1999; Gerard et al., 2002; Mousty, 2004; Ferrer et al., 2004) would introduce mass transport limitation and thus reduce the signal measured. Del Cerro et al. (1997) and Pe˜na et al. (2001) have reported graphite–Teflon electrode biosensors, creating a rigid disc of physically entrapped enzyme. Enzyme–Teflon discs supply an amount of enzyme proportional to their surface area and can be retained in the cell for subsequent measurements. Fig. 4a shows the results obtained over a period of 20 days using the same disc. However, as predicted from the foregoing discussion, even in a stirred sample, the mass transfer between the gas phase and the enzyme on the surface of the disc is not instantaneous. This is highlighted by changing VL (and thus sample depth) while keeping VG constant as shown in Fig. 5a. According to Eq. (2), when VG /VL KH RT increasing VG should increase the change in pressure. On the contrary, Fig. 6 shows that (a) the response is delayed for 150 s for the greater VG and (b) the rate of pressure change for t > 150 s is lower for VG = 2 mL than for VG = 1 mL. To overcome these limitations the Teflon–enzyme disc was constructed with an embedded stirring magnet (Fig. 5b). This kept the enzyme suspended and moving in the sample and led to faster/higher responses with improved stability. For 1815 units GOx/pellet, sensitivity 84.48 mM min−1
100
B.F. Serra et al. / Biosensors and Bioelectronics 22 (2006) 94–101
which required the addition of reagents, and a time of analysis of 20 min. 4. Conclusions Manometric transduction for enzymatic reactions is reported with good sensitivity and great robustness although it is clear that full efficiency is not attainted, probably due to mass transport limitations. If this is the limiting factor then Eq. (4), predicting a simple relationship between VG and VL in tuning the sensitivity cannot be applied without a full understanding of the dimensions of the pressure cell and the effects on mass transport. Nevertheless, the signal obtained can be tuned using VG and VL to give a robust measurement with good sensitivity. The system is also appropriate for other substrates of interest in fermentation processes: l-lactate oxidase (140 units/pellet, sensitivity 67.4 mM min−1 ; detection limit 0.02 mM), l-glutamate oxidase (5 units/pellet, sensitivity 60 mM min−1 ; detection limit 0.1 mM) and l-amino acid oxidase (10 units/pellet, sensitivity 18.43 mM min−1 ; detection limit 0.1 mM), immobilized in Teflon pellets have been used for the measurements of l-lactate, l-glutamate and l-cysteine stock solutions, respectively. For glutamate determination, glutamic dehydrogenase results in ammonia evolution: l-glutamate + H2 O + NAD(P)+ glutamic-dehydrogenase
−→
Fig. 5. (a) Differential pressure drop for VL = 1 and 2 mL; 1.25 mM glucose in 0.05 M phosphate buffer solution, pH 7.2. Sensor chamber contains GOx (2000 units)–Teflon (0.3 g) pellet. (b) 1 mL 0.5 mM glucose in 0.05 M phosphate, pH 7.2. 1 mm headspace. Sensor chamber contains either GOx (2000 units)–Teflon (0.3 g) pellet or GOx (2000 units)–Teflon (0.1 g) pellet + embedded magnet bar. Headspace: 1.14 mL. Stirred solution (R.S.D. = 6%, n = 10 for 0.5 mM glucose solutions). (c) Glucose estimations during the growth of a bacterial culture in a minimal medium. obtained with the manometric biosensor and the colorimetric method.
was found using the initial rate method, with a detection limit of 0.1 mM. 3.4. Monitoring of glucose in bacterial culture media To study the possibility of applying this system in the monitoring of a fermentation process, Pseudomonas putida was grown in a M9 minimal salts medium, using glucose as sole carbon source. Glucose concentration was measured in aliquots from the medium taken at different times. Results are compared in Fig. 5c. The initial concentration of glucose in the medium was approximately 27 mM, and both manometric and colorimetric methods gave a close estimate of this value. Both methods yielded similar responses, the manometric one having the advantage of not requiring any additional reagent, and a much shorter time for analysis (5 min), in contrast to the colorimetric assay,
2-oxoglutarate + NH3 + NAD(P)H
An increase in the pressure in the cavity due to ammonia formation might be predicted, but the solubility of ammonia in aqueous solution is high, even for warm alkaline solutions, so no pressure change is detected. In contrast, for glutamate oxidase: l-glutamate + O2 + H2 O glutamate-oxidase
−→
2-oxoglutarate + NH3 + H2 O2
Both O2 uptake and NH3 evolution is involved, but since ammonia is not detected, it is possible to follow the reaction via oxygen consumption. Future work includes the design and fabrication of an automated sensor battery of manometric biosensors for on line measurement of glucose, lactate, ethanol, amino acids and glutamate during fermentation processes with automated action control protocol. Acknowledgements BS acknowledges the financial support (fellowship) from the Ministry of Science and Technology (Ministerio de Ciencia y Tecnolog´ıa), Spain. References Barham, D., Trinder, P., 1972. Analyst 97, 142–145. Clark, L.C., Lyons, C., 1962. Ann. N. Y. Acad. Sci. 105, 20–45.
B.F. Serra et al. / Biosensors and Bioelectronics 22 (2006) 94–101 Copp, J.B., Spanjers, H., et al., 2002. Respirometry in control of the activated sludge process: benchmarkingcontrol strategies. Iowa, London. Cosnier, S., 1999. Biosens. Bioelectr. 14, 443–456. Del Cerro, M.A., Cayuela, G., Reviejo, A.J., Pingaron, J.M., Wang, J., 1997. Electroanalysis 9, 1113–1119. Dixon, M., 1934. Manometric Methods. Cambridge University Press, London, UK. Dixon, M., Webb, E.C., 1958. Enzymes. Longman Group Limited, London, UK. Ferrer, M.L., Levy, D., Gomez-Lor, B., Iglesias, M., 2004. J. Mol. Catal. B: Enzym. 27, 107–111. Gerard, M., Chaubey, A., Malhotra, B.D., 2002. Biosens. Bioelectr. 17, 345–359. Gooding, J.J., Hall, E.A.H., 1996. J. Electroanal. Chem. 417, 25–33. Jenkins, D.M., Delwiche, M.J., Depeters, E.J., Bondurant, R.H., 1999. J. Dairy Sci. 82, 1999–2004. Jenkins, D.M., Delwiche, M.J., 2002. Biosens. Bioelectr. 17, 557–563. Jenkins, D.M., Delwiche, M.J., 2003a. Biosens. Bioelectr. 18, 101–107. Jenkins, D.M., Delwiche, M.J., 2003b. Biosens. Bioelectr. 18, 1085–1093. Martens, N., Hall, E.A.H., 1994. Anal. Chem. 66 (17), 2763–2770. Mousty, C., 2004. Appl. Clay Sci., 159–177. Ngian, K.F., Lin, S.H., 1976. Biotech. Bioeng. 28, 1623–1627. Pe˜na, N., Ruiz, G., Reviejo, A.J., Pingaron, J.M., 2001. Anal. Chem. 73, 1190–1195.
101
Ramsay, G., 1998. Commercial Biosensors: Applications to Clinical, Bioprocess, and Environmental Samples. John Wiley & Sons Cop., New York. Rogers, K.R., Mulchandani, A., et al., 1995. Biosensors and Chemical Sensor Technology: Process Monitoring and Control. American Chemical Society, Washington, DC. Ros, M., 1993. Respirometry of Activated Sludge. Technomic Publishing Co., Lancaster, PA. Rozich, A.F., Gaudy, A.F., 1992. Design and Operation of Activated Sludge Processes Using Respirometry. Lewis Publishers, Boca Raton. Sand, T.T., Zielinski, J.E., Arthur, C., Bradley, D., Wie, S., 2003. Biosens. Bioelectr. 18, 797–804. Spanjers, H., Q. International Association on Water (1998). Respirometry in Control of the Activated Sludge Process: Principles. International Association on Water Quality, London. Takegawa, K., Fujiwara, K., Iwahara, S., Yamamoto, K., Tochikura, T., 1989. Biochem. Cell Biol. 67, 460–464. Thomas, L., 1992. Labor and Diagnose, fourth ed., Die Medizinische Verlagsgesellschaft, Marburg, 169. Tzoris, A., Cane, D., Maynardc, P., Hall, E.A.H., 2002a. Anal. Chim. Acta 460 (2), 257–270. Tzoris, A., Fearnside, D., Lovell, M., Hall, E.A.H., 2002b. Environ. Toxicol. 17 (3), 284–290.