摘要:

Critical Review Acoustic waves and the study of biochemical macromolecules and cells at the sensor–liquid interface Biljana A. � Cavi�c,a Gordon L. Haywardb and Michael Thompson*a a Department of Chemistry, University of Toronto, 80 St. George Street, Toronto, Ontario M5S 3H6, Canada b School of Engineering, University of Guelph, Guelph, Ontario N1G 2W1, Canada Received 22nd April 1999, Accepted 12th July 1999 1 Introduction 2 Overview of acoustic wave devices 3 Acoustic waves and the liquid–solid interface 4 Measurement techniques 5 Protein adsorption 6 Detection of interfacial immunochemical interactions 7 Nucleic acids and DNA/RNA–protein/peptide interactions.Drug discovery 8 Cell adhesion and cell function 9 Other applications 10 Concluding remarks and future perspectives 11 Acknowledgements 12 References I am skeptical of the principle of objectivity, which, in my view, is often simply the current popular viewpoint in disguise. Darryl Reanney1 1 Introduction The launching of acoustic waves in materials at ultrasonic frequencies, of the order of one to several hundred megahertz (MHz), depends upon the conversion of electromagnetic into acoustic energy. This process may involve piezoelectric substrates or the coupling of non-piezoelectric solids with devices capable of generating acoustic fields.Piezoelectricity is associated with the reversible electric polarization produced by mechanical strain in crystals belonging to certain classes.2 The Curie brothers have been universally credited with the discover- Professor Thompson is currently Professor of Analytical Chemistry at the University of Toronto, Canada. His research interests are concerned with the application of molecular recognition at interfaces, with emphasis on selective chemical and biological sensors for trace organic and clinical analysis.A graduate of the University College of Swansea and postgraduate of McMaster University, Ontario, Canada, he is on the Editorial Advisory Board of the journals ‘The Analyst’ and ‘Analytical Communications’ and was recently elected a Fellow of the Royal Society of Canada. Biljana A.. Cavi�c has been a researcher with Professor Michael Thompson at the Department of Chemistry, University of Toronto, since 1994. She received her PhD degree in 1993 from the University of Belgrade, where she was lecturing and participating in fundamental and industry related research projects. Biljana graduated from the University of Geneva in 1987 in environmental analytical chemistry and from the University of Belgrade in 1983 in chemical engineering. Her current research activities are in the areas of biosensors based on TSM acoustic wave devices, interfacial biomolecular interactions and hemocompatibility of silicone biomaterials.Gordon L. Hayward received his B.A.Sc., M.A.Sc. Ph.D. degrees in Chemical Engineering from the University of Waterloo in 1974, 1976 and 1981 respectively.His graduate research was in the field of aerodynamics and heat transfer. He then worked with Byte Craft Limited designing small computers for measurement systems until 1983 when he joined the faculty of the School of Engineering at the University of Guelph. The focus of his present research is the design of biological and chemical sensors, particularly ones based on bulk wave acoustic devices. In addition, some of his research involves fuzzy logic control systems, a natural extension of sensor technology into less tangible measurement problems.Dr. Hayward is a registered professional engineer in the province of Ontario. This journal is © The Royal Society of Chemistry 1999. Analyst, 1999, 124, 1405–1420 1405ing of this phenomenon (1880). From a chemical sensor standpoint it is interesting to note that the first application of piezoelectric effects in the development of such devices was that of Marie Curie.3 The Curies discovered the element radium and, in the course of their work, employed what was termed a ‘quartz electric balance’. It was about 60 years after the work of the Curies that Sauerbrey4 started using a thickness-shear mode (TSM) acoustic device, based on quartz plates, in the study of thin-deposited films.This research led to the famous Sauerbrey equation, which, essentially, treats any added layer on the device as being composed of quartz, that is, the new film is considered as extending the acoustic wavelength of the device.As we shall see, Sauerbrey’s expression is repeatedly invoked, even in modern times, to explain frequency responses in liquid media despite the fact that the original model was specified, reasonably, for the gas phase deposition of ultra-thin films. The first report of a TSM sensor being used as a selective analytical device concerned an application as a chromatographic detector.5 With respect to liquid phase operation, before the 1970s, it was commonly believed that the device could not function in such media because of viscous damping forces.6 However, in 1980, two separate groups were able to achieve operation in liquids by exposing only one face of the device to fluid.6,7 The first paper to suggest that acoustic wave devices could be employed in the field of biosensor technology involved the placement of antibody protein at the interface of a surface acoustic wave (SAW) sensor.8 Some 20 or so years earlier, White and Voltmer9 had pioneered the marriage of piezoelectricity with microelectronics through the demonstration that photolithographic techniques could be employed to deposit interdigital substrates to excite Rayleigh waves.Despite the obvious attractive possibilities for the SAW as a biosensor, later work cast serious doubt as to whether the sensor operates in a liquid in true Rayleigh format.10 Subsequent to this controversy, Thompson et al.11 successfully detected immunochemical interactions at the liquid–solid interface of a TSM device.Since this work, a number of different devices have been employed in conjunction with a wide variety of biological moieties, and much of this is detailed in the present review. Finally, before proceeding to review the theory and applications of acoustic wave technology in the field of biosensors, it is important to point out that not all such devices are based upon piezoelectric materials.These substrates can present a number of restrictions in terms of the attachment of biomolecules to the appropriate interface. In part, this is the reason that recent times have seen the introduction of thin rod, tube and magnetic acoustic resonator sensors where acoustic waves can be launched into non-piezoelectric materials. Although these structures have not yet seen use in biosensor development, we introduce them here because their physical structure and mode of excitation offer considerable potential for the future. 2 Overview of acoustic wave devices The operation of acoustic wave devices is based on the propagation of bulk or surface launched acoustic waves through piezoelectric and other materials.Quartz is the most frequently used piezoelectric matrix because of its stability with respect to temperature. Metal transducers launch acoustic waves at ultrasonic frequencies and the type and resonant frequency of such waves are determined by the crystal orientation, the thickness of the piezoelectric material and the geometry of the transducers.The generated waves are transverse or shear waves when particle displacement is perpendicular to the direction of propagation. In the case of compressional or longitudinal waves, particle displacement is parallel to the direction of travel. While bulk waves propagate through the volume of the piezoelectric substrate, the propagation of surface waves is restricted to a distance of about one acoustic wavelength from the piezoelectric substrate surface.With respect to shear wave devices, depending on the polarization of acoustic waves, particle displacement can be parallel or normal to the sensing surface creating horizontal or vertical shear waves.12–15 The most frequently used acoustic wave sensors are the thickness-shear mode devices that generate bulk transverse waves with particle dement parallel to the surface of the sensor. Such waves, on propagation into liquid, suffer attenuation in acoustic energy because of viscous forces in the liquid.However, the overall behaviour in liquids is governed not only by the viscosity of the liquid medium but also by the particular conditions present at the liquid–solid interface. The TSM device (Fig. 1) consists of an AT-cut quartz disc with metal electrodes on opposite sides to effect the application of an oscillating electric field. The applied transducer configuration categorizes the TSM as a one-port device.The thickness (t) determines the wavelengths (l) of the fundamental (n = 1) and harmonic (n = 3, 5, 7, …) resonances through the expression l = 2t/n. Consequently, the resonant frequency for the fundamental mode is restricted by the thickness of quartz wafers to the approximate range 5–20 MHz.12,13 SAW devices (Fig. 2) comprise thicker ST-cut quartz plates and interdigital metal transducers (IDTs) that generate Rayleigh waves propagating in both directions from the IDTs within the limit of approximately one acoustic wavelength from the sensing surface.The surface particles move elliptically, resulting in a wave consisting of both shear and compressional components. The latter provokes an important attenuation effect in liquids which prevents the application of SAW devices in such media. The particle displacements of the shear wave are transverse relative to the propagation direction and normal to the plane of the surface, so that the generated wave is categorized as shear vertical (SV).The most frequently applied, ‘delay line’, configuration of IDTs involves input and output transducers being positioned on two opposite sides of a Fig. 1 Top view of a TSM consisting of a quartz disc, metal electrodes and electric contacts. The smaller figure illustrates the wave motion and the direction of particle displacement. Fig. 2 Top view of a SAW delay line consisting of a quartz plate and interdigital metal transducers. The lower figure indicates the wave motion and direction of surface particle displacement. 1406 Analyst, 1999, 124, 1405–1420piezoelectric substrate. Such a structure forms a two-port device. The frequency of a SAW device is determined by the ‘finger’ spacing of the IDTs ranging from 30 to 500 MHz.12,15 An additional category of surface-launched acoustic wave sensors is those comprised of plate devices, where the thickness of a piezoelectric substrate is reduced to a dimension corresponding to the order of the acoustic wavelength. Although surface-excited, the generated waves travel through the bulk of the piezoelectric material.Lamb waves generated in a plate of finite thickness can be of both symmetric and antisymmetric modes when referring to the median plane of the plate. Antisymmetric Lamb waves exhibit flexural character and their velocity decreases with decreasing plate thickness. Flexural Love waves are composed of elliptical motion (as with Rayleigh waves) and, therefore, contain shear and compressional components.However, as a consequence of the minimal thickness of the plate, the wave velocity is lower than the compressional velocity of sound in liquids. Accordingly, compressional waves are not coupled in liquids, resulting in less attenuation of acoustic energy. The devices that generate such waves are termed flexural plate wave (FPW) devices [Fig. 3(A)] and can operate in liquids.FPW devices are microfabricated from very thin silicon or composite membranes with a sputtered ZnO piezoelectric layer. IDTs are positioned in a way to form a delay-line configuration. Since the piezoelectric plate is so thin, the operating frequencies of FPW devices are substantially lower than for other surface-launched acoustic wave devices (2–7 MHz).12,13,15 Another type of device in this category, shear horizontal acoustic plate (SH-APM) mode devices [Fig. 3(B)], produce waves where particle displacement is parallel to the sensing surface. Generated waves reflect between the plate surfaces which results in their superposition and the formation of a series of plate modes of different frequencies. With a configuration similar to typical SAW devices, SH-APM devices operate at frequencies in the range 25–200 MHz and are designed for tuning to a particular plate mode by adjusting the transducer bandwidth (determined by the number of finger pairs in the IDTs) and plate thickness. The device can operate in both liquid and gas media.12,15 Thick quartz plates of a different crystal orientation can produce shear horizontal surface waves (SH) parallel to the surface of devices.Structures producing this kind of wave are surface transverse wave (STW) devices [Fig. 3(C)] where a metal grating is used to prevent wave diffraction, and Lovewave devices which involve a structure where thin films are applied for the purpose of guiding acoustic waves.These acoustic devices also involve interdigital transducers.12 Lately, the development of acoustic wave devices that do not require piezoelectric materials for the propagation of acoustic waves has been under way. Thin-rod acoustic wave devices (TRAW) [Fig. 4(A)] are based on the propagation of acoustic waves in a cylindrical rod, which actually consists of a fibre with a radius much smaller than the acoustic wavelength. Piezoelectric input transducers generate and receive acoustic waves in a delay line configuration with respect to the thin rod.The rod itself does not have to be made of a piezoelectric substrate, because the acoustic wave propagation is based on a purely mechanical process. Two modes of vibration in TRAW devices are apparent, i.e., flexural and extensional. Whereas particle displacement for flexural waves is perpendicular to the plane of the thin rod surface, it is in the plane of the sensing surface for extensional waves, which results in an increased viscous damping on propagation into liquids.The possibility of increasing the mass sensitivity of the sensor by decreasing the fibre radius with facile operation of the device, particularly in the flexural mode, in liquid media, make this type of acoustic wave device appealing for the study of interfacial electrochemistry. 16,17 By introducing a tube instead of a thin rod, a sensor with a high mass sensitivity which is provided by a small thickness of the tube wall has been designed. Suitable geometry combined with only minor observed acoustic losses when the device is immersed in liquids makes it suitable for perspective application in on-line monitoring of physical and chemical changes in flow.18 The most recently described device exploits a coupling mechanism between electromagnetic and acoustic waves.A magnetic–acoustic–resonator sensor (MARS), presented in Fig. 4(B), consists of a non-piezoelectric acoustic plate which is excited from distance.Generation and detection of shear acoustic waves are performed by the means of a radiofrequency signal and a strong magnetic field. It was reported that the effect of mass loading of a metallized glass plate in air, incorporated into MARS device, can be predicted from Sauerbrey’s equation similarly to TSM devices.19 Table 1 depicts the comparison of some of the characteristics of the acoustic wave devices that have been described in brief above, together with the correlation of their responses to mass loading.Fig. 3 (A) FPW: schematic diagram of a wave motion and direction of particle displacement. (B) SH-APM: schematic diagram of a wave motion within the plate and direction of particle displacement for the n = 1 mode. (C) STW: the structure of the device and direction of particle displacement. Fig. 4 (A) The configuration of a TRAW system in flexural mode. (B) The configuration of a MARS device.Analyst, 1999, 124, 1405–1420 14073 Acoustic waves and the liquid–solid interface When a crystal whose lattice does not have a centre of symmetry is stressed, the asymmetric motion of charge centres produces an electric field.2 Conversely, when an electric field is placed across this type of crystal, the attraction and repulsion of the charge centres generate stress in the crystal. The converse piezoelectric effect provides a convenient method for generating motion at ultrasonic frequencies.The physical properties of various piezoelectric materials make each suitable for a particular application. For ultrasonic resonators, quartz is the material of choice. Its primary advantage is its high stiffness. Resonators made of quartz have low losses and very sharp resonances. Quartz is a hexagonal crystal with Cartesian axes shown in Fig. 5.20 The crystal may be rotated, but the X-axis passes through a vertex and the Y-axis passes at right-angles through a face.When a field is placed across the crystal parallel to the Xaxis, the thickness in the X direction increases. The X-axis is defined as positive at the side where a negative charge is generated by compressing the crystal. If the field is aligned with the Y-axis, the resulting deformation is a shear along the Z-axis. The Z-axis is electrically neutral. The quartz crystals used in sensor applications are wafers cut from a larger crystal. To achieve a shear motion, the cut must be normal to the Y-axis so that the field can be produced by charging electrodes placed on the two faces of the wafer.If the cut is rotated about the X-axis, the temperature coefficient changes markedly. The AT cut is normal to the Y-axis but rotated 35.25° from the Z-axis. This has a zero temperature coefficient at 40 °C, making the AT cut ideal for frequency control applications. The resulting mass production has made these crystals very inexpensive.Since quartz is anisotropic, most of its physical and electrical properties are direction dependent. The dielectric constant, elasticity and piezoelectric constants, and others, are expressed as matrices. Since the AT wafer is cut at a particular rotation and the electric field is placed across the wafer, the result is a set of scalar properties valid for the particular rotation. The field results in a shear deformation across the wafer with no thickness extension. When used as a thickness-shear mode sensor, the crystal is operated as a resonator.Since the applied field across an AT cut wafer causes both free surfaces to move in parallel but opposite directions, both surfaces are motion antinodes. This means that the thickness of the wafer will be an odd multiple of half of the acoustic wavelength in quartz. The velocity of the shear wave is given by20 n = (c/r)1/2 (1) where n is the acoustic shear wave velocity, r is the density of quartz and c is the stiffness calculated for the direction of motion and the wafer cut angle.This gives a resonant frequency f0 = nn/2dQ (2) where f0 is the resonant frequency in hertz, n is an odd integer and dQ is the thickness of the wafer. The quartz resonator is a mechanically resonant system which has two forms of energy storage, elastic deformation and inertia. Energy is transferred from one to the other at a characteristic rate dependent on the mass and stiffness of the resonator.These parameters are both represented in the above equations. This analysis works well for a crystal resonator operating in a vacuum. However, if the crystal is loaded by an additional mass attached to the surface or by immersion in a viscoelastic medium, the resonance behaviour will be affected. These loading effects provide the basis for the TSM and other resonant sensors. The first quartz sensors were developed by Sauerbrey4 to measure mass deposited on the crystal surface. The model, which is still widely used, considers the added mass to be simply an extension of the crystal thickness: Df = 22f0 2 (cr)21/2Dm/A (3) where Df is the resonant frequency shift due to the added mass, Dm is the added mass and A is the area of the crystal.This is valid for small amounts of added mass; however, the equation overpredicts the frequency shift when the added mass gives a frequency shift greater than about 2% of the unloaded resonant frequency. Miller and Bolef21 considered the loaded crystal as a compound resonator which gave Sauerbrey’s equation with higher order terms.This extends the model range to frequency shifts up to 15% of the unloaded value. When the resonator is immersed in a fluid, energy is lost to the fluid through viscous coupling. Kanazawa and Gordon 22 modelled the resonant frequency shift due to viscous coupling by assuming no slip between the fluid and the crystal surface. The resulting solution of the fluid equations of motion is a transverse shear wave that propagates into the liquid.This wave decays with a characteristic length d: d = (2hL/wrL)1/2 (4) where hL is the fluid viscosity and rL is the fluid density. In water this length is short, of the order of 0.25 mm. The frequency shift due to the fluid loading was also calculated: Df = 2f0 3/2 (rLhL/pcr)1/2 (5) Table 1 Survey of some of the properties of various acoustic wave devices12,15–18 Type of device Particle displacement (relative to the wave propagation direction) Typical operation frequency MHz Mass sensitivitya,b TSM Transverse 5–20 21/(rt) SAW Transverse and parallel component 30–500 2(f0K1)/(rV) FPW Transverse and parallel component 2–7 21/(2rt) SH-APM Transverse 25–200 2J/(rt) STW Transverse 200–500 2(f0K2)/(rV) TRAW Transverse (flexural mode) and parallel (extensional mode) 0.2–2 21/(nra) Tube Parallel (extensional mode) 2 21/[r(b 2 a1) (1 + a1/b)] a Mass sensitivity (Sm) is defined as Sm = lim (Dv)/(v0Dm)Dm?0, where: Dm is added mass, Dv is change in phase velocity and v0 is unperturbed phase velocity.b r, Density of the sensor material; t, plate thickness; f0, fundamental, unperturbed frequency of the device; K1, constant dependent on the properties of the plate of the SAW device; V, wave velocity; J, J = 1/2 for an isotropic plate mode (n1) n1 = 0 and J = 1 for n1 > 0; K2, constant dependent on the properties of the STW device; n, n = 1 for the first extension mode and n = 2 for the flexural mode; a, radius of the thin rod; b and a1, external and internal diameter of the tube, respectively.Fig. 5 Cartesian axes for a quartz crystal. 1408 Analyst, 1999, 124, 1405–1420When only the resonant frequency shift of a sensor operating in a fluid is measured, it is not possible to assess the individual contributions of mass and viscosity. To obtain additional information, the quartz resonator system can be modelled as an electrical equivalent circuit based on a simple analogy, the Butterworth–van Dyke model.20 A simple derivation may be made by performing a force balance on a damped mass and spring system as shown in Fig. 6. The force balance is given by m d2x/dt2 + kx + h dx/dt = F (6) where m is the mass, k is a spring constant, h is a damping coefficient and F is an applied force. Taking the Laplace transform of eqn. (6), dividing by the velocity and transforming the equation into the frequency domain with s = jw gives jwm + k/jw + h = F/V (7) where j is the square root of 21, w is an angular frequency and V is the velocity of the mass. F/V is a driving force divided by a flow like term, and so is analogous to a resistance.This provides the motional arm of the Butterworth–van Dyke (BVD) model of a quartz resonator. The impedance represented by eqn. (7) consists of an inductor arising from the inertial term, a capacitor arising from the elastic stress term and a resistor arising from the damping.The imaginary terms represent energy storage while the real resistance term represents energy dissipation. The complete BVD model of a resonator in an electric circuit, as shown in Fig. 7, includes a parallel capacitor representing dielectric energy storage. The previous model has been extended to include the effects of mass deposition and liquid loading.23 The deposited mass adds an inductor to the model, but the viscous loading adds both a resistor and an inductor.The resistance accounts for the energy dissipated by the fluid. Some of the energy transferred to the fluid is returned through viscous coupling so that the coupled fluid acts as an attached mass. Simpson24 showed that the impedance magnitude of the viscous resistance and the corresponding inductance are equal. By measuring the resistance and the resonant frequency shift, mass and viscous effects may be determined simultaneously.25 Hayward and Thompson26 developed a model of a chemical sensor based on the previous work of Reed et al.27 and Ferrante et al.28 The sensor was a coated crystal operating in a fluid where the coating represented a thin layer of a chemically or biologically sensitive agent.These models solved nine fundamental differential equations to predict the response of the sensor. The formulation of these models introduces additional sensor response mechanisms that have not yet been exploited. Reed et al.27 modelled the response of a resonator to the properties of an attached layer of fluid.Good agreement with experimental data was obtained for low molecular mass liquids; however, polymers gave lower losses and resonant frequency shifts than predicted. This was attributed to the elastic behaviour of these materials, particularly at the high frequencies used with quartz resonators. This film elasticity acts as an additional energy storage mechanism. Ferrante et al.28 developed a similar model, but the fluid layer was assumed to be only viscous and infinitely thick.This is the case when a resonator is immersed in a fluid medium. Changes in the losses and frequency shifts with the hydrophilic– hydrophobic nature of the surface were attributed to slip at the fluid-resonator interface. Here the fluid layer at the surface and the surface itself move at different velocities and out of phase. Amplitude calculations from the Hayward–Thompson model26 show that in water, the motion is approximately 30 nm V21.At this scale, slip at the surface is possible and will depend on the affinity between the surface and the medium. Hayward and Thompson also showed that the effect of this slip and that of viscoelasticity cannot be separated in the resonance model, however they occur together. Recently, Bandey et al. developed a very versatile, general equivalent circuit model describing TSM impedance response under different types of loadings, varying from rigid solids, Newtonian and Maxwellian fluids to viscoelastic layers and even multilayers and overlayers.29 According to the authors, the model permits the characterization of any combination of described physical conditions at the interface.These models indicate that a resonator can provide detailed information regarding the interaction between a resonator, a sensing layer and a medium. These can be changed by several factors: mass deposited on the resonator surface; energy dissipation by viscous action in the medium; energy storage by elastic deformation in the sensing film or the medium; and changes in the coupling between the surface and the medium. All of these process occur together, but in many systems one will dominate the others. 4 Measurement techniques The technique of choice for the monitoring acoustic wave device operation, in most studies, involves the oscillator method. Within such a configuration, the device constitutes the frequency-controlling element of a circuit.The oscillator method measures only one electrical parameter, the series resonant frequency of the resonating sensor. A typical oscillator circuit consists of two inverters connected in series producing a non-inverting amplifier. The sensor is connected from the output to the input of the amplifier to give a positive feedback.30 Mechanical damping of the generated acoustic wave results in a decrease in the amplitude of the received/reflected signal.At the point where admittance magnitude becomes too attenuated, as in a liquid medium where the viscosity exceeds a certain value, the oscillator circuit stops functioning.12 Alternatively, if an oscillator circuit is employed with automatic gain control, the amplitude of the output voltage of the amplifier is kept constant.11,25 In contrast to the oscillator technique, operation of acoustic devices in tandem with an impedance analyser involves the application of a sinusoidal voltage to the sensor resulting in a generated current which, by internal computing, provides admittance or impedance data.31,32 The method also permits the measurement of both series and parallel resonant frequencies together with corresponding energy dissipation factors.33 When the system is in series mode, the crystal oscillation is sensed by measuring the current that flows through the device, whereas when it is in parallel mode, the voltage over the crystal is detected.In both cases the decay curve is numerically fitted to the exponentially damped sinusoidal equation, yielding both resonant frequency and energy dissipation parameters. The application of the network analysis method, provides the most complete characterization of the electrical information Fig. 6 Mechanical analogy of a quartz resonator. Fig. 7 Extended Butterworth–van Dyke model. Analyst, 1999, 124, 1405–1420 1409obtained when an acoustic wave device operates in liquids.34 Values of the magnitude and phase of the impedance of the sensor are continuously calculated and registered. By fitting recorded data to the equivalent electrical circuit, the values of the circuit parameters can be calculated from the impedance– frequency curves. Among the quantities obtained from such an analysis are the series and parallel resonant frequencies, motional resistance and interfacial capacitance. The data obtained are related to the physical properties of the quartz, surface mass and contacting liquid, providing multiple information regarding the behaviour of the sensor in the liquid phase.23,35 The coupling of a flow injection analysis (FIA) system to an acoustic wave sensor [Fig. 8(a)]36 allows the monitoring of kinetic processes involved at the sensor surface–liquid interface. The design of a flow-through cell incorporating the sensing element varies with different experimental set-ups, but the essential requirement is always the attempt to generate a continuous laminar flow of liquid with minimized mechanical and electrical influence on sensor performance.As an approach to FIA-type analysis, a typical flow-through cell incorporating a TSM sensor [Fig. 8(b)]36 consists of two blocks clamped together to give a minimal volume of liquid in place over the device surface. Only one side of the sensor is exposed to the liquid, the other usually being kept under nitrogen. Generally, the sensing element is held between two O-rings, but in some designs it can be sealed or glued to the supporting holder.13 The fragility of quartz wafers makes the design of such a flowthrough cell a particularly difficult task in practical terms and significant further advances can be expected in the future. 5 Protein adsorption Research involving the study of protein adsorption at the solid– liquid interface has become an extremely important activity. A better understanding of such phenomena is essential for any investigation of crucial processes such as blood coagulation or solid-phase immunoassays, because the adsorption of proteins on solid surfaces appears to be the initial step in both events.The great number of related studies, with very often contradictory findings, makes this field even more scientifically intriguing. The complex structure of protein molecules and the existence of the considerable structural variety among the protein population, together with the essentially irreversible nature of certain adsorption processes, make the study of surface phenomena particularly complex and the description of appropriate kinetics very difficult to unravel. While traditional experimental techniques, predominantly involving radiolabelling protocols, provide important information concerning the amount of adsorbed protein, the general lack of adequate on-line techniques to study the kinetics of the process reliably has been widely recognized.Acoustic wave devices are among only a few of the available methods that can reach this goal, without the necessity for any modification of the protein molecule under study.13,37 A number of important studies dealing with basic protein adsorption to unmodified and chemically modified electrodes of TSM sensors have been published. Often, in many cases, the point has been made that frequency responses could not be correlated with the adsorbed mass of protein predicted by Sauerbrey’s equation.4 These findings indicate that for TSM devices operating in liquids, the mass response concept cannot be invoked to predict the amount of adsorbed protein. The change in frequency is probably governed not only by mass loading but also by boundary conditions at the sensor–liquid interface, such as interfacial free energy, liquid structure and coating film properties.The formation of a viscoelastic, highly hydrated protein layer at the interface produces changes in frequency response through viscous losses together with alteration of energy and dissipation processes.TSM devices coated with polysulfone were used to determine adsorption isotherms for albumin on a polymer surface.38 Massbased frequency responses were corrected by the introduction of a protein bulk concentration-dependent term to obtain a linear relationship. Higher values for the adsorption isotherms, obtained through the increased frequency shift, compared with predicted values, were attributed to the effect of hydration of adsorbed protein molecules.The adsorption of mellitin and b-globulin to phospholipid has been studied by the attachment of a lipid monolayer to the device surface by a Langmuir film balance.39 Frequency responses were measured both on-line from the solution and in air after drying, with the adsorbed mass of protein being evaluated from Sauerbrey’s equation. The discrepancy between the amount of adsorbed b-globulin calculated on the basis of the two different types of measurement was attributed to the incapacity of the large protein to penetrate into the lipid layer.As in the previous study, water hydration was assumed to be responsible for the increased frequency response. Acoustic network analysis was applied to characterize TSM device response for the adsorption of a-chymotrypsinogen on unmodified gold electrode surfaces.40 In addition to the series resonant frequency, other quantities such as parallel resonant frequency, motional resistance and static capacitance were also monitored.The shift in parallel resonant frequency, which is influenced by the composition of the charged interface, implied that the protein adsorbs initially in its native state, then undergoes later conformational changes that modify its surface charge characteristics. These findings were supported by the analysis of static capacitance response, which is primarily associated with the capacity of the electrical double layer at the charged interface.The importance of the size, shape and hydrophobicity of adsorbed protein has been associated with the enhancement of frequency response (in solution compared with air) for the adsorption of ferritin on the gold electrodes of piezoelectric crystals.41 On the basis of the amount of protein, calculated from Sauerbrey’s equation for the frequency measurement in air, the authors concluded that the protein adsorbed on the Fig. 8 (A) A single line manifold configuration for the FIA–TSM network analyser system. (B) Diagram of a flow-through cell incorporating a TSM acoustic wave device. 1410 Analyst, 1999, 124, 1405–1420electrode constituted less than monolayer coverage. The information extracted through protein film characterization by X-ray photoelectron spectroscopy, atomic force microscopy (AFM) and surface plasma resonance (SPR) spectroscopy supported the above finding. The analysis of the energy dissipation factor, rather than frequency response alone, was performed in a study where an enhanced sensor response was obtained in the case of the adsorption of viscoelastic, non-rigid layers of different biomolecules. 42 The experimental results showed that the adsorption of smaller proteins (myoglobin, haemoglobin, albumin) results in less dissipation shifts than for larger proteins such as ferritin or fibrinogen, after adsorption on hydrophilic gold electrodes or devices modified with a hydrophobic selfassembled thiol monolayer.The surveyed energy dissipation was defined by the co-existing action of three phenomena superimposed on the viscoelastic porous biofilm during the oscillation: the strain of the film, trapped liquid and the load from the bulk liquid. The authors used the same technique to study the adsorption of two forms of hemoglobin (met-Hb and HbCO) to the previously mentioned hydrophobic gold surface. 43 For both forms of protein, the results suggested the existence of two states of adsorption process exhibiting different viscoelastic properties. In the first phase the molecules are said to be more rigidly bound, forming a partially denatured protein layer, whereas in second phase an additional, more loosely bound layer of molecules is created, which is supposed to retain a native-like state.The adsorption kinetics of fibrinogen on the non-electrode surface of acoustic plate mode (APM) devices, sputtered with gold and modified with two differently terminated selfassembled monolayers (SAMs), has been examined.44 It was observed that while no protein adsorption occurred on hydrophilic hexa(ethylene glycol)-terminated SAM, large amounts of fibrinogen were adsorbed on hydrophobic methyl-terminated SAM which could be subsequently completely removed by surfactant.On the basis of these results, a two-state kinetic model was assumed to describe the adsorption process, where fibrinogen initially adsorbs on the surface in a native state and, subsequently, unfolds.The experimental results of another study indicated that the adsorption of fibrinogen on the gold electrodes of TSM devices modified with hydrophobic SAM resulted in an increased biosensor response compared with the case for a relatively hydrophilic gold surface.45 The same study presented evidence that the tertiary structure of avidin might have been compromised by adsorption on hydrophilic surfaces.The flow-through adsorption and characterization of the proteins human serum albumin, a-chymotripsinogen A, cytochrome c, fibrinogen, haemoglobin, immunoglobulin G and apo-transferrin to the two siloxane surfaces polymercaptopropylmethylsiloxane and octaphenylcyclotetrasiloxane and clean gold electrodes were monitored by acoustic network analysis.46 The nature of the responses for these unmodified and siloxanemodified TSM devices was discussed in terms of the various possibilities for interfacial molecular interactions, such as sulfur-containing moieties with gold, and side-chains with the phenyl rings present in octaphenylcyclotetrasiloxane.The overall direction of the responses, reduction in series resonance frequency and increase in motional resistance suggested the introduction of a viscoelastic protein film at the interface or perturbation of acoustic coupling to the liquid. TSM device measurements and data obtained by doublelayer capacitance were compared to study the kinetics of the adsorption of albumin on gold.47 It was concluded that sensor oscillations modified the adsorption kinetics by demanding an additional time constant to characterize the process.The authors considered that the biomolecular layer formed a stronger interaction with a non-oscillating than with an oscillating electrode and that, consequently, information concerning adsorption mechanisms obtained by TSM devices might be misleading.Adsorption and desorption of albumin and a thiol derivative on the gold electrode surfaces of TSMs have been monitored on-line in order to show the reversible adsorption of the former and irreversible adsorption of the latter.48 This very surprising result was used to support the findings obtained by application of an electrochemical ‘non-linear’ impedance method, where the measurement of non-linear electrochemical properties provided qualitative information on the adsorption state of the protein through the evaluation of the differential capacitance and conductance.The assumed physisorption of albumin resulted in loosely bound molecules and, consequently, a high voltage dependence of capacitance and conductance, while the same parameters in the case of the chemisorbed thiol derivative, tightly bound to surfaces, responded poorly to a polarity change associated with applied potential. In another study, the adsorption of albumin was monitored on the platinum electrodes of TSMs.49 The authors used Sauerbrey’s equation to estimate the mass loading of the protein, and suggested the existence of a monolayer of randomly oriented molecules.In contrast to a previous study, the adsorption of albumin was shown to be irreversible, based on experiments involving cyclic voltammetry. However, in situ stepping of the TSM electrode potential in the negative direction induced the minor desorption of albumin, which was attributed to an increase in electrostatic repulsion or decrease in electrostatic attraction within the protein molecule population, which is less tightly bound to the surface.The elutability of the milk proteins b-casein and blactoglobulin from clean gold and hydrophobic gold surfaces was studied by TSM devices.50 The experiments showed that proteins are harder to remove with surfactant from gold than from hydrophobic gold surfaces, possibly owing to the formation of covalent thiolate bonds between protein molecules and the gold surface.The enhanced frequency shifts that were observed in this study were attributed to changes in protein hydrodynamic layer thickness. An interesting attempt was made to develop an acoustic biosensor based on a thin (10 mm) film of a piezoelectric polymer, poly(vinylidene fluoride) (PVDF), which is sometimes also applied as a material for affinity membranes.51 An equation for mass sensitivity similar to that of Sauerbrey was derived and protein (IgG and BSA) adsorption was monitored in real time.Instability and numerous interferences were reported, with an enhanced Sauerbrey-like response being explained by involving the possibility of invoking multilayer protein adsorption and viscosity effects. Considerable research has been performed with respect to multilayer protein architecture, where TSM devices have been employed for monitoring the assemblage. Multilayers have been obtained by alternate adsorption of species including either opposite charges or specificity.Charged protein layers (cytochrome c, myoglobin, lysozyme, histone f3, haemoglobin, glucoamylase and glucose oxidase) were linked to positively charged polyethylenimine or negatively charged poly(styrene sulfonate) through alternate electrostatic adsorption.52 In this study, the assembly of up to 24 molecular layers of entire film thickness of 500 Å, incorporating adsorbed myoglobin and lysozyme, was deposited on the silver electrodes of 9 MHz piezoelectric crystals.The development of bioaffinity sensors, which are based on highly selective and sensitive binding interactions between a ligand and a biomolecule–receptor, has been the subject of many studies which will be discussed here and in the following sections of this review. The application of the capability of biological macromolecules for recognition of complementary substrate molecules and subsequent specific binding to them is the subject of a major research effort in the field of biosensor technology.However, many such systems have limited prac- Analyst, 1999, 124, 1405–1420 1411tical application because of factors generally connected with chemical and thermal instability. TSM devices with gold electrodes, modified with desthiobiotinylated albumin, were used to confirm the higher affinity for complementary coupling of avidin to biotin in solution rather than to the immobilized desthiobiotin.53 The high affinity and selectivity of the avidin–biotin interaction were applied for monitoring the in situ deposition of up to 20 protein layers on gold electrodes of 6 MHz piezoelectric crystals, involving streptavidin and biotinylated albumin.54 Frequency responses were measured in air, for dry protein assemblies as well, and were found to be in the accordance with Sauerbrey’s equation.The enhancement of frequency responses in situ was ascribed to the high hydration of protein layers.Considering the presumption that the sensitivity of TSM devices should decrease with increased viscoelastic coupling and after showing undiminished sensitivity of the device during the build-up of protein layers, the authors concluded that protein films behave as a rigid solid. In another study, 9 MHz TSM sensors with gold electrodes were applied for monitoring the formation of avidin–biotin conjugate multiple layer structures incorporating dextran, insulin and albumin.55 Although the build-up of multilayers was detected for more than a dozen avidin–biotin complexes, it was clearly observed that the resulting frequency shifts gradually diminished, which was explained by an increase in the thickness of the film which surpassed the decay length of shear wave motion.The specific binding of biomolecules to lipid receptors incorporated in supported membranes deposited on electrodes of TSM devices has been the subject of a number of studies.In some of this work, gold electrodes of 5 MHz TSM devices were functionalized with octanethiol and ganglioside-containing phospholipid vesicles were fused on the hydrophobic SAM.56,57 Impedance spectroscopy was used to confirm the complete immobilization of the bilyaer lipid membrane. The specific binding of various proteins, a lectin (peanut agglutinin)56 and bacterial toxins (tetanus, cholera and pertussis),57 to ganglioside receptors was detected by monitoring the series resonant frequency of TSM devices.On the basis of the experimental results, the authors concluded that the frequency shifts resulting from binding processes are specific parameters for particular ligand–receptor complexes. Discrepancies between measured frequency responses and those predicted by Sauerbrey’s equation for a complete protein monolayer were attributed to factors affecting the lipid–protein interface such as viscoelasticity, the electrochemical double layer and surface free energy.In another study, the binding affinity of a lectin for lactosylceramide was compared with the case where the glycolipid receptor monolayer was homogeneously mixed with matrix phospholipid to the situation when it was phaseseparated from a phospholipid monolayer, by means of a 27 MHz TSM horizontally attached to monolayers.58 On the basis of binding amounts calculated from frequency decreases, the authors concluded that the lectin shows higher affinity for the clustered glycolipid.TSM devices modified with photoactive materials have also been the subject of study. TSM devices (9 MHz) with gold electrodes modified with nitrospiropyran–pyridine photoisomerizable- mixed monolayers were applied to monitor the photostimulated association and dissociation of cytochrome c to and from a photoswitchable monolayer.59 Photoisomerization of the nitrospiropyran sites resulted in the protonated nitromerocyanine state which exerted electrical repulsive interactions with the positively charged cytochrome c associated with the pyridine sites.Consequently, the haemoprotein dissociated from the photoactive interface. It is important to mention that illumination cycles were actually performed outside the cell. As can be seen, in many studies the attempt has been made to associate experimental frequency responses with Sauerbrey’s equation. It appears that the prevailing explanation for the enhanced frequency response of TSM sensors in a liquid medium involves the role of loosely bound water molecules.It is likely that this type of association with proteins might increase their motional freedom and, therefore, energy losses through conformational changes in the structure. However, all the efforts made to emphasize the contribution of water molecules to the effective mass load appear to be completely inappropriate for a system where the protein film is surrounded by bulk water.By reducing the function of a TSM device to a microbalance, much readily available information on biomolecules at the TSM–liquid interface remains unappreciated. In reality, the complex response of a TSM sensor provides valuable qualitative data about the nature of the adsorbed protein layer and the functional consequences of the adsorption process. 6 Detection of interfacial immunochemical interactions Among the most important biological systems composed of molecules with specific binding properties are antibodies– immunogens (antigens), enzymes–substrates, lectins–oligosaccarides, receptors–various molecules (hormones, neurotransmitters, cytokines), avidin–biotin, protein A–antibodies or antibody–antigen complexes, DNA–DNA or RNA, transport proteins–various molecules and ions.In addition, the design of synthetic recognitive elements with particular binding specifities is the subject of much current research. Some of abovementioned molecular recognition processes have already been discussed in the previous section.A molecular interaction for an affinity sensor can be described by the equation A B AB + [| k k 2 1 where A is the molecule, B a substance with binding properties towards A, AB the complex between them, k1 rate of association and k2 rate of dissociation. The binding constant (K) represents the affinity between the two elements: K = [AB]/([A][B]) = k2/k1 An antigen can induce a polyclonal response, i.e., the production of a large number of antibodies with different binding constants.13,60–62 Compared with other applications of acoustic wave devices as biosensors, far more work has been carried out with respect to the development of immunosensors. In this section we shall discuss studies where various immunochemical reactions have been performed at the acoustic sensor–liquid interface.The earliest attempt to develop an immunosensor based on a TSM device was in 1972 when 5 MHz piezoelectric crystals were used to detect the formation of a BSA–anti-BSA sera complex by the ‘dip and dry’ method.63 Such a procedure consisted of dipping piezoelectric crystals, pre-coated with antibody/antigen, into a solution of antigen/antibody protein with subsequent measurement of the frequency of the dried sensor for comparison to the frequency obtained before immunoreaction. This type of approach has been adopted in a surprising number of later studies. For example, in a recent investigation, different human herpes viruses were detected with a reported excellent correlation between frequency change and the number of viral cells.64 Synthetic peptides have been used to generate herpes virus-specific monoclonal antibodies, which were, in turn, immobilized via protein A on modified gold electrodes of TSMs.The ‘dip and dry’ method was also used for the characterization of an atrazine immunosensor based 1412 Analyst, 1999, 124, 1405–1420on protein A-modified TSM with atrazine antibody immobilised on the device surface65 and for the detection of a microbe (Vibrio cholerae) by a TSM immunosensor.66 An assay with excellent precision for the detection of atrazine has been reported, in contrast with the observations of many researchers who have experienced great difficulty in obtaining reproducible frequency measurements in air after the devices have been submitted to any significant manipulation and (bio)chemical transformation.Experience shows that conclusions based on any relatively small, off-line, frequency changes in air related to the mass sensitivity of TSMs can only be regarded as, at best, highly speculative. Another system for the detection of the same herbicide was proposed later when the specific reaction between atrazine immobilized on modified TSMs and its monoclonal antibody was monitored in real time. 67 In contrast to the above, the complexity of the physics of propagation of acoustic shear waves through the TSM–bulk liquid interface has been recognized for some time.Silanized TSM devices and those modified with polyacrylamide gel were used to monitor the formation of the complex between the immobilized IgG antigen and the IgG from the solution in real time.11 In this study it was clearly pointed out, for the first time, that TSM systems with viscoelastic proteinaceous layers at the interface do not respond to absolute amount of added material associated with Sauerbrey-type mass frequency physics, but rather to perturbations in interfacial conditions.Similar immunoreactions were monitored on the surface of TSMs precoated with protein A, which specifically binds human IgG.68 In another study, silanized TSMs were used to monitor the preferential binding affinity of IgG subclasses to immobilized protein A.69 In both studies, unsuccessful attempts were made to correlate the amount of theoretically bound immunoreactant with the assumed mass sensitivity of the device.In a later study, where the interaction between anti-HSA and HSA on the surface of TSMs was observed, the enhanced frequency response was positively established by radiolabelling and attributed to interfacial factors.70 Recently, the binding of HSA to complementary monoclonal and polyclonal antibody precoated on TSMs was examined and compared.71 Assuming that the change in resonant frequency varies linearly with the mass of adsorbed protein, the authors were able to estimate association constants and rate constants for the immunoreactions.The detection of HIV antibodies by means of synthetic HIV peptide (the epitope of the protein) immobilized on 20 MHz TSM was performed with a selectivity comparable to that of a licensed ELISA-based test.72 The authors claimed that by application of a differential FIA method, which eliminated the damping effect, the validity of Sauerbrey’s mass–frequency dependence was assured.Other authors applied a dual modulation method that employed a sealed crystal providing frequency difference relative to the working crystal to overcome damping disturbances.73 A specific reaction between an enterotoxin and its antibody immobilized on modified electrodes of TSMs was monitored by this procedure. The enhancement of the frequency signal has been recognized in many immunoreaction-related studies, and various attempts to elucidate the observed phenomenon have been made.Many of the efforts have concentrated on the distinguishing of effective mass load and the presence of a viscoelastic biomolecular layer on the surface of the device. The oscillator frequency and the electroacoustic admittance were alternately measured during the selective reaction between a peroxidase and the corresponding antibody immobilized on a 27 MHz transducer, in order to compare the on-line change of parameters characterizing TSM equivalent circuit with the oscillator frequency which was supposed to depict only a mass variation. 74 The adsorption of the viscoelastic protein layer resulted in an increase in motional resistance and an alteration in rheological properties that produced an overestimation of the mass given by the oscillator frequency. However, the lack of the motional resistance increase during the immunoreaction itself was explained by the inadequacy of the performed immunoassay. In another study, where the acoustic network analyser method was used to monitor immunochemical interactions, the role of boundary viscoelastic properties of the surface of a TSM was examined through a shear acoustic impedance signal which was connected to rigid mass and interfacial liquid variation.75 Two types of immunoreactions were discussed.In the case of selective binding of erythrocytes to IgM, changes of boundary liquid viscosity dominated, whereas in the case of the formation of an IgG–anti-IgG complex, mass deposition appeared to be the controlling process. The simultaneous analysis of both frequency and dissipation factor was performed during HSA–anti-HSA immunoreaction on the surface of hydrophobic gold electrodes of TSMs.76 The enhanced frequency response and the increase in dissipation were attributed to the contribution made by the trapped interprotein capillary-like water to bare protein mass or to the non-specific binding in the second phase of the interaction.An entirely different experimental procedure was applied when the covalent coupling of IgG antibody fragments to linkers embedded in a phospholipid monolayer matrix and subsequent binding of the antigen were monitored at the air–water interface by means of a TSM device in the contact with a bioactive layer.77 The authors observed that a substantially larger surface density of bound biomolecules was determined on the basis of measured frequency response than by radioimmunoassay.Similarly to the study described earlier, the results were explained by hydration and/or slippage in the protein layer. As can be gleaned from the above analysis of different studies, the proteins that take part in immunoreactions have been immobilized on acoustic wave devices following different immobilization methods. While it is possible simply to adsorb the protein on the gold electrodes of TSMs through hydrophobic and sulfur–gold interactions,66,70–72,74 in other investigations gold TSM electrodes were modified with polyethylenimine via glutaraldehyde,64,65,73 aminosilane via glutaraldehyde, 11,64,65,67,69,75 polyacrylamide via glutaraldehyde,11 hydrophobic thiol SAMs76 or by protein A68 in order to attach further the active biolayer by hydrophobic or covalent bonds.In some of these studies, various immobilization procedures were evaluated.11,64,65 Recently, different immobilization methods, i.e., adsorption on gold, avidin–biotin binding and two types of bindings on thiol SAMs [dithiobis(succinimidyl propionate) and dextran-modified], were compared regarding BSA–anti- BSA and IgG–anti-IgG immunoreactions.78 Various specific interactions on TSM surfaces have been monitored in real time, which raises the interesting point that none of the suggested immobilization protocols was recommended as possessing any clear advantage over any other protocol.A similar conclusion might be derived from a study where specific interaction between IgG and anti-IgG was observed on the surfaces of unmodified TSMs and those modified following three different protocols: protein A binding, thiolation of gold electrodes and thiolation of anti-IgG.79 The estimation of adsorbed amounts of proteins was performed by the ‘dip and dry’ method, although the monitoring of the immunoreactions in real time was also presented, in order to evaluate orientational aspects and the immunological activity of immobilized proteins.A similar immunochemical interaction was observed in a related study where alternating polyelectrolyte films and anti-IgG multilayer films were used for the detection of IgG.80 Many other experimental procedures for immunoassays on TSM surfaces comprise various techniques for modification of gold electrodes. SAMs of a thiol-containing synthetic peptide constituting an epitope of the foot-and-mouth disease virus were covalently attached to gold electrodes and applied to the detection of different concentrations of the specific antibody in Analyst, 1999, 124, 1405–1420 1413the solution.32 It was further reported that Chlamydia trachomatis from urine samples might be detected on the surface of gold TSM electrodes primarily modified by a cystamine monolayer and linked to the corresponding antibody.81 In another investigation, by use of the Langmuir–Blodget technique, fluorescein lipids embedded in a phospholipid matrix were deposited on gold TSM electrodes and the binding of antifluorescyl antibody to fluorescein-conjugated lipid was monitored.82 Also, there have been attempts to modify gold electrodes of TSM devices by anodic polymerization of m-phenylenediamine in order to prepare them for the immobilization of bioactive molecules.83 The selective interaction of a microbe (S. paratyphi A) against its antibody, cross-linked within the electropolymerized film, was monitored.Somewhat earlier, the immobilization of antibodies on TSM surfaces modified by ethylenediamine plasma-polymerized films was examined and evaluated.84 The sensor was calibrated, in liquid, for the specific interaction between HSA and its antibody.The limited mass sensitivity of commercial TSM devices, which is of primary concern for many investigators, is imposed by manipulating difficulties with extremely thin quartz discs operating at higher resonant frequencies. Consequently, there have been a number of attempts to overcome this problem.A unique approach, described as an amplified-mass immunosorbent assay, was used for the detection of an enzyme (adenosine 5A-phosphosulfate reductase).85 Gold electrodes of TSMs were modified with the corresponding antibody and after the formation of the immuno-complex, which did not result in a significant frequency decrease in the applied range of concentrations, another antibody–enzyme conjugate reagent was employed to catalytically amplify the mass response.A histochemical staining agent, that was added to the solution, was enzymatically converted into an insoluble precipitate on the TSM surface, inducing a decrease in frequency with a magnitude corresponding to the level of enzyme bound to the antibody-modified TSM. The intensification of the signal by the enzymatically enhanced formation of the deposits, for another immunoreaction on the surface of TSM, was obtained using the amplified mass immunosorbent assay for detection of antibodies against the recombinant African swine fever virus attachment protein.86 Liposomes have been used as a different kind of signal-intensifying reagents for monitoring an immunoreaction on the surface of a TSM device in a type of competitive assay.87 The binding of antibody-bearing liposomes to the corresponding antigen (phenyloxazolone–BSA conjugate) immobilized on gold TSM electrodes has been examined in real time.The specific interaction was inhibited by the presence of the soluble hapten in a concentration-dependent manner.Different immunochemical binding processes have been monitored on the surfaces of other types of acoustic wave device. A surface transverse wave (STW) device operating at 250 MHz, silanized and modified with avidin–biotinylated atrazine complex, was used for the detection of atrazine. 88 Applying a type of a competitive assay, very low concentrations of the herbicide (0.06 ppb) were detected.The possibility of improving the sensitivity of Love wave devices by optimizing the thickness of the waveguiding layer has made them attractive for immunosensing applications. Low (ppb), levels of anti-IgG were detected using a Love wave sensor operating at 110 MHz, which was silanized and modified with IgG.89 It was reported that a Love wave device showed an improved performance in liquids when a thin gold film was deposited on top of the polymer waveguide layer to eliminate unwanted acoustoelectric coupling.90 In a later study, such a system was fully applied as an immunosensor.91 A biotinylated, supported phospholipid bilayer that bound streptavidin and biotinylated goat IgG was used to detect anti-goat IgG.An excellent suppression of non-specific binding, provided by streptavidin attachment to the phospholipid layer, was achieved. In a recent study, the viscoelastic behaviour of protein films, composed of IgG antibodies deposited on a shear horizontal surface acoustic wave device, was examined.92 The attenuation of Love waves in a film composed of up to 20 layers was discussed.The authors concluded that for the one- or two-layer films normally used in acoustic wave immunosensing, a pure mass effect is dominant, whereas for thick films, saturation of the sensor response was observed. Acoustic plate mode (APM) devices have also been applied to the detection of low levels of IgG in a static experimental cell.93 The IgG antigen–antibody system was examined in another study involving an APM device.94 A detailed investigation of the effect of viscoelasticity of the films of different reactants used to modify sensor surfaces prior to the immobilization of antibodies was performed.The authors concluded that the optimization of the device response would require minimizing viscoelastic effects of the coatings. A careful examination of the studies described above suggests that despite the existence of a significant amount of work devoted to the development of immunosensors based on acoustic wave devices, their full development is still problematic. Having in mind that antigenic sites are often conformation dependent and that hydrophobic, hydrophilic and ionic interactions are all involved in the binding process, it is understandable that the immobilization and adsorption of proteins on the surfaces compromises their activity in affinity based reactions, thus making the design of immunosensors particularly troublesome.Another generally acknowledged source of difficulty is misleading results related to non-specific adsorption of the species on the sensing surface. It appears that a more extensive analysis of these obvious difficulties is often conveniently neglected during the presentation of experimental results. 7 Nucleic acids and DNA/RNA–protein/peptide interactions. Drug discovery Recently, in an issue of Time magazine that was entirely devoted to the development of genetic engineering, the 21st century was proclaimed as the century of biotechnology.95 Indeed, having in mind the accelerated work and intensified efforts with respect to deciphering the human genome, it is realistic to assume that the coming decades will bring us perhaps the most exciting turning points in the history of medicine, through the full implementation of genetic profiling and genetically customized drugs.Much of the general availability of the above methods will depend on the further development of gene probe technology.Currently, radioactive, enzymatic or luminescence labelling are applied to obtain detectable signals during the hybridization of nucleic acids and for the determination of specific DNA sequences.96,97 Constant efforts have been made to develop simple, label-free methods to monitor not only hybridization processes, but also nucleic acid– protein interactions and the binding of small molecules and drugs. The compatibility of most biosensor technologies with microelectronics makes genosensors promising tools for fast and continuous screening of the above-mentioned processes.In this respect, research into genosensors based on acoustic wave devices presents exciting possibilities for the future. Most of the relevant work involves the monitoring of the hybridization of single-stranded nucleic acid molecules, immobilized on TSM surfaces by various techniques, to their complementary strands present in solution.In one of the earlier attempts, the hybridization of synthetic RNAs was identified on the surfaces of a polymer (styrene–acrylic acid copolymer)- coated 9 MHz PQCs using the often-mentioned ‘dip and dry’ method.98 Ebersole et al. used the procedure for amplifying TSM signals85 to demonstrate the potential of TSMs modified with avidin and streptavidin for performing hybridization 1414 Analyst, 1999, 124, 1405–1420assays with biotinylated oligonucleotides.99 The application of TSM surface modification by the formation of avidin–biotinylated DNA complexes was exploited in the number of hybridization studies.100–103 Multilayer biotinylated DNA films formed on 9 MHz TSM surfaces by the successive deposition of avidin and poly(styrene sulfonate) showed increased hybridization capacity compared with single-layer films.100 Another approach for building multilayer structures including DNA molecules on the surfaces of TSMs consisted of units built of double-stranded biotinylated DNA sandwiched between streptavidin layers.101 A biotinylated lipid matrix was primarily deposited on TSM electrodes.Neutravidin-modified TSM surfaces were also used to study the hybridization of complementary, non-complementary oligonucleotides and targets with a mutated base with a biotinylated 25-mer oligonucleotide probe.102 It was shown that distinctively different signals were obtained for the complementary and non-complementary cases and for different types and locations of induced mismatches.In a further study with 27 MHz TSM, involving an avidin–biotin nucleotide probe immobilization method, the comparison of the results obtained with those acquired by SPR (commercial system, BIAcore) was performed.103 The monitored hybridization process included alterations in the reacting oligonucleotide molecules, such as integration of different mismatched bases and changes in DNA chain length.Other methods to immobilize DNA probes on the gold electrodes of TSMs, have included thiol-derivatized nucleic acid molecules at the 5A-phosphate end.104–106 It was also shown, using the genoprobe obtained through the immobilization of a thiol-derivatized peptide nucleic acid on gold TSM electrodes, that it was possible to differentiate single-base mismatches during the hybridization event.107 In some studies, silver TSM electrodes were applied for the immobilization of TSMs through modification with thioglycollic acid108 or with a didodecyldithionooxamide–BSA complex.109 Other approaches to immobilize nucleic acids on TSMs for monitoring hybridization processes have been exploited.A DNA probe, where a nucleic acid was covalently immobilized on gold TSM electrodes hydroxylated with (3-glycidoxypropyl)trimethoxysilane, was reported earlier.110 In a unique attempt at adsorption and electropolymeric (polyphenol) entrapment for immobilizing a four-layer DNA dendrimer on gold TSM electrodes, trioxsalen (4,5A,8-trimethylpsoralen) was used to cross-link covalently dendrimer layers.111 The branched DNA structure consisting of multiple singlestranded arms showed significantly improved hybridization capacity compared with that for conventional DNA probes.In several earlier studies, the hybridization of nucleic acids adsorbed on PdO sputtered on gold TSM electrodes was investigated.112–117 Applying radiochemical labelling with 32P, the authors showed that the added mass resulting from the formation of a DNA hybridization complex corresponded to a considerably larger change of frequency than that predicted by Sauerbrey’s equation.113 The observation of an enhanced frequency response was interpreted in terms of the perturbation of acoustic energy transmission by changes in interfacial properties upon hybridization.By means of acoustic network analysis, DNA hybridization kinetics at the TSM–solution interface were examined in detail.114 The kinetics of hybridization of RNA homopolymer was also discussed and a hybridization mechanism was proposed after the examination of time profiles for the parameters measured by the network analyser such as series and parallel resonant frequency, motional resistance or maximum phase angle.117 The monitoring of the binding of proteins to nucleic acids immobilized on TSM surfaces has been the subject of a number of studies.The kinetics for sequence-specific binding of NFI (nuclear factor I) to DNA, including its recognition site, were monitored and compared with non-specific binding of histone f3.118 For another specific protein–DNA interaction, the authors compared the kinetic parameters calculated on the basis of data obtained by TSM measurements with those found by gel mobility shift assays and demonstrated a good agreement between the two techniques.119 A recent study involved the detection of the interactions between two Tat-derived peptides and HIV-1 trans-activation (TAR) RNA immobilized on TSM surfaces.120 The frequency responses related to two different binding processes showed opposite directions, reflecting a structural distinction between the two Tat–TAR complexes at the TSM interface.Small molecule binding to nucleic acids is very important for research efforts related to drug discovery. However, only a few studies with acoustic wave sensors have dealt with the monitoring of such processes.These mainly involve the kinetics of the binding of chemotherapeutic agents to DNA. In a study of the DNA–platinum-based drug interaction mechanism, both series frequency and motional resistance signals obtained from acoustic network analysis were monitored.121 A kinetic analysis showed the detection of DNA complexation with hydrolysis products of two platin isomers (cis and trans). In another study, the alkylation of DNA by different antitumour agents was monitored by TSM, through the change in viscosity of DNA solution induced by the alkylation process.122 The described method, unique because of the omission of a DNA immobilization step, could not, however, provide a more detailed kinetic analysis.Recently, a ruthenium complex binding to DNA immobilized on gold TSM electrodes through a cationic (methylated thiopyridine) monolayer was studied using cyclic voltammetry and an acoustic wave sensor.123 The binding properties of an antibiotic (doxorubicin) to DNA immobilized on TSMs were also studied.124 This investigation showed that, despite the differences in applied DNA immobilization procedures and sample injection methods, the affinity constants for ligand–DNA interactions determined by the acoustic wave sensor and by a commercial SPR device (BIAcore) were comparable.Regarding other types of acoustic wave sensor, work connected with the development of a gene probe has been performed using acoustic plate mode (APM) devices.In an earlier study, where a probe nucleic acid sequence was covalently attached to the sensing surface through an aminosilane film, the hybridization reaction was monitored.125 Depending on the concentration of the target nucleic acid, both elastic stiffening of the interface, due to the possibility of target DNA cross-linking the surface, and mass loading were observed. Later, the same authors demonstrated the sensing properties of an improved APM device design for the selective detection of chemically denatured DNA by means of polymerase chain reaction amplification.126 Finally, considering the fact that acoustic sensor technology might be conveniently incorporated into the design of highdensity arrays of genoprobes, widely known as gene chips, it can be expected that work in this field will intensify. 8 Cell adhesion and cell function The monitoring of cell adhesion, spreading and proliferation on solid surfaces is crucial for a better understanding and evaluation of the capability of biomaterials for assimilation within natural tissue and the activation of tissue repair mechanisms. However, the investigation of these phenomena by means of acoustic wave devices has been the subject of only a few studies, in spite of the obvious feasibility and the attractiveness that the method provides for investigating similar processes.In some earlier studies, the attachment of platelets,127 osteoblasts (bone-forming cells)128 and epithelial (monkey kidney) cells129 to gold electrodes of piezoelectric crystals was Analyst, 1999, 124, 1405–1420 1415monitored. In each study the amount of the adhered cells derived from a mass-based response could not be correlated with experimentally determined values, although different explanations were offered.Whereas in the former study127 it was supposed that the TSM sensor did not detect the total mass of whole bodies of adherent platelets, but only the weight of their focal contact region, the latter studies128,129 assumed that viscoelastic damping throughout the viscous layer of adhered cells was responsible for the decrease of the theoretically expected response.However, in the study with osteoblasts,128 it was shown that a linear correlation between frequency change and the surface coverage could be established and possibly applied to monitor cell growth.In the same study, a previous model22 for a purely viscous film was used to determine the apparent viscosity of osteoblasts. Similarly, the viscosities of epithelial (canine kidney) cell monolayers cultured on gold electrodes of TSM were determined by impedance analysis.130 The authors used an equivalent circuit model23 to quantify the viscous damping. The low values obtained for apparent viscosities were attributed to the existence of additional space between the cells immobilized through their focal contact regions and the TSM surface, where the shear wave might be considerably damped before contacting the cellular monolayers. Data obtained from simultaneously measured changes in frequency and energy dissipation during the adhesion of small colonies of epithelial (monkey kidney and hamster ovarian) cells to TSM surfaces modified with hydrophobic and hydrophilic polystyrene surfaces were used to evaluate the potential application of TSM for studying cell–surface interactions.131 The authors concluded that the TSM response provides a ‘fingerprint’ of the cell adhesion process reflecting different cell adsorption kinetics for various cell types and various surfaces. Experimental results suggesting that protein adsorption is the initial event during blood interaction with surfaces goes back to earlier studies.132 Further, it was shown that proteins such as fibrinogen, fibronectin or von Willebrand factor strongly promote platelet adhesion and spreading, whereas the same process is inhibited by pre-adsorbed albumin.133 Unfortunately, there are few TSM studies dealing with the problem of cell adhesion on a pre-adsorbed protein layer or any biospecific layer.The strong adhesion of bovine epithelial cells on TSM surfaces pre-coated with fibronectin was observed and confirmed using other techniques such as electron microscopy, interference reflection microscopy and fluorescence microscopy. 134 In another study it was shown that TSM devices with gold electrodes modified with collagen respond substantially more to the adhesion of platelets than corresponding surfaces modified with albumin.135 The smaller frequency response compared with a theoretically expected mass based response derived from the amount of platelets counted by radiolabelling was not a surprising finding considering the high water content of platelets. The sensor response was attributed to viscoelastic properties and morphology of the adhered platelet layer rather than to their mass.Finally, it was found that various tumour cells adhered and spread on galactose-bearing lipid film immobilized on the surface of TSM in serum-containing medium, but not in serum-free medium.136 The results suggested that serum components pre-adsorb on synthetic glycolipid films and further influence cell spreading. The necessity for the evaluation of the haemocompatibility of biomaterials and a better understanding of the processes controlling thrombus formation on solid surfaces in contact with blood suggests the importance of the described TSM studies and further investigation of relevant phenomena.It would be interesting to monitor the processes related to ‘multilayer protein passivation’ which has frequently been suggested for attaining long-term antithrombogeneity of implanted biomedical devices.137–139 Further studies involving monitoring of interactions of blood clotting factors with cell membranes would be extremely helpful for extending our knowledge of processes involved in the blood coagulation system and, consequently, the short-term blood compatibility of biomaterials. 9 Other applications In several studies, the dependence of the TSM frequency response on the interfacial viscosity was used to monitor various biochemical processes. The changing viscosity of the pre-adsorbed fibrin clot, as it dissolved after the addition of plasminogen activator, was monitored by a TSM device.140 The clot dissolution time was correlated with the amounts of two types of plasminogen activators in the examined samples.Whereas in the case of streptokinase the observed standard error was comparable to that exhibited in rapid fibrin plate assay, in the case of tissue plasminogen activator a much higher standard error was calculated. The discrepancy was attributed to diffusion effects and to acoustic compression wave resonances.In another study, the rheological behaviour of a blood drop deposited on the surface of silver TSM electrodes was investigated.141 Equivalent circuit parameters of the TSM were monitored with an impedance analyser. The authors examined processes such as blood clotting and urokinase activated fibrinolysis and concluded that the frequency response was dominated by a viscosity change only at the initial stage, followed subsequently by a mass effect and surface stress. A TSM sensor combined with a spectrum analyser was used to monitor the viscosity decrease during the depolymerization of a muscle protein, actomyosin, induced by the addition of adenosine 5A-triphosphate in its solution.142 On the basis of the experimental results, the authors suggested the further development of a sensor for monitoring the freshness of meat.In another study, the same authors applied the TSM device to monitor the viscosity change of the solution due to an immunoreaction in association with the specific agglutination of streptolysin-bearing latex particles induced by the introduction of lysin.143 In an early investigation, the resonance frequency and resistance of a palladium-plated TSM sensor were monitored with an impedance analyser during the gelation of limulus amoebocyte lysate induced by E.coli produced endotoxin.144 The monitoring of increased viscosity by a TSM sensor has been claimed to be superior in performance to a conventional optical method used for the determination of endotoxin.The growth of urease-producing bacteria, through detection of the dissolved ammonia in the culture medium using TSM devices, has also been reported.145 Ammonia, produced by the hydrolysis of urea, diffused across a gas-permeable membrane into an internal electrolyte solution, where conductivity alteration caused a change in the resonant frequency of the TSM sensor which was actually not in contact with the solution.The growth characteristics of Proteus vulgaris were investigated and the method apparently exhibited satisfactory accuracy and precision. Attempts have been made to develop a bioaffinity sensor for monosaccharides (glucose and ribose) based on a 6 MHz thiolmodified TSM devices coated with a synthetic receptor resorcinol–dodecanal cyclotetramer.146 The authors reported that no meaningful signal was obtained owing to the binding of low molecular mass substrates to the sensitive layer composed of a low surface concentration of receptor molecules.However, the binding of a disaccharide (sucrose) to a fully hydrated (thermally treated) Langmuir phospholipid film transferred onto the electrodes of 5 MHz TSM was possible, although not in situ.147 On the basis of frequency measurement of the dried 1416 Analyst, 1999, 124, 1405–1420substrate, the authors concluded that the amount of bound sucrose increases linearly with its concentration in the solution.The results were explained by the formation of a mingled layer with inseparably incorporated sucrose molecules within a hydrated phospholipid film. The adsorption behaviour of phospholipid containing liposomes on synthetic phospholipid polymers and poly[hydroxyethyl methacrylate] (HEMA) was monitored and compared, using 5 MHz TSM devices with gold electrodes spin-coated with the above mentioned polymers.148 On the basis of larger frequency responses, the authors concluded that liposomes can penetrate a hydrated poly(HEMA), and, thus, probably change their structure.The unchanged structure of liposomes adsorbed on phospholipids was, however, confirmed by AFM. It was emphasized that such findings might enhance the application of otherwise unstable liposomes, immobilized on phospholipid polymers, as model cell membranes for studying different interactions and processes. In another study, gold plated TSM devices with C- and Nterminated l-alanine immobilized on the surface were used to monitor artificial boundary lipid-containing liposome-induced release of membrane proteins and the subsequent reconstitution. 149 The authors suggested that different shapes of time– frequency responses obtained during the binding of the proteoliposome to two differently terminated l-alanine forms might make TSM devices a useful instrument for distinguishing conformational changes in amino acids. TSM devices were coated with cholestyramine resin and exposed to solutions of various bile acid salts (cholate, glycocholate, taurocholate) in a batch-operating one-side liquid cell.150 The authors reported apparently good sensitivity of the sensor obtained in such a way and suggested a multi-step regeneration protocol permitting the resin-based sensor to be rinsed numerous times over extended period.Such a sensor has been suggested as an important tool in the study of bile salts binding non-absorbable resins as a therapy for the elevated serum cholesterol.An interesting study of the inhibition of catalytic activity of a haem peptide, microperoxidase from cytochrome c by binding of strong ligands, was performed on the gold electrodes of 5 MHz TSM devices modified with a mercaptosilane and fabricated as a dual-response biosensor, giving simultaneous electrochemical and piezoelectric responses.151 The inhibition of the catalytic activity upon binding of ligands (imidazole, histamine and histidine) was observed through electrochemical response as a decrease in the cathodic current response to H2O2.The mass increase due to the binding of the same ligands was detected through the decrease in frequency response. Such a configuration apparently permitted both qualitative and quantitative analysis of interfering ligands present in the solution. It is interesting that the authors claimed, although supporting data were not shown, that by the mass-related decrease in frequency response it was possible to distinguish the binding of histamine (Mr = 111.15) from the binding of histidine (Mr = 155.16) to the peptide.Here we shall mention some attempts to develop enzyme based biosensors using acoustic wave devices. It was reported that glucose could be detected continuously by using the enzyme-catalysed formation of precipitates of an oxidised dye on the TSM surface.152 Similarly to previous studies,85,86 an enzyme-related reaction resulting in insoluble species was used to intensify the impedance response.In another study, the rate and extent of the inhibitory effect of organophosphorus and carbamate pesticides on the activity of acetylcholinesterase immobilised on the surface of TSMs were followed in real time by measuring the frequency changes associated with catalytically induced deposition of a pigment.153 However, the application of these methods may be problematic because of the simultaneous non-specific adsorption of other components present in the system. 10 Concluding remarks and future perspectives The literature associated with the operation of the TSM in liquid media offers a classic example of the nature of scientific dogma. When it comes to the interpretation of frequency responses for a wide variety of chemical systems instigated at the TSM device–liquid interface, the wavelength extension model has almost invariably been invoked. (Often, because the oscillator experimental configuration is employed, the frequency response is not the true series resonance frequency.) Although facile explanations based on a mass response model are convenient, particularly for situations where it is important to correlate other processes with deposited (or lost) amounts, it is unreasonable to ignore the fact that in liquids acoustic energy is propagated into the surrounding medium to decay lengths of the order of micrometres.Moreover, chemistry conducted at the TSM surface is occurring at precisely the location where acoustic coupling phenomena occur.Accordingly, the perturbation of the transmission of acoustic energy by biochemical pairs combined at the device–liquid interface provides an exciting biosensor mechanism. The reason for this lies not only in the fact that interfacial biochemical pairs often involve changes in physical properties connected to tertiary structural and counterion alterations, but also to the possibility to obtain multidimensioned information from the TSM.For example, the latter yields mass, charge and energy dissipation responses through consideration of inductance, parallel resonance frequency and motional resistance changes, respectively. Electrochemical and optical-based biosensors do not provide this range of versatility. Finally, it is worth noting that the sensitivity of acoustic wave sensors has generally been considered to be inferior to that of the above-mentioned devices.In reality, because of the sensing mechanisms outlined previously, the various acoustic wave structures offer unique sensitivity to interfacial biochemical phenomena. A brief summary of some of the features provided by acoustic wave sensor technology follows: 1. Basic and equivalent circuit parameters offer the potential to study the tertiary structure, charge, viscoelastic properties, etc., of biochemical macromolecules and cells attached to a device surface immersed in water. 2. Acoustic wave sensors yield responses to various properties of a wide variety of biological moieties such as proteins, nucleic acids and cells. 3. Signalling is direct in that no labels such as radiochemical or fluorescent tagging agents are required. 4. Most acoustic wave devices are functionally compatible with FIA technology. 5. The behaviour of devices in liquids is compatible with a number of different protocols available for attaching biochemical moieties to the sensor surface. 6.It is clearly possible to work with immobilized biochemical monolayers rather than the situation for other sensors where adjacent ‘amplifying’ layers such as dextran are required. 7. The detection of the binding of relatively small molecules to various receptor sites imposed on the device surface appears to be feasible, especially if such a binding event results in a change of receptor tertiary structure. 8. It is apparent that new acoustic wave structures such as the tuneable MARS device may offer not only enhanced sensitivity, but also new dimensions for the study of biochemical macromolecules in micro-assay format.We now turn to a brief look at what is on the horizon in the development of acoustic wave biosensor technology. One research activity that is common to all biosensors is the immobilization of various biological species on the device surface. With respect to acoustic wave devices this may, for example, involve gold surfaces such as those in place on TSM structures, or silicon in other devices.A key problem for all Analyst, 1999, 124, 1405–1420 1417sensors has been the stability of biochemical macromolecules attached to device surfaces, from the perspective of storage, heat and light effects. If a device is to be employed over relatively long periods of time, without calibration, denaturation of proteins, for example, cannot be tolerated. In this regard, we are beginning to see exciting new methods for the enhancement of macromolecular stability through the use of sol–gel and polymer-based systems.In terms of acoustic wave devices, we are likely to see considerably more research being directed at an understanding of the interactions between acoustic energy and the properties of the liquid–solid interface. This is likely to be reflected in attempts to dissect out from sensor responses, mass, charge, viscoelastic and acoustic coupling processes, particularly with respect to frequency and equivalent circuit parameters.An exciting possible result of such understanding will be the generation of new ways of studying biochemical phenomena such as tertiary structural changes and counter-ion release. In this respect, it is interesting that the link between acoustic wave responses and biochemical structure may require a new detailed understanding of the theory of acoustic wave propagation into liquids. This is so because explicit correlation with other techniques may be unavailable because of the dearth of optical and other methods for the detection of subtle biochemical structural alterations.The comment given above leads one to conclude naturally that we are likely to see future significant advances in the development of true hybrid sensors, including those involving acoustic wave devices. This type of structure will incorporate a substrate together with its immobilized biochemical apparatus being interrogated by completely different physical techniques.Possibilities include opto-acoustic devices and the use of materials that are both pyroelectric and piezoelectric. Such combinations in tandem with chemometric procedures may lead to enhanced analyses in terms of the information generated by sensor response. Finally, as we mentioned above, we will see the introduction of new acoustic wave sensors in the near future. Such structures will involve better sensitivity, facile attachment of biochemical macromolecules and cells, use of non-piezoelectric materials, the generation of several levels of chemical information from physical parameters and commercial presentation of equipment incorporating a sensor into sophisticated FIA set-ups. 11 Acknowledgements We are grateful to the Natural Sciences and Engineering Research Council of Canada and Dow-Corning, Midland, MI, USA for the support of this work. 12 References 1 D. Reanney, After Death, Morrow, New York, 1991. 2 W. G. Cady, Piezoelectricity. An Introduction to the Theory and Applications of Electromechanical Phenomena in Crystals, Dover, New York, 1964. 3 M. Curie, Radioactive Substances, (translation of the Thesis presented to the Faculty of Science, Paris), Philosophical Library, New York, 1961. 4 G. Sauerbrey, Z. Phys., 1959, 155, 206. 5 W. H. King, Anal. Chem., 1964, 36, 1735. 6 T. Nomura and A. Minemura, Nippon Kagaku Kaishi, 1980, 1621. 7 P. L. Konash and G. J. Bastiaans, Anal.Chem., 1980, 52, 1929. 8 J. E. Roederer and G. J. Bastiaans, Anal. Chem., 1983, 55, 2333. 9 R. M. White and F. W. Voltmer, Appl. Phys. Lett., 1965, 7, 314. 10 G. S. Calabrese, H. Wohltjen and M. K. Roy, Anal. Chem., 1987, 59, 833. 11 M. Thompson, G. K. Dhaliwal, C. L. Arthur and G. S. Calabrese, IEEE Trans. Ultrason. Ferroelec. Freq. Contr., 1987, UFFC-34, 127. 12 M. Thompson and D. C. Stone, Surface-Launched Acoustic Wave Sensors, John Wiley and Sons, New York, 1997, pp. 34–52. 13 J. Rickert, G. L. Hayward, B. A. Cavic, M. Thompson and W. G�opel, in Sensors Update: Sensor Technology—Applications—Markets, ed. H. Baltes, W. G�opel and J. Hesse, Wiley-VCH, Weinheim, 1999, vol. 5, pp. 105–139. 14 J. C. Andle and J. F. Vetelino, Sens. Actuators A, 1994, 44, 167. 15 J. W. Grate, S. J. Martin and R. M. White, Anal. Chem., 1993, 65, 940A. 16 P. C. H. Li, D. C. Stone and M. Thompson, Anal. Chem., 1993, 65, 2177. 17 P. C. H. Li and M. Thompson, Anal. Chem., 1996, 68, 2590. 18 P. C. H. Li and M. Thompson, Anal. Chim. Acta, 1996, 336, 13. 19 A. C. Stevenson and C. R. Lowe, Sens. Actuators A, 1999, 72, 32. 20 V. I. Bottom, Introduction to Quartz Crystal Unit Design, Van Nostrand Reinhold, New York, 1982. 21 J. G. Miller and D. I. Bolef, J. Appl. Phys, 1968, 39, 4589. 22 K. K. Kanazawa and J. G. Gordon, Anal. Chim. Acta, 1985, 175, 99. 23 S. J. Martin, V. E. Granstaff and G. C. Frye, Anal. Chem., 1991, 63, 2272. 24 R. L. Simpson, PhD Thesis, University of Washington, 1985. 25 G. L. Hayward and G. Z. Chu, Anal. Chim. Acta, 1994, 288, 179. 26 G. L. Hayward and M. Thompson, J. Appl. Phys., 1998, 83, 2194. 27 C. E. Reed, K. K. Kanazawa and J. H. Kaufman, J. Appl. Phys., 1990, 68, 1993. 28 F. Ferrante, A. L. Kipling and M. Thompson, J. Appl. Phys., 1994, 76, 3448. 29 H. L. Bandey, S. J. Martin, R. W. Cernosek and A. R. Hillman, Anal. Chem.,, 1999, 71, 2205. 30 S. Bruckenstein and M. Shay, Electochim. Acta, 1985, 30, 1295. 31 H.Muramatsu, E. Tamiya and I. Karube, Anal. Chem., 1988, 60, 2142. 32 J. Rickert, T. Weiss, W. Kraas, G. Jung and W. G�opel, Biosens. Bioelectron., 1996, 11, 591. 33 M. Rodahl, J. H�o�ok and B. Kasemo, Anal. Chem., 1996, 68, 2219. 34 A. L. Kipling and M. Thompson, Anal. Chem., 1990, 62, 1514. 35 M. Yang and M. Thompson, Anal. Chem., 1993, 65, 1158. 36 M. Yang and M. Thompson, Anal. Chim. Acta,, 1992, 269, 167. 37 J. J. Ramsden, Q. Rev. Biophys., 1994, 27, 41. 38 M. Laatikainen and M.Lindstrom, J. Colloid Interface Sci., 1988, 125, 610. 39 Y. Ebara and Y. Okahata, Langmuir, 1993, 9, 574. 40 M. Yang, F. L. Chung and M. Thompson, Anal. Chem., 1993, 65, 3713. 41 F. Caruso, D. N. Furlong and P. Kingshott, J. Colloid Interface Sci., 1997, 186, 129. 42 M. Rodahl, F. Hook, C. Fredriksson, C. Keller, A. Krozer, P. Brzezinski, M. Voinova and B. Kasemo, Faraday Discuss. Chem. Soc., 1998, 107, 229. 43M. Rodahl, B. Kasemo and P. Brzezinski, Proc. Natl.Acad. Sci. USA, 1998, 95, 12271. 44 R. R. Seigel, P. Harder, R. Dahint, M. Grunze, F. Josse, M. Mrksich and G. M. Whitesides, Anal. Chem., 1997, 69, 3321. 45 B. A. Cavic, F. L. Chu, M. Furtado, S. Ghafouri, G. L. Hayward, D. P. Mack, M. E. McGovern, H. Su and M. Thompson, Faraday Discuss. Chem. Soc., 1997, 107, 159. 46 B. A. Cavic and M. Thompson, Analyst, 1998, 123, 2191. 47 F. Lacour, R. Torresi, C. Gabrielli and A. Caprani, J. Electrochem. Soc., 1992, 139, 1619. 48 S. Nakata, N.Kido, M. Hayashi, M. Hara, H. Sasabe, T. Sugawara and T. Matsuda, Biophys. Chem., 1996, 62, 63. 49 B. Guo, J. Anzai and T. Osa, Chem. Pharm. Bull., 1996, 44, 800. 50 B. S. Murray and L. Cros, Colloids Surf. B: Biointerfaces, 1998, 10, 227. 51 P. W. Walton, M. R. O’Flaherty, M. E. Butler and P. Compton, Biosens. Bioelectron., 1993, 8, 401. 52 Y. Lvov, K. Ariga, I. Ichinose and T. Kunitake, J. Am. Chem. Soc., 1995, 117, 6117. 53 M. Masson, K. Yun, T. Haruyama, E. Kobatake and M.Aizawa, Anal. Chem., 1995, 67, 2212. 54 J. Rickert, A. Brecht and W. Gopel, Anal. Chem., 1997, 69, 1441. 55 S. Ghafouri and M. Thompson, Langmuir, 1999, 15, 564. 56 A. Janshoff, C. Steinem, M. Sieber and H.-J. Galla, Eur. Biophys. J., 1996, 25, 105. 57 A. Janshoff, C. Steinem, M. Sieber, A. Bay�a, M. A. Schmidt and H.-J. Galla, Eur. Biophys. J., 1997, 26, 261. 58 M. Hashizume, T. Sato and Y. Okahata, Chem. Lett., 1998, 399. 1418 Analyst, 1999, 124, 1405–142059 I.Willner, A. Doron and E. Katz, J. Phys. Org. Chem., 1998, 11, 546. 60 L. J. Kricka, in Chemical Sensors, ed. T. E. Edmonds, Blackie, London, 1988, pp. 3–14. 61 L. Bui, R. De Bono, V. Ghaemmaghami, K. M. R. Kallury, P. Li, N. McKeown, D. Stone, H. Su, L. Tessier, S. Vigmond and M. Thompson, in Advances in Biosensors, ed. A. P. F. Turner, JAI Press, Greenwich, CT, 1992, vol. 2, pp. 181–213. 62 D. Leech, Chem. Soc. Rev., 1994, 23, 205. 63 A. Shons, F. Dorman and J. Najarian, J.Biomed. Mater. Res., 1972, 6, 565. 64 B. K�onig and M. Gr�atzel, Anal. Chem., 1994, 66, 341. 65 G. G. Guilbault, B. Hock and R. Schmid, Biosens. Bioelectron., 1992, 7, 411. 66 R. M. Carter, J. J. Mekalanos, M. B. Jacobs, G. J. Lubrano and G. G. Guilbault, J. Immunol. Methods, 1995, 187, 121. 67 C. Steegborn and P. Skl�adal, Biosens. Bioelectron., 1997, 12, 19. 68 K. A. Davis and T. R. Leary, Anal. Chem., 1989, 61, 1227. 69 H. Muramatsu, J. M. Dicks, E. Tamiya and I. Karube, Anal.Chem., 1987, 59, 2760. 70 M. Muratsugu, J. Ohta, Y. Miya, T. Hosokawa, S. Kurosawa, N. Kamo and H. Ikeda, Anal. Chem., 1993, 65, 2933. 71 G. Sakai, T. Saiki, T. Uda, N. Miura and N. Yamazoe, Sens. Actuators B, 1997, 42, 89. 72 C. K�osslinger, S. Drost, F. Aberl and H. Wolf, Fresenius’ J. Anal. Chem., 1994, 349, 349. 73 C. Zhang, G. Feng and Z. Gao, Biosens. Bioelectron., 1997, 12, 1219. 74 K. Bizet, C. Gabrielli, H. Perrot and J. Therasse, Biosens. Bioelectron., 1998, 13, 259. 75 L. Tessier, N. Schmitt, H. Watier, V. Brumas and F. Patat, Anal. Chim. Acta, 1997, 347, 207. 76 F. H�o�ok, M. Rodhal, P. Brzezinski and B. Kasemo, Langmuir, 1998, 14, 729. 77 I. Vikholm and W. M. Albers, Langmuir, 1998, 14, 3865. 78 S. Storri, T. Santoni, M. Minunni and M. Mascini, Biosens. Bioelectron, 1998, 13, 347. 79 F. Caruso, E. Rodda and D. N. Furlong, J. Colloid Interface Sci., 1996, 178, 104. 80 F. Caruso, K. Niikura, D. N. Furlong and Y. Okahata, Langmuir, 1997, 13, 3427. 81 I. Ben-Dov, I. Willner and E. Zisman, Anal. Chem., 1997, 69, 3506. 82 H. Ebato, C. A. Gentry, J. N. Herron, W. M�uller, Y. Okahata, H. Ringsdorf and P. A. Suci, Anal. Chem., 1994, 66, 1683. 83 S. Si, F. Ren, W. Cheng and S. Yao, Fresenius’ J. Anal. Chem., 1997, 357, 1101. 84 K. Nakanishi, H. Muguruma and I. Karube, Anal. Chem., 1996, 68, 1695. 85 R. C. Ebersole and M. D. Ward, J. Am. Chem. Soc., 1988, 110, 8623. 86 J. M. Abad, F. Pariente, L. Hernandez and E. Lorenzo, Anal.Chim. Acta, 1998, 368, 183. 87 K. Yun, E. Kobatake, T. Haruyama, M.-L. Laukkanen, K. Kein�anen and M. Aizawa, Anal. Chem., 1998, 70, 260. 88 M. Tom-Moy, R. L. Bae, D. Spira-Solomon and T. P. Doherty, Anal. Chem., 1995, 67, 1510. 89 G. L. Harding, J. Du, P.R. Dencher, D. Barnett and E. Howe, Sens. Actuators A, 1997, 61, 279. 90 E. Gizeli, Smart Mater. Struct., 1997, 6, 700. 91 E. Gizeli, M. Liley, C. R. Lowe and H. Vogel, Anal. Chem., 1997, 69, 4808. 92 M. Weiss, W. Welsch, M.Schickfus and S. Hunklinger, Anal. Chem., 1998, 70, 2881. 93 J. C. Andle, J. T. Weaver, D. J. McAllister, F. Josse and J. F. Vetelino, Sens. Actuators B, 1993, 13, 437. 94 J. Renken, R. Dahint, M. Grunze and F. Josse, Anal. Chem., 1996, 68, 176. 95 Time, 1999, 153 (1), 26. 96 R. L. P. Adams, J. T. Knowler and D. P. Leader, The Biochemistry of the Nucleic Acids, Chapman and Hal, London, 11th edn., 1992, pp. 593–656. 97 M. Yang, M. E. McGovern and M. Thompson, Anal. Chim.Acta, 1997, 346, 259. 98 N. C. Fawcett, J. A. Evans, L. Chien and N. Flowers, Anal. Lett., 1988, 21, 1099. 99 R. C. Ebersole, J. A. Miller, J. R. Moran and M. D. Ward, J. Am. Chem. Soc., 1990, 112, 3239. 100 F. Caruso, E. Rodda, D. N. Furlong, K. Niikura and Y. Okahata, Anal. Chem., 1997, 69, 2043. 101 K. Ijiro, H. Ringsdorf, E. Birch-Hirschfeld, S. Hoffman, Y. Schilken and M. Strube, Langmuir, 1998, 14, 2796. 102 L. M. Furtado and M. Thompson, Analyst, 1998, 123, 1937. 103 Y. Okahata, M. Kawase, K. Niikura, F. Ohtake, H. Furusawa and Y. Ebara, Anal. Chem., 1998, 70, 1288. 104 Y. Okahata, Y. Matsunabu, K. Ijiro, M. Mukae, A.Murakami and K. Makino, J. Am. Chem. Soc., 1992, 114, 8299. 105 K. Ito, K. Hashimoto and Y. Ishimori, Anal. Chim. Acta, 1996, 327, 29. 106 F. Caruso, E. Rodda, D. N. Furlong and V. Haring, Sens. Actuators B, 1997, 41, 189. 107 J. Wang, P. E. Nielsen, M. Jiang, X. Cai, J. R. Fernandes, D. H. Grant, N. Ozsoz, A. Beglieter and M. Mowat, Anal. Chem., 1997, 69, 5200. 108 H. Zhang, H. Tan, R. Wang, W. Wei and S. Yao, Anal. Chim. Acta, 1998, 374, 31. 109 H. Zhang, R. Wang, H. Tan, L. Nie and S. Yao, Talanta, 1998, 46, 171. 110 S. Yamaguchi, T. Shimomura, T. Tatsuma and N. Oyama, Anal. Chem., 1993, 65, 1925. 111 J. Wang, M. Jiang, T. W. Nielsen and R. C. Getts, J. Am. Chem. Soc., 1998, 120, 8281. 112 H. Su, M. Yang, K. M. R. Kallury and M. Thompson, Analyst, 1993, 118, 309. 113 H. Su, K. M. R. Kallury, M. Thompson and A. Roach, Anal. Chem., 1994, 66, 769. 114 H. Su and M. Thompson, Biosens. Bioelectron., 1995, 10, 329. 115 H. Su, S. Chong and M. Thompson, Langmuir, 1996, 12, 2247. 116 H. Su and M. Thompson, Can. J. Chem., 1996, 74, 344. 117 H. Su, S. Chong and M. Thompson, Biosens. Bioelectron., 1997, 12, 161. 118 K. Niikura, K. Nagata and Y. Okahata, Chem. Lett., 1996, 863. 119 Y. Okahata, K. Niikura, Y. Sugiura, M. Sawada and T. Morii, Biochemistry, 1998, 37, 5666. 120 L. M. Furtado, H. Su, M. Thompson, D. P. Mack and G. L. Hayward, Anal. Chem., 1999, 71, 1167. 121 H. Su, P. Williams and M. Thompson, Anal. Chem., 1995, 67, 1010. 122 Z.-H. Lin, G.-L. Shen, Y. Lin and R.-Q. Yu, Fresenius’ J. Anal. Chem., 1997, 357, 921. 123 M. Aslanoglu, A. Houlton and B. R. Horrocks, Analyst, 1998, 123, 753. 124 M. Yang, H. C. M. Yau and H. L. Chan, Langmuir, 1998, 14, 6121. 125 J. C. Andle, J. F. Vetelino, M. W. Lade and D. J . McAllister, Sens. Actuators B, 1992, 8, 191. 126 J. C. Andle, J. T. Weaver, J. F. Vetelino and D. J. McAllister, Sens. Actuators B, 1995, 24–25, 129. 127 T. Matsuda, A. Kishida, H. Ebato and Y. Okahata, ASAIO J., 1992, 38, M171. 128 J. Redepenning, T. K. Schlesinger, E. J. Mechalke, D. A. Puleo and R. Bizios, Anal. Chem., 1993, 65, 3378. 129 D. M. Gryte, M. D. Ward and W. S. Hu, Biotechnol. Prog., 1993, 9, 105. 130 A. Janshoff, J. Wegener, M. Sieber and H.-J. Galla, Eur. Biophys. J., 1996, 25, 93. 131 C. Fredriksson, S. Kihlman, M. Rodahl and B. Kasemo, Langmuir, 1998, 14, 248. 132 J. L. Brash and D.J. Lyman, J. Biomed. Mater. Res., 1969, 3, 175. 133 A. T. Poot, J. P. Bengeling, J. P. Casenave, A. Banthes and W. G. van Aken, Biomaterials, 1988, 9, 126. 134 H. Ebato, Y. Okahata and T. Matsuda, Kobunshi Ronbunshu, 1993, 50, 463. 135 M. Muratsugu, A. D. Romaschin and M. Thompson, Anal. Ch2, 23. 136 T. Sato, M. Endo and Y. Okahata, J. Biomater. Sci., Polym. Ed., 1995, 7, 587. 137 L. Vroman, Science, 1974, 184, 585. 138 J. L. Brash, in Blood Compatible Materials and Devices, ed. C. P. Sharma and M. Szycher, Technomic Press, Lancaster, 1991, pp. 3– 24. 139 V. I. Sevastianov, in High Performance Biomaterials, ed. M. Szycher, Technomic Press, Lancaster, 1991, pp. 313–341. 140 G. L. Hayward, R. L. Dutton, Z. Zhang, J. M. Scharer and M. Moo Young, Anal. Commun., 1998, 35, 25. 141 S. H. Si, T. A. Zhou, D. Z. Liu, L. H. Nie and S. Z. Yao, Anal. Lett., 1994, 27, 2027. 142 S. Kurosawa, E. Nemoto, M. Muratsugu, M. Yoshimoto, Y. Mori and N. Kamo, Anal. Chim. Acta, 1994, 289, 307. Analyst, 1999, 124, 1405–1420 1419143 M. Muratsugu, S. Kurosawa and N. Kamo, Anal. Chem., 1992, 64, 2483. 144 H. Muramatsu, E. Tamiya, M. Suzuki and I. Karube, Anal. Chim. Acta, 1988, 215, 91. 145 H. Tan, L. Deng, L. Nie and S. Yao, Analyst, 1997, 122, 179. 146 J. Falter, T. Medina and H.-L. Schmidt, Sens. Actuators, 1994, 18–19, 694. 147 T. Hasegawa, J. Nishijo and J. Umemura, J. Phys. Chem. B, 1998, 102, 8498. 148 Y. Iwasaki, S. Tanaka, M. Hara, K. Ishihara and N. Nakabayashi, J. Colloid Interface Sci., 1997, 192, 432. 149 M. Nakamura, K. Tsujii and J. Sunamoto, J. Med. Biol. Eng. Comput., 1998, 36, 645. 150 J. J. Chance and W. C. Purdy, Anal. Chem., 1996, 68, 3104. 151 T. Tatsuma and D. A. Buttry, Anal. Chem., 1997, 69, 887. 152 S. M. Reddy, J. P. Jones, T. J. Lewis and P. M. Vadgama, Anal. Chim. Acta, 1998, 363, 203. 153 J. M. Abad, F. Pariente, L. Hern�andez, J. D. Abr�una and E. Lorenzo, Anal. Chem., 1998, 70, 2848. Paper 9/03236C 1420 Analyst, 1999, 124, 1405–

ISSN:0003-2654    

DOI:10.1039/a903236c

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

A new purge-and-membrane mass spectrometric (PAM-MS) instrument for analysis of volatile organic compounds in soil samples Marja Ojala, Ismo Mattila, Timo Särme, Raimo A. Ketola and Tapio Kotiaho* VTT Chemical Technology, P.O. Box 1401, FIN-02044 VTT, Finland. E-mail: tapio.kotiaho@vtt.fi Received 25th June 1999, Accepted 26th August 1999 Purge-and-membrane mass spectrometry (PAM-MS) is a combination of dynamic headspace technique and membrane extraction. A new purge-and-membrane sampler is introduced and its basic test results, such as the effects of sample pre-heating/desorption temperature, pre-heating time, moisture content in soil, methanol concentration in the samples, purge gas and soil type are briefly reported.Three different soil types were used in the method development, namely sand, peat and commercial garden soil. Some authentic soil samples were also analyzed and the results were compared to those obtained with static headspace gas chromatography.The agreement was generally good. Introduction Volatile organic compounds (VOCs) have caused serious environmental problems during the recent decades. The main emission sources are industry, traffic and energy production. Several analytical methods have been applied for VOC determination from water and soil samples,1–5 including on-site and on-line analytical methods.6 Among these methods membrane inlet mass spectrometry (MIMS) is one of the most suitable techniques for the determination of VOCs in air7 and water samples.8 However, MIMS is not suitable for analysis of soil or sludge samples.A recently introduced technique, purgeand- membrane mass spectrometry (PAM-MS), which combines dynamic headspace and membrane inlet mass spectrometry, 9 is also suitable for the determination of VOCs in soil samples. In this technique VOCs are purged from the samples with inert gas, the gas stream is directed through a membrane inlet and analyte/matrix molecules permeate through the membrane into the ion source of a mass spectrometer for mass spectrometric measurement.In the following a new prototype of a PAM-sampler, representing the first step of full automation of the PAM-MS method, is described. The results obtained in detailed characterization of the instrumental/analytical properties of the PAM-MS method, such as the effect of soil type, humidity and desorption temperature on peak areas and desorption times are also briefly presented.Experimental The mass spectrometers used were a Balzers QMG 421C quadrupole mass spectrometer (mass range 1 to 500 u) equipped with an open cross-beam electron impact (70 eV) ion source (Balzers Aktiengesellschaft, Balzers, Liechtenstein) and a Balzers Omnistar quadrupole mass spectrometer (mass range 1 to 300 u) equipped with a closed electron impact ion source. Custom-made membrane inlets utilizing a sheet membrane, built at VTT Chemical Technology,7 were used in both mass spectrometers.The temperature of the membrane inlet in both mass spectrometers was typically 70 °C. The material of the sheet membrane was dimethylpolysiloxane (SSP-M100, Specialty Silicone Products Inc., Ballston Spa, NY, USA) with dimensions: thickness 25 mm and contact area 28 mm2 for QMG 421C and 10 mm2 for Omnistar. Most of the testing was performed using the selected ion monitoring mode (SIM). In the analysis of the authentic soil samples the scanning mode was also used, the mass range measured being 46 to 200 u.A schematic diagram of the new purge-and-membrane mass spectrometric apparatus is presented in Fig. 1. The main parts of Fig. 1 A schematic diagram of a PAM-sampler showing the standby mode of operation (a) and the sampling mode (b). Direction of gas flow is indicated by arrows. 1, Membrane inlet mass spectrometer; 2, Convection oven; 3, Six-port valve; 4, Gas flow control unit for purge gas; 5, Sample vessel. This journal is © The Royal Society of Chemistry 1999.Analyst, 1999, 124, 1421–1424 1421the device are a membrane inlet mass spectrometer (1, see above), a convection oven with adjustable temperature up to 300 °C (2, Meyer-vastus, Monninkylä, Finland), a six-port valve (3, E3C6UWT, Valco Co, Schenkon, Switzerland) and a custom-made gas flow control unit for purge gas (4). The gas flow control unit was built using standard Swagelok® parts (Helsinki Valve & Fitting, Helsinki, Finland), a mechanical HP gas chromatograph flow controller (Hewlett Packard, Palo Alto, CA, USA), 1/8 in (3.18 mm) S-series standard needle valves (Nupro Co, Ohio, USA), a Bürkert 127 3-way on/off valve (Bürkert, Ingelfinssen, Germany) and using 1/8 in (3.18 mm) or 1/16 in (1.59 mm) copper or nickel tubing.All gas lines in contact with the gas stream containing VOCs were 1/16 in (1.59 mm) nickel tubing and they were heated to 150 °C in order to minimize contamination.The sample vessels (Fig. 1, 5) were made by cutting off bottoms of two commercial 10 mL headspace vials and connecting the truncated vials together from the cut ends. A glass sinter for supporting the sample was mounted on the bottom of the vessel during the manufacturing process. The gas chromatograph (GC) used for headspace GC analysis was an HP 5890 Series II, equipped with two flame ionization detectors and with an HP 7694 headspace sampler (Hewlett Packard). Two separate columns were used in order to improve the identification capability of the headspace GC method, namely DP-1 and a DP-1701 (J & W Scientific, Folsom, CA, USA). Both columns were 30 m 30.32 mm id, with a stationary phase thickness of 0.25 mm.The stock solutions of test compounds and an internal standard (m-fluorotoluene) were prepared by weighing 1 g of the compound and dissolving it in 100 mL of methanol. Further dilutions of the stock solutions were made with methanol and 5–10 mL were spiked to the soil samples.Typically, 5 g of a soil sample was weighed to a sample vessel and the internal standard, m-fluorotoluene, was injected through the septum of the sample vessel to the soil sample. The soils used in testing were sand (Seesand, purum, Fluka, Switzerland), commercial garden soil (Kekkilä Oy, Parkano, Finland) and peat (garden peat, unmanured, GardenPeatt Oy, Mellilä, Finland). The water and organic content of the soils were determined by drying samples at 102 °C and thereafter burning off the organic compounds at 550 °C.The following results were obtained: garden soil had a water content of 17% and 9% organics; sand had a water content of 0% and 0% organics and peat had a water content of 39% and 60% organics. The authentic soil samples used in testing were obtained from customers of VTT Chemical Technology. Results and discussion The new PAM-MS instrument is presented in Fig. 1 in the stand-by and sampling modes.In the stand-by mode a purge gas flows through all the gas lines in contact with VOCs and the membrane inlet in order to prevent contamination and to provide a constant background for the mass spectrometer. The measurement procedure starts by preheating the sample in the oven. After a selected pre-heating time the sample is mounted into the sample holder (see Fig. 1), in the next step the six-port valve is switched to the sampling position and finally the sampling lines with needles are manually punctured through the sample vessel septums.This last step directs the flow of purge gas through the sample, desorption of VOCs occurs and ion chromatograms for VOCs can be measured with the mass spectrometer while the purge gas containing the VOCs flows through the membrane inlet. An example of the typical desorption curves obtained with the PAM-MS method is presented in Fig. 2. The curves presented were measured for a garden soil sample spiked with benzene, toluene, o-xylene and 1,3,5-trimethylbenzene at concentrations of 2 mg kg21.The operation of the instrument has been thoroughly tested and detailed characterization of the instrumental/analytical properties of the PAM-MS method are presented below. Effect of soil moisture content The effects of moisture on desorption peak areas and desorption times were studied using garden soil, sand and peat samples. The amounts of moisture in the samples were adjusted by adding water to the samples before the analysis.A selected set of test compounds was used, e.g., MTBE (methyl tert-butyl ether), tetrachloroethene, 1,1,1-trichloroethane and o-xylene. According to the measurements the variation of moisture in the samples had no significant effect on desorption peak areas, i.e. the areas under the measured desorption curves of the individual compounds were the same within ±25% when measured by SIM. However, increasing moisture content in the samples increased the standard deviation in the peak area measurements.The desorption times, defined as the time between 0 and 90% of the desorption yield, were normally longer when the moisture content of the samples was higher. For example, the desorption time of MTBE for dry sand increased from 17 to 26 s when measured from sand with a moisture content of 17%. A clear dependence of the desorption times on soil type was also observed, desorption times increasing as the amount of organic material in the samples increased.Further investigations of the effects of moisture on desorption peak areas and desorption times are in progress. Effect of pre-heating/desorption temperature and pre-heating time of the sample The effects of the pre-heating/desorption temperature and preheating time on the desorption times and desorption peak areas were studied using a selected set of compounds and garden soil. The effects of the pre-heating/desorption temperature on the desorption times of the selected compounds are presented in Table 1 (the pre-heating time was 20 min) and as an example the effect of pre-heating/desorption temperature on the recovery curve of toluene measured from garden soil at various preheating/ desorption temperatures is presented in Fig. 3. The yield curves of toluene, calculated from similar desorption curves as presented in Fig. 2, are shown since these provide a very clear visual demonstration of the effects of increasing pre-heating/ desorption temperature on the desorption times (see also Table 1), i.e., the desorption times decreased considerably when the pre-heating/desorption temperature was increased. Interestingly, it was observed that the desorption peak areas decreased Fig. 2 Determination of 2 mg kg21 benzene (m/z 78), toluene (m/z 92), oxylene (m/z 106) and 1,3,5-trimethylbenzene (m/z 105) in a garden soil sample by PAM-MS. 1422 Analyst, 1999, 124, 1421–1424as the pre-heating/desorption temperature increased.It was also observed that the peak height of the desorption peaks increased slightly as the pre-heating/desorption temperature increased. Higher desorption peaks are observed at higher pre-heating/ desorption temperature, because the amount of VOCs in the headspace of the sample vessels increases as a function of temperature due to the increased vapour pressures. However, the desorption peak areas decrease because at higher preheating/ desorption temperatures a smaller portion of VOCs is sampled due to rapid desorption times and due to the fact that only a small part of the VOCs in the purge gas stream is sampled via the membrane inlet.On the basis of the results presented a pre-heating/desorption temperature of 80 °C was selected to be used in further studies. Different pre-heating times, 10, 15, 20, 30 and 40 min were also tested, but no significant differences in desorption times or desorption peak areas were observed, a result which is in good agreement with our earlier static headspace gas chromatographic results.10 A pre-heating time of at least 10 min has been used in further studies.Effect of methanol concentration of the soil samples The effects of methanol concentration on desorption peak areas were studied because the standard mixtures for spiking and calibration were made in methanol. This was done by spiking garden soil samples (5 g) with selected VOCs (MTBE, 1,1,1-trichloroethane, o-xylene, tetrachloroethene and mfluorotoluene, concentration of each 0.2 mg kg21) and changing the amount of methanol in the samples (methanol added 20, 110 or 210 mL).The desorption peak areas of all the analytes decreased considerably as the methanol content increased from 20 mL to either 110 or 210 mL, the decrease being typically at least 50%. In order to minimize the effect of methanol its amount in further analyses was reduced to 5 or 10 mL per sample. Effect of purge gas The effects of purge gas on the desorption peak areas and on the desorption times were studied using helium, carbon dioxide, nitrogen or synthetic air as purge gas.Garden soil was used in these experiments and selected compounds were spiked to a concentration of about 2 mg kg21 two days before analysis. The effects of various purge gases on the desorption peak areas are presented in Table 2. It can be seen that the differences in the desorption peak areas between the various purge gases are not very great for any of the compounds used in testing.From Table 2 it can be also observed that the relative standard deviations in the desorption peak area determinations are generally of the same order of magnitude for all the purge gases. In addition, it was observed that the desorption times were the same for all the purge gases. On the basis of these results nitrogen and synthetic air were selected as purge gases for further studies. The low price of these gases also guided this selection, and purified air is relatively easy to make using mobile air purifiers, a fact which is important when a purge gas is selected for on-site applications of the PAM-MS method.Linearity and detection limits The linear dynamic range of the new PAM-MS apparatus was measured using garden soil (moisture 17% and organic concentration 9%) as a standard soil and MTBE, 1,1,1-trichloroethane, o-xylene and tetrachloroethene as test compounds. The linear dynamic ranges for the selected compounds were rather wide, from 100 mg kg21 to at least 1 g kg21 (11 different concentrations).The correlation coefficients of the calibration lines were typically at least 0.994. It is worthy of notice that the limits for remediation of contaminated soil in Finland are well within the linear dynamic range of the PAMMS method, for example the limit of remediation is 25 mg kg21 for benzene and 600 mg kg21 for 4-nonylphenol.11 The detection limits (S/N = 3) measured using the Omnistar mass spectrometer were of the same order of magnitude as in our earlier studies,9 being 1 mg kg21 for o-xylene, 10 mg kg21 for 1,1,1-trichloroethane, 10 mg kg21 for tetrachloroethene and 5 mg kg21 for MTBE.Table 1 Effects of pre-heating/desorption temperatures on the desorption times of a selected set of compounds. Temperature/°C Compound 50 65 80 95 Benzene 34a 22 16 11 Toluene 51 33 22 12 1,2,4,5-Tetramethylbenzene nmb 227 175 154 1,2-Dichloroethene 34 22 13 7 Trichloroethene 41 26 14 8 a Desorption time/s.b Not measured. Fig. 3 The effect of pre-heating/desorption temperature on the desorption time of toluene from garden soil. Table 2 Effect of purge gas on peak areas (four or five replicate measurements) of a selected set of compounds. Garden soil was used in testing MTBE 1,1,1-Trichloroethane o-Xylene Tetrachloroethene Gas Mean peak areaa RSD (%) Mean peak areaa RSD (%) Mean peak areaa RSD (%) Mean peak areaa RSD (%) Synthetic air 4500 4.6 3900 4.1 31 000 3.1 10 000 1.9 Helium 6200 9.4 3700 9.3 40 000 2.2 13 000 5.6 Carbon dioxide 4200 9.4 3400 12.3 34 000 3.6 12 000 5.6 Nitrogen 5900 3.1 3900 2.5 38 000 2.5 13 000 4.2 a Arbitrary units Analyst, 1999, 124, 1421–1424 1423Analysis of authentic soil samples Three authentic soil samples were analyzed with the PAM-MS method combined with the Solver calculation program,12 PAMMS method with selected ion monitoring (SIM) and a headspace gas chromatographic (HSGC) method (Table 3).Samples 1 and 2 were clay and sample 3 was moist sand. The Solver calculation program is a custom-made computer program which performs identification and quantification of VOCs in unknown samples on the basis of the measured MIMS mass spectra.9 The results obtained with the three different methods were in relatively good agreement. The amount of toluene could not be determined reliably with the SIM method due to large amounts of xylenes and ethylbenzene, which have spectral overlap with toluene. Note also that the mass spectra of MTBE and TAME (tert-amyl methyl ether) are so similar that neither Solver nor SIM can quantitate them separately, but their total amount was in good agreement with the amount obtained with the HSGC method.In conclusion, the tests show that the new purge-andmembrane mass spectrometric apparatus is a very promising system for the determination of VOCs in soil samples.In the PAM-MS method all volatiles are purged from the sample and therefore the results are independent of soil type and humidity. Other advantages of the method are short analysis times, the non-requirement for pre-treatment of samples and the fact that solvents are not used. The studies will continue by constructing a new smaller PAM-sampler for on-site applications and by more detailed studies of the instrumental/analytical properties of the PAM-MS method reported.Acknowledgements The authors acknowledge financial support from the Technology Development Centre (Tekes), Fortum Oil and Gas, Golder Associates and Finnish Measurement Systems (FMS). We thank Risto Kostiainen, Marjo Poutanen, Katja Pihlainen and Anniina Määttänen for their assistance. References 1 R. A. Ketola, V. T. Virkki, M. Ojala, V. Komppa and T. Kotiaho, Talanta, 1997, 44, 373. 2 C. E. Koester and R. E. Clement, Cri. Rev. Anal. Chem., 1993, 24, 263. 3 R. E. Clement, P. W. Yang and G. A. Eiceman, Anal. Chem., 1997, 69, 251R. 4 B. A. Schumacher and S. E. Ward, Environ. Sci. Technol., 1997, 31, 2287. 5 A. D. Hewitt, Environ. Sci. Technol., 1998, 32, 143. 6 T. Kotiaho, J. Mass Spectrom., 1996, 31, 1. 7 R. A. Ketola, M. Ojala, H. Sorsa, T. Kotiaho and R. K. Kostiainen, Anal. Chim. Acta, 1997, 349, 359. 8 F. R. Lauritsen and T. Kotiaho, Rev. Anal. Chem., 1996, 15, 237. 9 R. Kostiainen, T. Kotiaho, I. Mattila, T. Mansikka, M. Ojala and R. A. Ketola, Anal. Chem., 1998, 70, 3028. 10 M. Ojala, Licentiate’s Thesis, University of Helsinki, 1998. 11 Saastuneet maa-alueet ja niiden käsittely Suomessa, Ympäristöministeriö/ ympäristönsuojeluosasto, Helsinki, Finland, 1994. 12 R. A. Ketola, M. Ojala, V. Komppa, T. Kotiaho, J. Juujärvi and J. Heikkonen, Rapid Commun. Mass Spectrom., 1999, 13, 654. Paper 9/05106F Table 3 The quantitative results of some authentic soil samples using different analytical methods Content/mg kg21 Sample Compound Solver SIM HSGC 1 Toluene 1.8 2.8 Xylenesa 19 17 10 MTBE 43b 31b 31 TAME < 1 2 Toluene 1.0 3.1 Xylenesa 7.3 9.6 2.6 MTBE 18b 21b 17 3 Toluene 1.3 2.4 Xylenesa 8.9 7.5 10 a Results are sums of xylenes and ethylbenzene. b Results are sums of MTBE and TAME. 1424 Analyst, 1999, 124, 1421–1424

ISSN:0003-2654    

DOI:10.1039/a905106f

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Characterisation and determination of phytochelatins in plant extracts by electrospray tandem mass spectrometry Véronique Vacchina,a Hubert Chassaigne,a Matjaz Oven,b Meinhard H. Zenkb and Ryszard Oobi �nski*a a CNRS EP132, Hélioparc, 2, av. du Président Angot, 64000 Pau, France. E-mail: Ryszard.Lobinski@univ-pau.fr b Lehrstuhl für Pharmazeutische Biologie, Ludwig-Maximilians-Universität, Karlstr. 29, D-80333 München, Germany Received 28th June 1999, Accepted 20th August 1999 A method based on pneumatically assisted electrospray ionisation tandem mass spectrometry (ESI MS-MS) was developed for the identification, sequencing and determination of phytochelatin (PC) peptides in plant tissue and plant cell cytosols.The ionization and fragmentation conditions were optimized using a series of (GluCys)2Gly (PC2), (GluCys)3Gly (PC3), and (GluCys)4Gly (PC4) standards prepared from glutathione by enzymatically (g-glutamylcysteine dipeptyl transpeptidase) assisted biosynthesis in the presence of Cd2+.Phytochelatins were found to ionize readily to produce a characteristic mono-protonated ion. The collision-induced dissociation (CID) of this ion followed by mass spectrometry (MS-MS mode) allowed the determination of the amino acid sequence of each of the PCs. Calibration curves were linear up to a concentration of 2 mg ml21 in the MS and MS-MS modes with the detection limits at the low ng ml21 level. The method was applied to the determination of phytochelatin peptides biosynthesized by a number of plant cell cultures exposed to the Cd stress. The results agreed with those obtained by an independent procedure based on reversed-phase HPLC with post-column derivatization of the –SH groups with 5,5A-dithiobis-2-nitrobenzoic acid and spectrophotometric detection.Introduction Speciation of heavy metals in plants has been attracting considerable interest as a way to understand the internal mechanisms allowing the living organisms to grow in an environment contaminated by heavy metals.1 Most frequently, as a response to the metal stress, plants biosynthesize a ligand, such as a phenolic compound, organic acid, or oligo- or polypeptide that is able to complex the excess of the toxic element into a compound innocuous to the organism.1–4 Particular attention has been paid to phytochelatins (PCs) which are a class of peptides composed only of three amino acids: cysteine (Cys), glutamic acid (Glu) and glycine (Gly) and in which glutamic acid is linked to cysteine through a g-peptide linkage.Their general formula is (GluCys)nGly where n is between 2 and 11.2–4 They are synthesized from glutathione (GSH) in the presence of some heavy metals during a reaction catalysed by the enzyme g-glutamylcysteine dipeptyl transpeptidase (PC-synthase).5 PCs can detoxify these metals by forming a metal–PC complex in which the metal is bound to the thiol group of the cysteine unit.2–5 The general structure of phytochelatins is conservative in a wide variety of plants but some modifications may occur on the C-terminal amino acid. For example, instead of glycine, b-alanine was found in some plants (Fabacea),6,7 serine8 and glutamic acid9 were reported in rice and maize, respectively, whereas des-Gly phytochelatins (GluCys)n were found in yeast.10 These modified PCs are named iso-PCs.Even though some hyphenated techniques, such as sizeexclusion chromatography (SEC) with inductively coupled plasma mass spectrometric (ICP-MS) detection offer an attractive way to monitor the induction of PCs and binding of heavy metals to these ligands, the poor resolution of this separation technique and the lack of the molecular specificity of the detector allow neither the differentiation between the different PCs and iso-PCs complexes nor the differentiation between PCs and other ligands.11,12 The key to the understanding of the heavy metal detoxification mechanisms in plants is the unambiguous identification, characterization and quantification of the bio-induced ligands.The classical approach to the analysis of PCs is reversed phase HPLC with post-column derivatization of the sulfhydryl groups with 5,5’-dithiobis(2-nitrobenzoic acid) (DTNB, Ellman’s reagent) and spectrophotometric detection at 410 nm.13–16 The detection is not specific to PCs; any compound containing a sulfhydryl group is able to produce a signal.The signal identification needs therefore to be based on matching the retention times of the analyte compounds with the corresponding standards. The latter are usually unavailable. Even if they were available, ambiguities with the identification using this approach may occur, especially in the case of iso-PCs which have a similar structure and retention times to the corresponding PCs. The use of an analytical technique able to detect compounds specifically, for example mass spectrometry, is therefore required.Positive ion fast atom bombardment tandem mass spectrometry (FAB MS-MS) was proposed as an elegant and species specific method for fingerprinting of PC peptides.17,18 This technique lacks the sensitivity and is difficult to use for quantitative analysis. These drawbacks can be overcome by using electrospray tandem mass spectrometry (ESI MS-MS) that is becoming a well-established tool for peptide identification and sequencing in biological samples.19,20 ESI MS-MS using a triple quadrupole configuration has also an emerging potential as a sensitive species-selective quantification technique.Recently, it was proposed for quantitative speciesselective analysis of cobalamin analogues at ng levels.21 The objective of this work was to develop a simple sensitive method for the identification, sequencing and quantitative determination of Cd-induced phytochelatins in plants and plant cell cultures by ESI MS-MS following a custom-designed sample preparation procedure.This journal is © The Royal Society of Chemistry 1999. Analyst, 1999, 124, 1425–1430 1425Experimental Apparatus All experiments were performed using a Perkin-Elmer SCIEX (Thornhill, ON, Canada) API 300 pneumatically-assisted electrospray (ion-spray) triple-quadrupole mass spectrometer. A Model 1100 HPLC pump (Hewlett-Packard, Wilmington, NC, USA) was used as the sample delivery system for the purification of the plant and plant cell extracts by reversedphase chromatography.A Hitachi (Tokyo, Japan) Model CS 120 GX refrigerated ultracentrifuge was used for the separation of the supernatant after leaching of Cd species from plant tissues and cell cultures. The solvents were degassed by means of an ultrasonic bath. Materials A reversed-phase Vydac (Hesperia, CA) C8 150 mm 3 4.6 mm id column was used for the purification of the plant tissues and plant cell extract. Acetonitrile and methanol (Sigma-Aldrich, Saint-Quentin Fallaviec, France) were of HPLC grade.Trifluoroacetic acid (TFA) and dithiothreitol (DTT) were purchased from Sigma-Aldrich. Water purified using a Milli-Q system (Millipore, Bedford, MA, USA) was used throughout. Cell cultures of plants: Silene cucubalus, Agrostis tenuis, and Rauvolfia serpentina were investigated. They were grown for four days in a 300 mM Cd2+ solution. Phytochelatin standards The synthesis of phytochelatins was carried out according to Grill et al.2 In brief, 2000–10 000 pkat of PC-synthase5 were incubated at 25 °C with 1 mM GSH and 0.8 mM Cd(NO3)2 in 1.2 l of buffer solution (pH 8.0).An amount of 0.02% of NaN3 was added to retard the bacterial growth. Proteins were precipitated from the resulting solution by the addition of (NH4)2SO4 to reach a concentration of 85% followed by centrifugation at 8000g for 30 min. Phytochelatins were precipitated from the supernatant by the addition of 20 ml of 1 M Cd(NO3)2 solution.The precipitated mixture of phytochelatins was centrifuged and washed twice with water. The washed precipitate was stable for months when stored at 220 °C. The precipitated phytochelatins were dissolved in 3.5 M HCl and separated by semi-preparative HPLC using elution with a concentration gradient of acetonitrile in water.2 The fractions with the individual phytochelatin (PC2, PC PC4) peaks were heart-cut. The acetonitrile was rotavaporated and the fractions were lyophilized.The heart-cut fractions were repurified in the same way to produce compounds used as standards below. Analytical procedures Sample preparation. Cells were vacuum filtered and washed. Plant roots were cut from the rest of the plant and washed (in both cases only with water). From here on, both cells and plants were treated the same way. They were frozen in liquid nitrogen to break the cell wall, ground with a pestle and mortar and extracted with water or with 10 mM TRIS-HCl buffer (pH = 8).They were centrifuged (30 min, 10 000g, 4 °C), filtered and lyophilized. A sample of 30 mg of the freeze-dried material was dissolved in 500 ml of water containing 0.1% of TFA (pH = 2.3). The solution was filtered and injected on the reverse-phase purification column. An aliquot of 100 ml was eluted with 0.1% TFA in water for 5 min followed by a linear increase of acetonitrile concentration in the eluent up to 50% during 30 min. The flow rate was set at 0.75 ml min21.The eluate containing acetonitrile was collected and incubated for 20 min at 25 °C with 5 mM DTT. The acetonitrile was removed by rotavaporation and the aqueous residue was freeze-dried. The dried residue was dissolved in 200 ml of 0.06 M acetic acid in 30% methanol, diluted if necessary to fit the linearity range of the calibration curve, and analysed by ESI MS. Instrumental ESI MS-MS conditions. In the MS mode, Q1 was swept over a given mass range and Q3 was operated in rfonly mode.The orifice potential was 40 V, the ion spray voltage was 4100 V and the ion multiplier potential was 2400 V. The total spectra of PCs were acquired in the range 50–1100 Da using a 10 ms dwell time and a 0.5 Da step size during 10 scans. In the MS-MS mode, the product ion scan mode was chosen for data acquisition. The mass of Q1 was fixed and Q3 was swept over a given mass range to determine the ions which result from the fragmentation of the precursor ion.The collision gas was nitrogen and the collision energies were 24 eV for GSH, 26 eV for PC2, 34 eV for PC3 and 49 eV for PC4. For the calibration curves in direct introduction, the step size was 0.05 and the dwell time was 5 ms. The mass ranges selected for the calibration curves in the MS mode were 290–320 u for GSH, 530–550 u for PC2, 760–790 u for PC3 and 990–1020 u for PC4. In the MS-MS mode the mass ranges selected were: 305–311 u, 536–543 u, 769–775 u and 1001–1007 u for GSH, PC2, PC3 and PC4, respectively.The biggest y-fragment in the MS-MS mode was monitored in the ranges: 176–182, 408–414, 640–646, and 872–878 for GSH, PC2, PC3 and PC4, respectively. Results and discussion Preliminary experiments were aimed at maximizing the signal/ background ratio obtained in the MS and MS-MS modes, by direct introduction of the sample solution. Methanol and acetonitrile were investigated as organic solvents whereas acetic acid and HCl were used for acidification.The most intense signals were obtained for a mixture of methanol with water 30 + 70 v/v). The maximum signal was obtained at a concentration of 0.06 M acetic acid. Characterization of PCs standards in the MS mode Fig. 1 shows mass spectra obtained for the individual GSH, PC2, PC3 and PC4 standards. These spectra are principally constituted by the protonated molecule ion [M + H]+ which allows the identification of each PC according to its mass. There are also intense but unidentified signals present at 159 and 288 u and some other minor background signals.In the case of PCs, even if the dissociation of the protonated molecule ion is weak, the formation of certain fragments cannot be prevented, especially that of the heaviest y-fragment of the protonated molecule ion. This is, for example, the case of the m/z = 411 u peak in the PC2 spectrum or of the m/z = 643 u peak in the PC3 spectrum. Characterization of PCs standards in the MS-MS mode In the MS-MS mode, the protonated molecule ion is fragmented by collision induced dissociation (CID) with inert gas molecules. Peptides fragment primarily at the amine bonds to produce a ladder of sequence ions.20 The charge can be retained on the amino terminus (type b-ion) or on the carboxy terminus (type y-ion).Thus a complete series made of ions from both 1426 Analyst, 1999, 124, 1425–1430types allows the determination of the amino acid sequence by subtraction of the masses of adjacent sequence ions.Fig. 2 shows the mass spectra produced by the fragmentation of the molecular peaks observed in Fig. 2 for each of the compounds investigated. Most of the expected b- and y-type fragments can be identified. As observed elsewhere in FAB MS the g-GluCys linkage is hardly broken.18 The peaks corresponding to the loss of a g-GluCys group are less intense. This is for example the case for the peak at 362 u in the PC2 spectrum.In the MS-MS mode, two ions are characteristic of a PC: the protonated molecule ion and the y-type fragment of the biggest mass. For lower masses, the MS-MS spectrum of PCn overlaps Fig. 1 Electrospray mass spectra of GSH and purified PCs standards subtracted from the blank. (a) GSH; (b) PC2; (c) PC3; (d) PC4. Ca. 1 mg ml21 of each PC. Fig. 2 Electrospray tandem mass spectra of GSH and purified PCs standards. (a) GSH, collision energy = 24 eV; (b) PC2, collision energy = 26 eV; (c) PC3, collision energy = 34 eV; (d) PC4, collision energy = 49 eV.Ca. 1 mg ml21 of each PC. Analyst, 1999, 124, 1425–1430 1427with the PCn21 one. Nevertheless the MS-MS mode offers a possibility to detect a PC unambiguously, even if its molecular peak overlaps with one or two more compounds in the ESI MS spectrum. Precision, linearity and detection limits ESI MS offers an until now unexplored opportunity of the quantitative species-selective determination of the individual phytochelatins in plant extracts.Calibration curves were established for the molecular MS mode and the tandem (MSMS) mode. Precision, determined by five fold injection at the 1 mg ml21 level, was in the range 3–5%. Calibration curves were plotted for the characteristic ion of each of the compounds determined. In the MS mode it was the protonated molecule ion [M + H]+, whereas in the MS-MS mode the use of the protonated molecule ion and the y-type fragment having the biggest mass was investigated for the purpose of quantification.The linearity of the instrumental response as a function of PC concentration was investigated in the range 0–10 mg ml21. It was found that, both in the MS and the MS-MS mode, the signal intensity is a linear function of the PCs concentration up to 2 mg ml21. Table 1 summarizes the sensitivities (slopes), correlation coefficients and detection limits (DL) obtained in each case. In the case of interferences with compounds of the same mass, PCs can be identified and quantified using the largest y-type fragment.The detection limits were calculated as three times the standard deviation of the blank measured for a given ion. The detection limits in the MS and the MS-MS modes are similar. They deteriorate rapidly with the increasing molecular mass of the PC-species investigated. Analysis of the biosynthesis post-reaction mixture The method developed was applied to the analysis of PCs mixture resulting from the reaction of glutathione with the enzyme phytochelatin synthase.The mass spectrum obtained is shown in Fig. 3 The intense peaks at 288 and 159 u are still present but the molecular peak of GSH at 308 u, PC2 at 540 u, PC3 at 772 u and PC4 at 1004 u can be identified unambiguously. The CID fragmentation of all those molecular peaks leads to the characteristic peaks of PCs as in Fig. 2 (data not shown). This allows the identification of the PCs present in the mixture without the need for HPLC. The different analytical modes: ESI MS, ESI MS-MS using the molecular peak and ESI MS-MS using the biggest y-type fragment were compared for the determination of phytochelatins in the crude post-reaction mixture.The results are shown in Table 2. It is evident that glutathione which is the substrate for the enzyme-assisted PC synthesis is almost totally consumed. PC2, PC3 and PC4 are synthesized in proportions 10+5+3. The analytical result is practically independent of the instrumental operating mode used for quantification.Determination of phytochelatins in plant extracts The method developed was further applied to the characterization and determination of PCs in extracts of three different plant cell cultures (Silene cucubalus, Agrostis tenuis, and Rauvolfia serpentina) known to biosynthesize phytochelatins when exposed to Cd2+. The procedures investigated to extract the Cd complexes included extraction with water and with 10 mM TRIS-HCl buffer (pH 8.0).Table 3 shows that the extraction procedure has hardly any effect on the recovery of PC from a sample. Water was therefore used for extraction of metal–peptide complexes in order not to introduce additional salt to the extract that would be preconcentrated by freezedrying. Even when water is used as extractant, the extracts contain a considerable concentration of salt and cannot be analysed directly by ESI MS. The expected concentrations in the plant cell cultures analysed are much lower than in the case of the post-reaction mixture referred to above so the reduction of the salt concentration by dilution cannot be applied.The use of a size-exclusion desalting column as used in the protein analytical biochemistry fails because the PC-species are much smaller Table 1 Figures of merit of ESI MS and ESI MS-MS for the quantitative determination of glutathione and phytochelatins using different data acquisition modes Protonated molecule ion in the MS mode Protonated molecule ion in the MS-MS mode Heaviest y-type fragment in the MS-MS mode Compound Slope, counts s21/mg l21 r2 DL/mg l21 Slope, counts s21/mg l21 r2 DL/mg l21 Slope, counts s21/mg l21 r2 DL/mg l21 GSH 8602 0.9985 1 118.9 0.9918 17 727.4 0.9982 0.5 PC2 9471 0.9982 9 209.2 0.9964 9 91.0 0.9952 6 PC3 585.7 0.9985 7 17.3 0.9967 38 6.1 0.9837 43 PC4 144.1 0.9962 28 2.5 0.9979 42 1.0 0.9960 40 Fig. 3 Electrospray mass spectrum obtained for the post-reaction mixture in the enzymatically mediated biosynthesis of PC from glutathione.Peak identification: 1, GSH; 2, PC2; 3, PC3; 4, PC4. Table 2 Comparison of the results of the quantitative determination of glutathione and phytochelatins in the post-rection mixture of enzymatically assisted biosynthesis using the different data acquisition modes Concentration in the analysed solution/mg l21 Compound Protonated molecule ion in the MS mode Protonated molecule ion in the MS-MS mode Heaviest y-type fragment in the MS-MS mode GSH 35 ± 2 37 ± 2 29 ± 2 PC2 1056 ± 53 1104 ± 56 1193 ± 60 PC3 519 ± 26 462 ± 24 607 ± 31 PC4 354 ± 18 392 ± 20 411 ± 21 1428 Analyst, 1999, 124, 1425–1430than proteins and may co-elute with the salt when diluted from a desalting column.Therefore, reversed-phase chromatography was applied (see Fig. 4). Apophytochelatins (non-metallated PCs) are known to be retained on a C8 support from aqueous and slightly (up to 5%) organic media containing 0.1% TFA5 under which conditions the salts are eluted. PCs can be recovered from the column by increasing the concentration of an organic modifier (acetonitrile) in the mobile phase.Such a procedure eliminates salts but introduces TFA that is known to suppress electrospray ionization, and leads to a considerable dilution of the PCs solution. Freeze-drying was attempted to remove both TFA and the solvent. It turned out, however, that PC present before freeze-drying underwent oxidation and no original non-oxidized PC forms could be recovered.The use of b-mercaptoethanol, commonly used as an anti-oxidant in analytical chemistry of thiol compounds22 offered only a limited improvement. The best results were obtained with dithiothreitol (DTT) that allowed the recovery of the original PCs and did not influence the ESI MS signal. The matrix effect of the solution obtained after dissolving the freeze-dried eluate was investigated by spiking a PC standard and examining its signal.It turned out that the slopes of the PC standard calibration curve and a standard addition curve obtained by spiking the sample solution with the PC standards were identical. This indicates the possibility of using external calibration for the determination of PCs in extract purified by reversed-phase chromatography. The results obtained, expressed as the concentration of PC in dry mass of the cell culture extract, are shown in Table 3.Validation of the determination of phytochelatins in plant extracts Since reference materials with certified phytochelatin concentrations in plant samples are not available, the only way to validate the method developed was to compare the results obtained with those obtained by an independent analytical method. Table 3 shows results of the determination of phytochelatins in the analysed samples obtained by reversedphase HPLC with post column derivatization of the sulfhydryl groups of the individual peptides with Ellman’s reagent (5,5Adithiobis- 2-nitrobenzoic acid), and spectrophotometric detection at 410 nm.The measurements were realised in a different laboratory by a different operator. As can be seen from Table 3 the agreement can be considered satisfactory, irrespective of whether water or buffer solution were used for the extraction of phytochelatins. Conclusion Pneumatically assisted electrospray mass spectrometry allows not only the identification of phytochelatins but also their quantitative determination with detection limits in the low ng ml21 range.Tandem mass spectrometry enables the on-line determination of the amino acid sequence of a PC and thus unambiguous identification of the compound determined even in a complex mixture. A purification step using a reversedphase column is necessary prior to ESI MS-MS to separate the bulk of phytochelatins from the matrix salts that would suppress the ionization. The on-line coupling with RP HPLC suffers from definitely poorer detection limits because of the need for TFA in the mobile phase that affects negatively the electrospray ionization.Acknowledgement The work was financed by grant No. 130201 Z0025 (Region Aquitaine-FEDER). Table 3 Determination of phytochelatins in plant cell cultures. Comparison of ESI MS with HPLC with post-column derivatization. The values correspond to the amount of each phytochelatin (in mg) in 1 mg of the powder obtained after purification by reversed-phase HPLC and lyophilization Extracted with water Extracted with TRIS buffer ESI MS analysis Reference method ESI MS analysis Reference method Silene cucubalus— PC2 3.11 ± 0.16 3.6 3.48 ± 0.17 2.8 PC3 10.9 ± 0.5 10.5 6.21 ± 0.31 9.3 PC4 0.68 ± 0.03 3.6 0.69 ± 0.03 3.1 Agrostis tenuis— PC2 1.14 ± 0.16 0.9 0.55 ± 0.03 0.98 PC3 2.69 ± 0.14 2.3 3.3 ± 0.17 2.7 PC4 — 0.16 0.16 ± 0.019 0.22 Rauvolfia serpentina— PC2 0.28 ± 0.02 0.3 0.56 ± 0.03 0.3 PC3 3.3 ± 0.17 2.7 3.4 ± 0.17 2.7 PC4 0.66 ± 0.03 1.3 0.65 ± 0.03 1.3 Fig. 4 Electrospray mass spectra obtained for plant culture extracts after their purification by reversed-phase HPLC and lyophilization. (a) Silene cucubalus; (b) Agrostis tenuis; (c) Rauvolfia serpentina. Peak identification: 1, PC2; 2, PC3; 3, PC4. The signals in Fig. 4(c) correspond to the oxidized forms of the corresponding phytochelatins. Analyst, 1999, 124, 1425–1430 1429References 1 Heavy Metal Stress in Plants—From Molecules to Ecosystem, ed.M. N. V. Prasad and J. Hagemeyer, Springer, Heidelberg, 1999, p. 401. 2 E. Grill, E. L. Winnacker and M. H. Zenk, Methods Enzymol., 1991, 205, 333. 3 W. E. Rauser, Annu. Rev. Biochem., 1990, 59, 61. 4 M. H. Zenk, Gene, 1996, 179, 21. 5 E. Grill, S. L�offler, E. L. Winnacker and M. H. Zenk, Proc. Natl. Acad. Sci., 1989, 86, 6838. 6 W. Gekeler, E. Grill, E. L. Winnacker and M. H. Zenk, Naturforsch., 1989, 44C, 361. 7 E. Grill, W. Gekeler, E. L. Winnacker and M. H. Zenk, FEBS Lett., 1986, 205, 47. 8 S. Klaeck, W. Fliegner and I. Zimmer, Plant Physiol., 1994, 104, 1325. 9 P. Meuwly, P. Thibault, A. L. Schwan and W. E. Rauser, Plant J., 1995, 7, 391. 10 J. Barbas, V. Santhanagopalan, M. Blasczynski, W. R. Ellis Jr. and D. R. Winge, J. Inorg. Biochem., 1992, 48, 95. 11 I. Leopold and D. G�unther, Fresenius’ J. Anal. Chem., 1997, 359, 364. 12 I. Leopold, D. G�unther and D. Neumann, Analusis, 1998, 26, M28. 13 E. Grill, J. Thumann, E. L. Winnacker and M. H. Zenk, Plant Cell Rep., 1998, 7, 375. 14 H. Harmens, P. R. Den Hartog, W. M. Ten Bookum and J. A. C. Verkleij, Plant Physiol., 1993, 103, 1305. 15 A. Tukendorf and W. E. Rauser, Plant Sci., 1990, 70, 155. 16 U. N. Rai, R. D. Tripathi, M. Gupta and P. Chandra, J. Environ. Sci. Health, 1995, A30, 2007. 17 S. Klapheck, W. Fliegner and I. Zimmer, Plant Physiol., 1994, 104, 1325. 18 M. Isobe, D. Uyakul, K. Liu and T. Goto, Agric. Biol. Chem., 1990, 54, 1651. 19 S. A. Hofstadler, R. Bakhtiar and R. D. Smith, J. Chem.Educ., 1996, 73, A82. 20 J. R. Yates, III, A. L. McCormack, A. J. Link, D. Schieltz, J. Eng and L. Hays, Analyst, 1996, 121, 65R. 21 H. Chassaigne and R. Oobi�nski, Analyst, 1998, 113, 131. 22 R. Oobi�nski, H. Chassaigne and J. Szpunar, Talanta, 1998, 46, 271. Paper 9/05163E 1430 Analyst, 1999, 124, 1425&nd

ISSN:0003-2654    

DOI:10.1039/a905163e

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Determination of nicarbazin in feeds using liquid chromatography–electrospray mass spectrometry Andrew Cannavan, Glyn Ball and D. Glenn Kennedy* Veterinary Sciences Division, Department of Agriculture for Northern Ireland, Stoney Road, Stormont, Belfast, UK BT4 3SD. E-mail: glenn.kennedy@dani.gov.uk Received 8th June 1999, Accepted 11th August 1999 A method is presented for the determination of the 4,4A-dinitrocarbanilide component of the coccidiostat nicarbazin in animal feeds. Samples are extracted by shaking with methanol and analysed, without further clean-up, using liquid chromatography–electrospray mass spectrometry. A deuterated form of the analyte is employed as internal standard to improve the repeatability of the method.The method has been validated at levels between 0.1 and 100 mg kg21 with internal standard corrected recoveries between 88 and 101% and RSD values < 8%. Introduction Nicarbazin, a mixture of 4,6-dimethyl-2-hydroxypyrimidine (DMHP) and 4,4A-dinitrocarbanilide (DNC) in a 1+1 molar ratio, is a drug that is used globally in the prevention of coccidiosis in broiler chickens.In the UK, nicarbazin is licensed for use as a feed additive, at concentrations of 100–125 mg kg21, in broiler chickens up to a maximum age of 28 d. The licence requires that treatment must be withdrawn for at least 9 d prior to slaughter. Nicarbazin is not licensed for use in commercial egg-laying chickens in the UK, and consequently eggs should be free from nicarbazin residues.A joint FAO/ WHO Expert Committee recommended the use of DNC alone as the marker residue for nicarbazin, and fixed a maximum residue limit (MRL) of 200 mg kg21 in broiler chicken tissues. However, no MRL has yet been fixed by the EU. The preparation of drug-free withdrawal diets can be difficult because nicarbazin powder is strongly electrostatic. This property can lead to contamination of feed mill production lines after milling a medicated feed, and hence to contamination of supposedly nicarbazin-free feeds with the drug. Feeding of diets containing contamination-level concentrations of nicarbazin can cause unwanted residues in eggs1,2 and broiler chickens.3 Although drug manufacturers have responded to this challenge by the introduction of granular preparations of the drug that are less prone to contaminate feed milling equipment, there are still persistent reports of the occurrence of nicarbazin in eggs and poultry tissues.4 It is therefore important that methods capable of the determination of nicarbazin in feeds at low levels are available to aid the animal feed industry in identification and elimination of contaminating processes. Several groups have reported methods for nicarbazin in feeds using high performance liquid chromatography (HPLC) with UV detection.The DNC and DMHP components were extracted and chromatographed separately by Macy and Loh,5 with quantification down to 25 mg kg21.Extraction and quantification of the DNC component at 0.1 mg kg21 was reported by Hurlbut et al.6 Micro HPLC was used by Draisci et al.7 to measure DNC down to 4 mg kg21. Liquid chromatography–thermospray mass spectrometry has been applied as a confirmatory method for DNC in chicken tissues after determination by HPLC-UV.8,9 A method has also been reported1 for the simultaneous determination of both DNC and DMHP in eggs by liquid chromatography–atmospheric pressure chemical ionisation mass spectrometry (LC-APCI-MS). The present study describes a method for the determination of nicarbazin in feeds using liquid chromatography–electrospray mass spectrometry (LC-ESI-MS). Nicarbazin is extracted with methanol and an aliquot of the extract is evaporated to dryness and redissolved in methanol + water (75 + 25, v/v).DNC is analysed, without any further clean-up, by reversed phase chromatography with detection by ESI-MS. Deuterated (d8) DNC is employed as an internal standard. The method was validated for DNC at concentrations between 0.1 and 100 mg kg21.Experimental Materials All solvents were of HPLC grade and other chemicals were of analytical reagent grade. Distilled or de-ionised water was used throughout. DNC was obtained from Sigma-Aldrich Co Ltd (Gillingham, Dorset, UK). The deuterium-labelled internal standard (d8-DNC) was custom-synthesised by Quchem (The Queen’s University of Belfast, UK).Stock standard solutions (1 mg ml21) of DNC and d8-DNC were prepared by sonication in dimethylacetamide. Dilute standards (10 mg ml21) were prepared by dilution of the stock standards in methanol. All stock and dilute standard solutions were stable for at least 1 month when stored in a refrigerator. Working standards (0.2 mg ml21) were prepared weekly by mixing aliquots of each of the dilute standards and diluting with methanol + water (75 + 25, v/v) and were stored in a refrigerator.Equipment The HPLC system consisted of an L6200A intelligent pump and an AS2000 autosampler (Merck, Poole, Dorset, UK). The LC column was a Luna 5m C18 (2), 250 3 4.6 mm (Phenomenex, Macclesfield, Cheshire, UK). The mobile phase was acetonitrile + water (75 + 25, v/v) containing ammonium acetate (0.05 M). The LC was coupled via an ESI probe to a Platform LC-MS system (Micromass, Altrincham, Cheshire, UK). The source was maintained at 125 °C. Nitrogen was used as the drying gas and ESI nebulising gas at flow rates of 300 and 15 l h21, respectively. Spectra were obtained in negative mode over the range m/z 50–400 with the cone set to 15 V, by flow injection of This journal is © The Royal Society of Chemistry 1999.Analyst, 1999, 124, 1431–1434 1431standard solutions (10 mg ml21) with no column installed. Selected ion monitoring (SIM) was used for sample analysis. The negative ions at m/z 301 and 309 were monitored for DNC and d8-DNC, respectively, with dwell times of 0.5 s for each ion.The mobile phase flow rate was 1 ml min21, with the column effluent split so that approximately 100 ml min21 was introduced into the MS. The run time was 7 min. Sample extraction Meal samples were pulverised, if necessary, in a Knifetec 1095 sample mill (Tecator, Hoganas, Sweden) and aliquots (10 g) weighed into 125 ml polyethylene bottles. Aliquots of known negative meal were weighed out for spiking and to provide blank chromatograms with each batch of samples.Medicated feeds (100–125 mg kg21) Spiked negatives were prepared by adding DNC (1 mg ml21, 1 ml). Internal standard (d8-DNC, 1 mg ml21, 1 ml) was added to all samples, negatives and spiked negatives. The bottles were allowed to stand for approximately 15 min before proceeding with the extraction. Methanol (50 ml) was added and the bottles were capped and shaken on a mechanical shaker for 30 min. The extracts were allowed to settle and aliquots (10 ml) transferred to centrifuge tubes and centrifuged (600g, 4 °C, 10 min).Aliquots of the supernatants (100 ml) were transferred to volumetric flasks and diluted to 10 ml with methanol + water (75 + 25, v/v). Aliquots (200 ml) of these dilute solutions were transferred to microvials for analysis. Contamination levels (0.1, 1.0 and 10.0 mg kg21) Spiked negatives were prepared by adding DNC (10 mg ml21, 1 ml). Internal standard (d8-DNC, 10 mg ml21, 1 ml) was added to all samples, negatives and spiked negatives.The samples were shaken with methanol and centrifuged as described above. Aliquots (1 ml) of the supernatants were evaporated to dryness under nitrogen at 60 °C using a Techne Sample Concentrator (Jencons Scientific), allowed to cool and re-dissolved in methanol + water (75 + 25, v/v, 1 ml). Aliquots (200 ml) were transferred to microvials for analysis. LC-MS analysis The system was equilibrated by pumping mobile phase for 15 min.Peak reproducibility was checked by injecting, typically, three aliquots (25 ml) of the working standard before beginning an analytical sequence. Samples were then injected, with a standard injection after every 3–4 samples. Peak area data were collected for the ions at m/z 301 (DNC) and 309 (d8-DNC). Results were calculated by comparing the peak area ratio of analyte to internal standard in a sample with the ratio in the mean of the standards bracketing the sample. Results and discussion The structures and LC-ESI-MS spectra of DNC and d8-DNC are shown in Fig. 1.With the sample cone voltage set to 15 V, both DNC and d8-DNC gave a strong response in negative mode, with prominent peaks for the [M 2 H]2 ions at m/z 301 and 309, respectively. Peaks were also present at m/z 361 and 369 for DNC and d8-DNC, respectively, probably corresponding to acetate adducts. Our previously published method1 used LC-APCI-MS to determine both the DNC and the DMHP components of nicarbazin in eggs.In the present study, which focussed specifically on the determination of DNC alone in animal feedingstuffs, we chose to use electrospray ionisation because this ionisation mode is more sensitive for this compound. In contrast to the LC-APCI-MS spectra produced by Blanchflower et al.,1 in which the [M 2 H]2 ion was the only prominent ion formed regardless of sample cone voltage, it was possible, using ESI, to initiate collision induced dissociation of the DNC molecule by increasing the sample cone voltage. This apparently anomalous result (more fragments formed by a softer ionisation technique) may be explained by the fact that the earlier study1 was performed on an earlier version of the VG Platform than that used in the current study. We have previously noticed differences in the extent of fragmentation induced by the two machines.Fragment ions were formed at m/z 137 and 107. The ion at m/z 137 results from cleavage of the molecule between the carbonyl and either of the adjoining secondary amino groups.The fragment at m/z 107 may result from the neutral loss of NO•. These ions could be useful for the unequivocal identification of the compound, if required. Fig. 2 shows SIM chromatograms for a DNC standard (equivalent to 1 mg kg21 in meal), a negative meal and a meal fortified at 1 mg kg21 with DNC. Peak symmetry is good and the chromatograms are free from interference at the retention time of DNC.Inter- and intra-assay precision and recovery were assessed by extracting and analysing five replicates of known nicarbazinnegative meal spiked at 0.1, 1, 10 and 100 mg kg21 with DNC on three separate occasions. The results are presented in Table 1. The absolute recovery achieved by the method was approximately 65%. Internal standard corrected recoveries ranged from 88–101% and RSDs were < 8%. The limit of determination, defined as the lowest level at which the method was validated, is 0.1 mg kg21.Quantification to this level should be adequate to identify any significant contamination of meals produced by the animal feed industry. However, the method is sufficiently sensitive to achieve quantitative results at levels at least a factor of 10 lower than this if required. The LC–UV method described by Hurlbut et al.6 is also capable of measuring DNC at 0.1 mg kg21 levels, and provides a possible alternative in laboratories where LC-ESIMS instrumentation is not available.However, that method involves the use of hot solvents and requires an alumina cleanup step, which makes the procedure more time consuming than the method described in this paper. Mass spectrometric detection provides a greater degree of specificity than does UV Fig. 1 Structures and LC-ECI-MS spectra of (A) DNC and (B) d8-DNC. The sample cone voltage was 15 V. 1432 Analyst, 1999, 124, 1431–1434detection, and minimizes the requirement for sample clean-up before analysis.The LC-ESI-MS response was shown to be linear by preparing and analysing a series of eight standard solutions containing DNC over the range 0–2 mg ml21 (equivalent to 0–1000 mg kg21 in meal). Each standard solution also contained d8-DNC at a concentration of 0.2 mg ml21. A standard curve was prepared by plotting the peak area ratio (m/z 301/309) against DNC concentration. The correlation coefficient (r) was 0.9999. The inclusion of a deuterated internal standard contributes significantly to the repeatability of the method since it corrects not only for any minor losses of the analyte during the extraction, but also for changes in the response of the MS during an analytical sequence.Over a period of hours a gradual decrease in abundance is frequently observed, possibly due to contamination of the ESI source or the sample cone. The sample matrix may also suppress, or in some instances enhance, the ionisation of the analyte in the ESI source.These effects can make calculation of results, including recovery values, using absolute areas difficult. Calculations based on the comparison of the ratios of analyte to internal standard in unknown samples with the ratios found in standards overcome these difficulties. In conclusion, the method presented is simple and rapid, and has been used in this laboratory to measure nicarbazin in commercially produced meals, and in meals formulated for experimental purposes, at medicated and contamination levels.Approximately 15 samples, plus controls and negatives, can be extracted in duplicate in one working day. The inclusion of an autosampler in the LC-ESI-MS system facilitates overnight analysis of the extracts. Acknowledgement The authors acknowledge the generous financial support of The Wellcome Trust for the vacation scholarship awarded to Glyn Ball (Award VS98/BEL/007/CH/TG/JG). References 1 W. J. Blanchflower, P.J. Hughes and D. G. Kennedy, J. Assoc. Off. Anal. Chem., 1997, 80, 1177. 2 A. Cannavan, G. Ball and D. G. Kennedy, Food Addit. Contam., 1999, submitted. 3 A. Cannavan and D. G. Kennedy, Food Addit. Contam., 1999, submitted. 4 Anonymous, in MAVIS, Veterinary Medicines Directorate, Addlestone, Surrey, UK, 29th edn., 1999, pp. 9–18. 5 T. D. Macy and A. Loh, J. Assoc. Off. Anal. Chem., 1984, 67, 1115. 6 J. A. Hurlbut, C. T. Nightengale and R. G. Burkepile, J. Assoc. Off. Anal.Chem., 1985, 68, 596. Fig. 2 SIM chromatograms for (A) a DNC standard at a concentration of 1.0 mg kg21, (B) a negative meal sample and (C) a negative meal sample fortified with DNC at 1.0 mg kg21. The top row shows traces at m/z 301, normalised to 100% = 5.5 e,5 for the [M 2 H]2 ion of DNC. The bottom row shows traces at m/z 309, normalised to 100% = 3.8 e,5 for the [M 2 H]2 ion of d8-DNC. Table 1 Inter- and intra-assay reproducibility and recovery for meal spiked with DNC at 0.1, 1, 10 and 100 mg kg21. Results are calculated using the internal standard Day 1 Day 2 Day 3 Overall 0.1 mg kg21— Mean/mg kg21 0.10 0.09 0.09 0.09 s/mg kg21 0.005 0.004 0.005 0.006 RSD (%) 5.7 5.1 5.8 6.4 Mean recovery (%) 96.0 88.0 94.0 92.7 n 5 5 5 15 1.0 mg kg21— Mean/mg kg21 0.99 0.93 1.01 0.98 s/mg kg21 0.031 0.014 0.021 0.043 RSD (%) 3.2 1.5 2.1 4.4 Mean recovery (%) 99.3 92.7 101.0 97.7 n 5 5 5 15 10.0 mg kg21— Mean/mg kg21 8.83 9.44 10.06 9.44 s/mg kg21 0.323 0.718 0.472 0.714 RSD (%) 3.7 7.6 4.7 7.6 Mean recovery (%) 88.3 94.4 100.6 94.4 n 5 5 5 15 100 mg kg21— Mean/mg kg21 91.1 93.4 95.4 93.3 s/mg kg21 0.792 1.199 1.178 2.043 RSD (%) 0.9 1.3 1.2 2.2 Mean recovery (%) 91.1 93.4 95.4 93.3 n 5 5 5 15 Analyst, 1999, 124, 1431–1434 14337 R. Draisci, L. Lucentini, P. Boria and C. Lucarelli, J. Chromatogr. A, 1995, 697, 407. 8 J. L. Lewis, T. D. Macy and D. A. Garteiz, J. Assoc. Off. Anal. Chem., 1989, 72, 577. 9 M. G. Leadbetter and J. E. Matusik, J. Assoc. Off. Anal. Chem., 1993, 76, 420. Paper 9/04557K 1434 Analyst, 1999, 124, 1431–1434

ISSN:0003-2654    

DOI:10.1039/a904557k

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Fractionation of soluble selenium compounds from fish using size-exclusion chromatography with on-line detection by inductively coupled plasma mass spectrometry Gunilla Önning*a and Ingvar A. Bergdahlbc a Biomedical Nutrition, Center for Chemistry and Chemical Engineering, Lund University, PO Box 124, SE-221 00 Lund, Sweden. E-mail: Gunilla.Onning@kc.lu.se b Department of Occupational and Environmental Medicine, Lund University, Lund, Sweden c Occupational Medicine, Department of Public Health and Clinical Medicine, Umeå University, Umeå, Sweden Received 19th May 1999, Accepted 16th August 1999 Fish accumulate significant amounts of selenium and are an important dietary source of this element. Some studies have however indicated a low bioavailability of the selenium from fish.Since little is known of the selenium forms in fish, we have studied soluble selenium compounds in fish species, and compared different techniques for fractionation of selenocompounds (size-exclusion chromatography, ultrafiltration, and precipitation with trichloroacetic acid).The size-exclusion column (Superdex 200 HR 10/30) was coupled on-line to inductively coupled plasma mass spectrometry (ICP-MS). The limit of detection was 0.20 mg l21 and the selenium response was linear in the investigated concentration range of 0–20 mg l21 (r2 = 0.98). For plaice 47% of the selenium was extractable while the extraction efficiency for cod was 23%. The fish extracts were injected onto the column four times each and the variation in the quantitative data for different selenium-containing fractions between the runs was small (RSD < 10%).The recovery of selenium in the chromatographic step was about 70%, indicating some interaction between the fish extracts and the column material. Ultrafiltration using a membrane with a cut-off at Mr 10 000 gave results similar to the size-exclusion fractionation, for cod about 20% of the soluble selenium had a Mr < 10 000 and the corresponding value for plaice was 69%.Removal of high-molecular-weight compounds from the sample by trichloroacetic acid precipitation showed a similar proportion of low-molecular-weight compounds for plaice (77%), while the obtained value for cod was higher (38%) compared with the other techniques. Introduction Selenium is an essential element and it forms part of at least two groups of enzymes in the human body: glutathione peroxidases and iodothyronine deiodinases.1 The selenium concentration in soil varies between different regions, leading to varied selenium concentrations in plants.Therefore, the intake of selenium is low in certain countries or regions. For example, the intake in Sweden is about 25% below the recommended intake for both women and men,2 but compared to other European countries the selenium status is in the middle of the observed range.3 A high selenium intake is believed to reduce the risk for some forms of cancer and for cardiovascular disease.For example, selenium supplementation to patients with a history of carcinomas of the skin gave a reduced incidence of other kinds of cancer.4 Several forms of selenium exist in the diet. In plant and animal foods selenium predominantly exists in organic forms such as the selenoamino acids selenocysteine and selenomethionine, which are built into proteins. In the glutathione peroxidases and iodothyronine deiodinases, the selenocysteine is situated in the active site, while selenomethionine seems to be incorporated unspecifically in proteins at methionine positions.5 It is not known if the inorganic forms, selenite, selenate and the metal selenides, occur in the diet, but the former compounds are found in water and used in some diet supplements.The bioavailability of the different selenocompounds varies. In a long-term study on New Zealand women,6 supplementation with selenomethionine increased the blood selenium levels more effectively than selenate, but the glutathione peroxidase activities plateaued at similar levels.Selenomethionine is retained in tissue proteins to a greater extent than selenocysteine and the inorganic forms, and there is probably a restricted release of the selenium from selenomethionine and thus a limited availability for incorporation into functional proteins.5 Fish is rich in selenium and the consumption of fish is correlated to plasma concentrations of selenium,7–9 glutathione peroxidase and selenoprotein P.8 However, some studies have indicated a low availability of the selenium from fish.Thorngren and Åkesson,10 increased the fish intake for human subjects with 150–200 g per day for 6 or 11 weeks but the mean increase in plasma selenium was modest, 13%. Meltzer et al.11 found no increase in the plasma selenium when the diet was supplemented with fish, but wheat gave a raised level. Little is, however, known about the forms of selenium in different fish species.The selenium seems to be associated with free amino acids or proteins and is probably not present as inorganic species.12 Analyses of fish have also shown that the distribution of low- and high-molecular-weight compounds varies in different species.13 Plaice seems to have a high amount of lowmolecular- weight compounds that so far are unidentified. More information on different selenium forms is needed for evaluation of the bioavailability of selenium from fish.Several methods can be used for selenium speciation. The selenocompounds can be extracted from the fish using water, methanol or enzymes, and the extracts can then be analysed with different chromatographic techniques. Size-exclusion and ionexchange chromatography have been used to speciate selenocompounds from fish.13–15 The selenium in the eluates can be analysed with different techniques. The selenium is often determined with hydride generation atomic absorption spectrometry (HGAAS).The method is generally time consuming, since the samples have to be digested with acids, and the selenium reduced before analyses. A more rapid method is graphite furnace atomic absorption spectrometry (GFAAS) with Zeeman background correction.13 However, the selenium concentration in chromatographic eluates is often low and in some samples the selenium level can be close to the detection limit. To improve the detection, size-exclusion chromatography This journal is © The Royal Society of Chemistry 1999.Analyst, 1999, 124, 1435–1438 1435may also be coupled on-line to inductively coupled plasma mass spectrometry (ICP-MS),16–18 and ICP-MS has also been used with ion-exchange chromatography in the speciation of selenium. 14,19 In this study we use size-exclusion chromatography to separate water-soluble selenium-containing macromolecules from fish, and determine the selenium on-line with ICP-MS.The method is compared with two alternatives to chromatographic separation: ultrafiltration and precipitation of highmolecular- weight proteins with trichloroacetic acid (TCA). Experimental Reagents Distilled water was prepared by a Milli-Q system (Millipore Corporation, Bedford, MA, USA). Tris(hydroxymethyl)-aminomethane, ammonium acetate, TCA, methanol, hydrochloric acid and perchloric acid were obtained from E. Merck (Darmstadt, Germany). The sulfuric acid was purchased from Merck Ltd.(Poole, Dorset, UK) and the selenium standards were prepared from a 1000 mg l21 SpectrosoL solution (Merck Ltd.). Sample preparation Cod and plaice homogenate were obtained frozen from DLO Netherlands Institute for Fisheries Research, Ijmuiden, The Netherlands. The water-soluble components were extracted with buffer (1 + 1; 20 mM TRIS acetate, pH 7.5, 0.15 M ammonium acetate) using a Polytron homogeniser. The soluble components were thereafter separated by centrifugation at 4 °C for 30 min (Beckman J2-21 high-speed centrifuge, 20 000g).The selenocompounds in the supernatant were fractionated with size-exclusion chromatography (see below). Aliquots of the supernatant were also fractionated by TCA precipitation and ultrafiltration, respectively. For the TCA precipitation 1 ml of 5% TCA was added to 5 ml supernatant, mixed for 30 min and the TCA–supernatant was thereafter recovered by centrifugation. In the ultrafiltration procedure 3 ml of supernatant was added to a centrifugal concentrator with a membrane cut-off at Mr 10 000 ( MIKROSEP, Filtron Technology Corporation, Northborough, MA, USA).The supernatant was centrifuged (Beckman CRP Centrifuge, 4 °C, 2800g) until half of the sample had passed through the membrane. Determination of selenium The supernatant, TCA–supernatant and the ultrafiltration fractions were digested with nitric and perchloric acids and the selenium in the digested samples was reduced by heating in 7 M hydrochloric acid.The selenium concentration was thereafter determined with a flow injection analysis selenium hydride system connected to GFAAS equipment (AAnalyst 800, Perkin Elmer, Überlingen, Germany). A calibration curve was made using selenium standard concentrations of 0, 2, 4, 6, 8, and 10 mg l-1 in 10% hydrochloric acid. Size-exclusion chromatography (SEC) The fish supernatant was filtered (Millex-LCR filter unit, 0.5 mm, Millipore) and injected onto a Superdex 200 HR10/30 column (Pharmacia Biotech, Pharmacia, Uppsala, Sweden) via a 500 ml loop and a Rheodyne valve.The column separates in the molecular weight range 10 000–600 000 and was calibrated with a marker kit from Sigma (MW GF-200, St. Louis, MO, USA) plus cyanocobalamin. The total column volume was 24 ml and the eluent was the same buffer as was used for the fish extraction (flow rate: 0.75 ml min-1). The column outlet was connected via a UV spectrometer (at 280 nm) to the ICP-MS.Inductively coupled plasma mass spectrometry (ICP-MS) The ICP-MS instrument was a PlasmaQuad 2+ (Fisons Elemental, Winsford, Cheshire, UK) fitted with a water-cooled spray chamber (+5 °C) and a V-groove nebuliser. The instrument settings were first optimised for indium, and then adjusted for selenium. Typical settings were: forward power, 1350W; reflected power, < 5W; nebuliser gas, 1.0 l min21; auxiliary gas, 0.7 l min21; coolant gas, 13 l min21.The peak-jumping mode under the software TRVision (Fisons Elemental) was used for acquisition. Obtained data were exported to a spreadsheet software (Microsoft Excel) for graphic presentation and quantification. For quantification freshly prepared selenium standard solutions (0, 2.5, 5, 10 mg l21) were injected via a loop into a flow injection valve mounted after the column. The loop mounted after the column (used for standards) had the same size as the loop mounted before the column (used for samples).This was certified by injection of a 5 mg l21 standard four times through the different loops. No difference between the obtained data was found (sample/standard loop ratio: 0.99–1.01). A standard (5 mg l21) was also injected between each run to be able to correct for fluctuations in the ICP-MS response. Seleno-l-methionine (Se-Met, Sigma, St. Louis, MO, USA) was injected onto the column to test the recovery and to evaluate if the samples contained this compound.Calculations of recoveries were based on the ratio between the amount of selenium determined from the chromatogram, and the amount of selenium in the solution entered into the system. Results and discussion SEC-ICP-MS Initially, chromatograms obtained with (3%, v/v) and without methanol in the elution buffer were compared, monitoring at m/z 77 and m/z 82. No significant difference in the sensitivity for selenium was found when a plaice supernatant was applied (Fig. 1). Peaks detected in only one of the chromatograms were designated as non-selenium peaks. One non-selenium peak at m/z 77 (presumably corresponding to 40Ar37Cl) appeared at 30 min but it was almost completely absent when methanol was used. Another non-selenium peak appeared at m/z 82 at about 40 min. The abundance of selenium is somewhat higher for m/z 82 than for m/z 77, and since some non-selenium peaks dis- Fig. 1 SEC-ICP-MS chromatogram of plaice supernatant. (a) Without methanol, (b) with 3% methanol in the elution buffer (dark: m/z 82; grey: m/z 77). 1436 Analyst, 1999, 124, 1435–1438appeared when methanol was used, we chose to use methanol in the elution buffer, and to monitor m/z 82 for quantitative determinations. Using this method the limit of detection (calculated as three times the standard deviation of the analytical blank, n = 5) was 0.20 mg l21 Se. The selenium response was linear in the investigated concentration range 0–20 mg l21 (r2 = 0.9829).Earlier, Kölbl et al.20 used methanol in the elution buffer and, contrary to our results, found that it improved the intensity of the ICP-MS signal; a concentration of 2% gave a maximum intensity while higher concentrations (10%) depressed the selenium signals to 10–40%. Larsen and Stürup,21 also used 3% methanol and reported an enhancement of the ICP-MS signal. A large amount of selenium was extracted from plaice, 104 mg l21 supernatant, which was equivalent to 47% of the total selenium content of the fish sample.The extraction efficiency for cod was 23% (36 mg l21). The 82Se size-exclusion chromatograms for cod and plaice supernatants are shown in Fig. 2 and the quantification of the chromatographic peaks in Table 1. Each sample was injected five times but the first run was disregarded due to a large baseline drift. For both fish species two clear peaks appeared (fractions B and D), and the variations in the quantitative data for these fractions between runs were small (RSD 3–7%).The smaller, less defined, peaks (fractions A and C) varied more (RSD 7–11%), while the total selenium content recovered gave a RSD of less than 5%. The recovery of selenium in the chromatographic step was 70% for cod and 68% for plaice. Experiments were also made by adding selenomethionine (50 mg l21, corresponding to 20 mg l21 Se) to cod supernatant. Applying pure selenomethionine to the column gave a recovery of 113% (n = 2, 113 and 114%, respectively) but when it was applied together with cod supernatant the recovery was decreased to 75%.Thus, there seems to be some interaction between the fish extracts and the column material. During a day of analysis when ten samples were applied an increased backpressure in the column was also observed, indicating accumulation of proteins in the column. Incomplete recoveries from gel chromatography have been observed before. For example, Owen et al.17 analysed intestinal extracts for aluminium with SEC-ICP-MS and got recoveries of 14–21%, and Bergdahl et al.18 got concentration dependent recoveries of lead from erythrocytes.Most of the selenium (76%) in plaice supernatant appeared in fraction D. Its peak maximum is outside the calibrated molecular range, but extending the calibration curve gives an apparent molecular weight of 1100. For cod, fraction B with a peak maximum at an apparent molecular weight of 57 000 contained most of the selenium (66%).The peak maximum for selenomethionine appeared a little later than for fraction D (0.3–0.4 min). Analysis of plaice supernatant diluted 1 + 2 and 1 + 4 was made in order to find if the concentration of other compounds in the sample (salts, proteins, lipids) influenced the detection. No such influence could be observed: the amount of selenium in fraction B was 103–107% in the diluted samples compared with the concentrated sample and the corresponding values for fractions C and D were 100–112% and 102–104%, respectively.Fraction A was too small to permit an evaluation. Since selenomethionine is unspecifically bound into proteins, a wide range of proteins can contain selenium. In cod, the selenium chromatogram resembled the A280 chromatogram to a relatively large extent, suggesting that much of the selenium is unspecifically bound in proteins. In plaice, on the other hand, there was a large difference between the selenium and the A280 chromatogram, indicating a specific occurrence of selenium in the D peak.Shen et al.15 have also analysed cod and plaice, using SEC-HGAAS. Similar amounts of selenium were extractable in comparison with the present study, 26% in cod and 44% in plaice. The size distribution of selenocompounds was also similar especially for plaice where most of the selenium was found in two peaks eluting late in the low-molecular-weight range. A lower amount of selenium was extractable in another study,13 19% for cod and 29% for plaice, but the size distribution was similar compared with the present study.They used SEC-GFAAS-Zeeman and over 70% of the selenium was found in high-molecular-weight compounds for cod, while over 70% was found in low-molecular-weight compounds for plaice. We used frozen fish in this study and there could be important textural changes in fish during freeze-storage due to increased salt concentrations and decreased pH values. In the study by Åkesson and Srikumar,13 the distribution of selenium was, however, similar for fresh and frozen fish (cod, herring).The ICP-MS technique has previously mostly been used in combination with ion-exchange chromatography. Most of the studies report qualitative data and rarely are quantitative data presented. Selenocompounds previously separated using this method are mainly selenocystine, selenomethionine, selenite and selenate but other compounds have also been detected. 14,19,22,23 Crews et al.14 studied cod that was cooked and digested with gastrointestinal enzymes and in total 61% of the selenium was extracted. When analysing the extract with ionexchange chromatography-ICP-MS two peaks were seen: one was probably selenite but the larger peak was not identified. Fig. 2 SEC-ICP-MS chromatogram of cod and plaice supernatant. The m/z ratio 82 was monitored. The thin line shows the UV absorbance at 280 nm. The first peak (at 1–2 min) is a calibration standard containing 5 mg l-1 Se. The peaks A–D are referred to in Table 1 and in the text.Table 1 The amount of selenium (mg l-1) in size-exclusion chromatography fractions for cod and plaice supernatant (mean ± s, n = 4). The total volume of the column is at 30–31 min. See Fig. 2 for identification of the fractions Mr: Retention time: Fraction A > 250 000 11–17 min Fraction B 15 000–250 000 17–23 min Fraction C 3400–15 000 23–26 min Fraction D < 3400 26–31 min Sum (recovery) Cod 2.1 ± 0.2 16.9 ± 0.6 1.7 ± 0.1 5.9 ± 0.1 25.6 ± 0.8 (70%) Plaice 2.8 ± 0.3 10.4 ± 0.8 4.2 ± 0.5 53.7 ± 2.6 71.0 ± 3.7 (68%) Analyst, 1999, 124, 1435–1438 1437Comparison of size-exclusion fractionation with ultrafiltration and TCA precipitation Ultrafiltration using a membrane with a cut-off at Mr 10 000 gave results for the proportion of low-molecular-weight compounds close to the size-exclusion fractionation, for cod about 20% of the soluble selenium had a Mr < 10 000 and the corresponding value for plaice was 70% (Table 2).The recoveries were 100 and 114%, respectively. A similar value for soluble compounds after TCA precipitation was obtained for plaice, while the value obtained for cod was higher (38%) compared with the other techniques. If a technique is to be used to separate compounds that will later be further characterised it should be mild. In this context the ultrafiltration technique is preferable since the TCA precipitation leads to a low pH.TCA precipitation is generally used to precipitate large proteins but it is not clear which molecules remain soluble, since both the size and the hydrophobicity of proteins are important.24 Peptides with less than seven amino acid residues seem to be soluble even at high TCA concentrations (12%), but for larger peptides the solubility is dependent on the hydrophobicity, since TCA decreases their hydration potential. This study shows large differences between fish species as regards selenium-containing proteins. It is not known if this affects the bioavailability of the selenium for humans but different fish species have been studied in rats.25 The efficiency of different dietary selenium sources to restore glutathione peroxidase activity after depletion was investigated.The relative liver glutathione peroxidase activity compared with control rats (100%) was found to be higher after a flounder diet (106%) than after a tuna diet (101%).Both fish species were more effective in restoring the selenium compared with the selenium from other sources (beef, chicken, veal, lamb, pork). There are some indications also for humans that fish is a good selenium source, for example plasma selenium for Latvian men with a high fish intake (21–50 fish meals per month) was 81% higher than in those with the lowest fish intake.8 In Sweden, the median fish intake is much lower compared with the high fish consumers in Latvia and the most commonly consumed species is cod (1–3 meals per month).2 In the present study raw fish was investigated.Studies on the effect of food processing and gastrointestinal digestion on the selenocompounds in fish are planned, and the relation of selenium speciation to its bioavailability will be further evaluated. Conclusions The ICP-MS method gave reproducible results for selenocompounds of different size in fish extracts detected on-line coupled to a size-exclusion column. The advantage with the ICP-MS technique compared with the GFAAS or HGAAS technique is that the analyses can be performed on-line, avoiding loss of chromatographic resolution, making it less time-consuming and attaining low limits of detection.Some selenium was, however, retained on the column. Also ultrafiltration may be used to study the partition of selenium between low- and high-molecular-weight compounds. The resulting information is then of course less detailed than from a chromatogram.The TCA precipitation gave results that are more difficult to interpret, since it is somewhat unclear which compounds are precipitated. Financial support by grants from the FAIR programme (project CT95-0077) of the European Union is acknowledged by G.Ö. The study was also supported by the Påhlsson Foundation and the J. Andersson Foundation. The authors thank other partners in the FAIR project and Dr Andrejs Schütz and Dr Björn Åkesson for their encouraging support and for valuable discussions.References 1 L. Johnsson, B. Åkesson and J. Alexander, Report 4711, Swedish Environmental Protection Agency, Stockholm, 1997. 2 W. Becker, Vår Föda, 1999, 51, 24. 3 W. van Dokkum, Nutr. Res. Rev., 1995, 8, 271. 4 L. C. Clark, G. F. Combs, B. W. Turnbull, E. H. Slate, D. K. Chalker, J. Chow, L. S. Davis, R. A. Glover, G. F. Graham, E. G. Gross, A. Krongrad, J. L. Lesher, H. K. Park, B. B. Sanders, C. L. Smith and J.R. Taylor, JAMA, 1996, 276, 1957. 5 C. D. Thomson, Analyst, 1998, 23, 827. 6 C. D. Thomson, M. F. Robinson, J. A. Butler and P. D. Whanger, Br. J. Nutr., 1993, 69, 577. 7 S. Bergmann, V. Neumeister, R. Siekmeier, C. Mix, V. Wahrburg and W. Jaross, Toxicol. Lett., 1998, 96–97, 181. 8 L. Hagmar, M. Persson-Moschos, B. Åkesson and A. Schütz, Eur. J. Clin. Nutr., 1998, 52, 796. 9 D. Luoma, Int. J. Circumpolar Health, 1998, 57, 109. 10 M. Thorngren and B. Åkesson, Int.J. Vit. Nutr. Res., 1987, 57, 429. 11 H. M. Meltzer, K. Bobow, I. T. Paulsen, H. H. Mundal, G. Norheim and H. Holm, Biol. Trace Elem. Res., 1993, 36, 229. 12 W. Maher, S. Baldwin, M. Deaker and M. Irving, Appl. Organomet. Chem., 1992, 6, 103. 13 B. Åkesson and T. S. Srikumar, Food Chem., 1994, 51, 45. 14 H. M. Crews, P. A. Clarke, D. J. Lewis, L. M. Owen, P. R. Strutt and A. Izquierdo, J. Anal. At. Spectrom., 1996, 11, 1177. 15 L. H. Shen, M. Hoek-van Nieuwenhuizen and J. B.Luten, in Seafood from Producer to Consumer, Integrated Approach to Quality, ed. J. B. Luten, T. Børresen and J. Oehlenschläger, Elsevier, Amsterdam, 1997, pp. 653–663. 16 B. Gercken and R. M. Barnes, Anal. Chem., 1991, 63, 283. 17 L. M. W. Owen, H. M. Crews and R. C. Massey, Analyst, 1995, 120, 705. 18 I. A. Bergdahl, A. Schütz and A. Grubb, J. Anal. At. Spectrom., 1996, 11, 735. 19 G. A. Pedersen and E. H. Larsen, Fresenius’ J. Anal. Chem., 1997, 358, 591. 20 G. Kölbl, M. Krachler, K. Kalcher and K. J. Irgolic, in Proceedings of the Fifth International Symposium on Uses of Selenium and Tellurium, ed. S. C. Carapella, J. E. Oldfield and Y. Palmieri, STDA, Grimbergen, Belgium, 1994, pp. 291–300. 21 E. H. Larsen and S. Stürup, J. Anal. At. Spectrom., 1994, 9, 1099. 22 S. M. Bird, G. Honghong, P. C. Uden, J. F. Tyson, E. Block and E. J. Denoyer, J. Chromatogr. A, 1997, 789, 349. 23 H. Emteborg, G. Bordin and A.R. Rodriguez, Analyst, 1998, 123, 245. 24 H. Y. Wen, R. L. Davis, B. Shi, J. J. Chen, L. Chen, M. Boylen and J. E. Spallholz, Biol. Trace Elem. Res., 1997, 58, 43. 25 M. Yvon, C. Chabanet and J-P. Pélessier, Int. J. Peptide Protein Res., 1989, 34, 166. Paper 9/04024B Table 2 Comparison of the distribution of selenium in cod and plaice supernatant after ultrafiltration, TCA precipitation and size-exclusion fractionation, respectively (mean ± s) Cod Plaice n Ultrafiltration— % Selenium, Mr, < 10 000 (a) Of soluble selenium 20 ± 2.5 69 ± 3.5 4 (b) Of total selenium 5 ± 0.6 32 ± 1.7 4 TCA precipitation— % TCA-soluble selenium (a) Of soluble selenium 38 ± 0.0 77 ± 2.7 5 (b) Of total selenium 9 ± 0.1 36 ± 1.2 5 Size-exclusion fractionation— % Selenium, Mr, < 10 000 (a) Of soluble selenium 27 ± 0.3 80 ± 0.9 4 (b) Of total selenium 6 ± 0.1 38 ± 0.3 4 1438 Analyst, 1999, 124, 1435–1438

ISSN:0003-2654    

DOI:10.1039/a904024b

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

The use of nitrite ion in the chromatographic determination of large amounts of hypochlorite ion and of traces of chlorite and chlorate ions A. Gallina, P. Pastore* and F. Magno Department of Inorganic, Metallorganic and Analytical Chemistry, University of Padova, via Marzolo 1, I-35131, Padova, Italy Received 8th June 1999, Accepted 16th July 1999 An ion chromatographic procedure was developed for the one-run determination of ClO2, ClO22 and ClO32 present at very different concentration levels in the same sample.The method is based on the quantitative and fast reduction of HClO in slightly acidic medium by added sodium nitrite, followed by the ion chromatographic determination of ClO22, ClO32 and the generated NO32. The possibility of amperometric titration of the ‘free chlorine’ with a standard solution of NaNO2 was also ascertained. The performance characteristics of the analysis were defined. Chlorine and hypochlorite ion find wide use in different applications such as washing, bleaching and drinking water disinfection.1 Since chlorite and chlorate, mainly originating from concentrated aged aqueous solutions of hypochlorite,2 are suspected to be health hazards, in addition to the methods for the analysis of free and combined chlorine,3 a direct and selective procedure for determining ClO2, ClO22 and ClO32 present in the same solution at very different concentration levels is required.The difficulties arising from the circumstance that these ions present similar reactivities4 can be successfully overcome by resorting to an ion chromatographic method able to detect minor amounts of ClO22 and ClO32 in the presence of a large excess of ClO2.5 As the reported method5 suffers from the drawbacks of drift of the baseline of the chromatographic response, due to the excess of ethylenediamine added to mask ClO2, and of requiring separate tests, iodimetric titrations and chromatographic analysis, to determine ClO2, ClO22 and ClO32, this paper describes a new procedure which advantageously substitutes the nitrite ion for ethylenediamine as the masking agent.A further goal was also to prove that under appropriate experimental conditions the redox reaction between ClO2 and NO22 is quantitative and fast, so that ClO2 can be, in turn, indirectly quantified by measuring the stoichiometric amount of the generated nitrate ion. To validate this analytical approach, the reaction between ClO2 and NO22 has to be monitored by some complementary techniques, such as spectrophotometric and voltammetric methods, and some significant analytical parameters of the chromatographic analysis have to be quantified.Finally, the potential of the nitrite ion as a selective titrant of only the free chlorine was evaluated. Experimental Reagents For the preparation of solutions distilled water obtained from a Milli-Q system (Millipore, Bedford, MA, USA), with no appreciable chlorine demand, was always used.Standard solutions of chlorite ion were prepared from repeatedly recrystallised NaClO2 6 (BDH, Poole, Dorset, UK). The resulting purity, checked by an iodimetric titration carried out at pH 1.3 for H2SO4, was > 99%. Standard solutions of chlorate ion were prepared from NaClO3 (Fluka, Buchs, Switzerland) of purity > 99.5%. Commercially available solutions of ClO2 were standardised for the total chlorine content by the iodimetric method working at pH Å 4 with acetic acid to avoid interference of ClO22 and ClO32.7 Standard solutions of NO22 were prepared either by dilution of a 200 mg l21 solution supplied by Alltech (Deerfield, IL, USA) or by standardisation of nitrite solutions with the MnO42–I2 method.8 Standard solutions of nitrate ions were prepared by dissolving in water weighed amounts of 99.9% purity KNO3 (Merck, Darmstadt, Germany), previously dried at 120 °C.All other reagents used were of analytical-reagent grade.Apparatus Solutions containing NO22, NO32 (after removal of ClO2 by NO22), ClO22 and ClO32, alone and in mixtures, were analysed by ion chromatography (IC) with suppressed conductivity detection. A Dionex (Sunnyvale, CA, USA) Model 2000i S/P ion chromatograph, equipped with an ASRS-II selfregenerating suppressor, able to produce the H3O+ regenerating ion by electrolysis of the eluent, was used together with an ED40 electrochemical detector. The column, obtained from Dionex, was a 250 3 4.6 mm id IonPac AS9-SC in conjunction with a 50 3 4.6 mm id IonPac AG9-SC guard column. The chromatograms were recorded with a Dionex PeakNet acquisition system. A 20 ml sample loop was used for all injections.For voltammetric measurements and amperometric titrations, a laboratory-made potentiostat equipped with an EG&G Princeton Applied Research (Princeton, NJ, USA) Model 175 function generator was used. The spectrophotometer used was a Perkin- Elmer (Norwalk, CT, USA) Lambda 5 instrument.Procedures (a) Direct amperometric titration of free chlorine with NO22. Even if straightforward thermodynamic considerations indicate that the redox reaction between ClO2 and NO22 has to be quantitative:9 E°(NO32/HNO2) = 0.94 V vs NHE Ka(HNO2) = 1023.3 E°(HClO/Cl2) = 1.5 V vs NHE Ka(HClO) = 1027.3 This journal is © The Royal Society of Chemistry 1999. Analyst, 1999, 124, 1439–1442 1439amperometric titrations of ClO2-containing solutions with NO22 were carried out at different pH values with a double purpose: (i) to check the real completeness and the speed of the reaction ClO2 + NO22 ? Cl2 + NO32 under the different experimental conditions adopted and (ii) to verify the selectivity of NO22 as a redox reagent only towards the free chlorine. For these purposes solutions containing known amounts of ClO2 were titrated with a standard solution of NaNO2 at pH 4.75 and 9, respectively, following the decrease in the limiting cathodic current of HClO–ClO2 recorded at a glassy carbon electrode polarised at E = 21.0 V vs.SCE. After completion of the titration, the resulting solution was analysed by IC to quantify the stoichiometric amount of NO32 produced, the excess of NO22 and possibly ClO22 and ClO32. Analogous experiments were tried in which the hypochlorite content was monitored spectrophotometrically.10 In a different series of runs, voltammetric titrations, followed by IC analyses, were performed to test the selectivity of NO22 towards free and combined chlorine.Finally, separate tests were carried out to verify the complete absence of reaction between NO22 and ClO22 or ClO32. (b) Ion chromatographic determination of ClO22, ClO32, NO32 and NO22. The IC determination of ClO22 and ClO32 in the presence of a large excess of hypochlorite ion presents two main difficulties, the interference due to the large concentration of Cl2 and the requirement to remove ClO2 prior to injection.5 The first difficulty can be overcome either by using Ag+ cartridges or by optimising the IC experimental conditions, and for the second the use of a masking agent such as ethylenediamine has been suggested.5 This procedure, effective in removing the excess of ClO2, produces, however, some irregularities in the chromatographic baseline with a consequent deterioration of the repeatability of the results and of the detection limit.Since NO22 does not react at all with ClO22 and ClO32 but is quantitatively oxidised to NO32 by ClO2, it offers the double opportunity to remove ClO2 and to allow the subsequent IC determination of ClO2 as NO32, after suitable dilution.To quantify ClO22, ClO32 and NO32 present in the solution, the corresponding calibration curves were prepared (see later). The chromatographic conditions adopted are reported in Table 1. The coincidence between the results obtained with amperometric and IC analysis proves the absence of nitrate ion in the original test solution.It must be remarked that, in our experience, no commercially available product ever contained detectable nitrate amounts. Results and discussion Determination of ClO2, ClO22 and ClO32 Preliminary investigations indicated that the reaction rate between ClO2 and NO22 is heavily dependent on the pH of the solution. In fact, both voltammetric and spectrophotometric measurements carried out at pH 9 on NO22– and ClO2– containing solutions were not stable with time but indicated the occurrence of a slow reaction between the two reagents.Only after about 60 min did the signals become stable (Fig. 1). In contrast, in fairly acidic medium (pH 4.75 for acetate buffer), the reaction rate was very fast so that this condition was always chosen for any determination. Fig. 2 and 3 show several voltammetric curves obtained during the titrations of an aged HClO solution with NO22 in the absence and presence of NH3, respectively.It is evident that in the former case HClO reacts quantitatively with the added NO22, allowing a correct titration (see also the inset of Fig. 2), whereas in the latter case the reaction proceeds up to the reduction of the HClO excess and is then hindered by the reduced oxidising power of the ‘combined chlorine’ produced by the reaction between ClO2 and NH3. Analogous results were found by IC analyses which correspondingly indicated in the former case the completeness of the Table 1 Ion chromatographic experimental conditions Guard/separator columns Dionex Ionpac AG9-SC/Ionpac AS9-SC Detection Dionex ED40 conductimetric detector Suppressor Dionex ASRS-II Eluent 1.7 mmol l21 NaHCO3 Eluent flow rate 2 ml min21 Injection loop volume 20 ml System back-pressure 1250–1350 psi Fig. 1 Time dependence of the UV absorption of a mixture of NaClO and NaNO2 at pH 9. (a) 7 3 1024 mol l21 NaClO alone (1 ml); (b) spectrum recorded immediately after the addition of 0.2 ml of 1.4 3 1023 mol l21 NaNO2 ; (c), (d) and (e) as (b) after 5, 10 and 60 min, respectively.Fig. 2 Voltammetric curves recorded during the titration of NaClO with standard NaNO2 solution at pH 4.7 for acetate buffer. Experimental conditions: initial volume, 40.2 ml; NaClO concentration, 2.7 3 1023 mol l21; NaNO2 concentration, 0.1 mol l21; additions of NaNO2 in aliquots of 70 ml; working electrode, rotating glassy carbon; counter electrode, Pt foil; reference electrode, SCE; scan rate, 10 mV s21; rotating speed, 105 rad s21.(a) No NaNO2 added; (b) background current. Inset: titration curve monitored at –1.0 V vs. SCE. 1440 Analyst, 1999, 124, 1439–1442redox reaction, in the absence of the formation of chloroamine, and, in the latter case, the partial failure of the formation of NO32 in the presence of NH3. The results indicate that NO22 has to be considered as a selective reducing agent towards the ‘free chlorine’ and that its use allows the direct voltammetric titration of ClO2.Spectrophotometric titrations were not allowed as in basic media, where ClO2 strongly absorbs, the reaction rate is very low and in acidic media HClO weakly absorbs. Moreover, in the slightly acidic medium chosen, the spectra of chloroamines, NO22 and NO32 are heavily overlapped. Work is in progress to ascertain the possibility of distinguishing among the different chloroamines. An immediate consequence of this finding is the possibility of determining ClO2 via NO32.Fig. 4 shows the linear correlation between the amount of ClO2 taken and the amount of NO32 generated, confirming the correctness of the proposed procedure. The validation of the overall proposed method relies on the characteristics of the IC analyses, on the properties of the calibration plots relative to NO32, ClO22 and ClO32 (see Table 2) and on the statistics of a comparison between replicate determinations of ClO2 via iodimetric and IC procedures (see Table 3). From Table 2, the good performance of the IC procedure can be seen in terms of both detection limits and dynamic range for the considered analytes.From Table 3, it appears that the two compared procedures furnished statistically equal results in terms of both accuracy and precision since both t- and F-tests satisfied the null hypothesis. In fact, the critical values of t and F (tcr and Fcr) exceed the experimental values.The coincidence between the results obtained with amperometric and IC analysis proves the absence of nitrate ion in the original test solution. It must be noted that, in our experience, no commercially available product ever contained detectable amounts of nitrate. Analysis of real samples The obvious application of the analytical procedure examined is in the analysis of commercial samples of NaClO, which contain large amounts of ClO2, and variable amounts of ClO22 and ClO32 depending on the solution ageing, together with some additives.For this kind of analysis, the approach described was completely satisfactory since no interference in the IC tests was found. The problem arising from the very different concentration levels of NO32 (coming from ClO2), ClO22 and ClO32 can easily be overcome by using an appropriate data acquisition system (see Experimental). As a typical example, Fig. 5 shows the chromatogram and Table 4 the analyte concentrations relative to a commercially available product.As expected, an aged sample contains appreciable amounts of ClO22 and ClO32 that are easily quantified by the proposed method. The fairly high dilution necessary for the optimum performance of the IC analysis always avoided the overlapping of the large nitrite peak on the smaller chlorate peak. Anyway, IC ‘blank’ tests performed with the same chlorate concentration as in Fig. 5 and a five-fold more concentrated nitrite concentration Fig. 3 Voltammetric curves recorded during the titration of NaClO in the presence of NH3 with standard NaNO2 solution. (a) 0, (b) 70, (c) 140, (d) 210 and (e) 280 ml of NaNO2 solution. Experimental conditions: pH 4.7 for acetate buffer; initial volume, 40.2 ml; formal NaClO concentration, 2.7 3 1023 mol l21; formal NH3 concentration, 1.7 3 1023 mol l21; NaNO2 concentration, 0.1 mol l21; other experimental conditions as in Fig. 2. Fig. 4 Correlation between the amounts of ClO2 taken and the amounts of NO32 generated from the oxidation of NO22. Regression coefficients (y = mx + q): q = 2 3 1023 ± 0.011 mmol; m = 1.011 ± 8 3 1023; r2 = 0.999.Table 2 Characteristics of the IC analyses and regression parameters of the calibration plots (y = mx + q) Parameter NO32 ClO22 ClO32 m/arbitrary units mol21 l 4 221 212 680 6 478 804 181 6 505 518 083 q (arbitrary units) 845 15198 26274 r 0.999 0.999 0.999 sm/arbitrary units mol21 l 51 384 565.5 81 419 251 83 655 173 sq (arbitrary unit) 472 36068 37059 Detection limita/mol l21 4.8 3 1027 9.0 3 1028 3.6 3 1027 Linear dynamic range/ mol l21 1026–1023 5 3 1027– 5 3 1024 8 3 1027– 7 3 1024 Retention time repeatabilityb (%) 0.3c, 0.4d 0.1c, 0.5d 0.3c, 0.4d Peak area repeatabilityb (%) 6c, 5d 3c, 6d 6c, 5d a Calculated according to the Hubaux–Vos method.11 b As relative standard deviation for five replicate runs.c Sample concentration used: 6 3 1026 mol l21. d Sample concentration used: 1 3 1024 mol l21.Table 3 Determination of ClO2: comparison between the titration and the IC procedures Parameter Titration IC Standard deviationa/mol l21 0.007 0.011 Averagea/mol l21 0.558 0.559 F-test Fexp. = 2.78 (Fcr,0.05,4,4 = 6.39) t-testb texp. = 0.19 (tcr,0.05/2,8 = 2.31) a From five replicate tests. b Calculated as t x x s n n = - + ( ) 1 2 1 2 1 1 p with s n s n s n n p = - + - + - ( ) ( ) . 1 1 2 2 2 2 1 2 1 1 2 Analyst, 1999, 124, 1439–1442 1441than that in Fig. 5 again gave complete resolution between the two peaks. The possible interference of other anions present in the sample solution or used as buffers in the proposed method was taken into account by injecting standard solutions of them. No interference was detected for fluoride, silicate, bromide, bromate and acetate (used as the buffer solution for the nitrite oxidation and eluted inside the huge chloride peak). Conclusions The results clearly indicate that the proposed procedure is suitable for the quantification of ClO2, ClO22 and ClO32.A single chromatographic run furnishes the concentrations of ClO2, ClO22 and ClO32 present in the same sample. Statistical evaluations of the results obtained demonstrate that the determination of ClO2 as NO32 corresponds perfectly to the classical iodimetric procedure. In particular, it has been verified that NO22 is not only a convenient masking agent of ClO2 but also a selective reagent of the free chlorine so that when only the ClO2 content is required, voltammetric titration with NO22 is a selective procedure.Acknowledgements We gratefully acknowledge the financial support of the Italian National Council of Research (CNR) and the Ministry of University and Scientific and Technological Research. References 1 A. Girelli, L. Matteoli and F. Parisi, in Trattato di Chimica Industriale ed Applicata, Zanichelli, Bologna, 1973, p. 675. 2 G. Gordon, L. C. Adam, B. P. Bubnis, B. Hoyt, S. J. Gillette and A. Wilczak, J. Am. Water Works Assoc., 1993, 9, 89. 3 APHA, AWWA and WEF, Standard Methods for the Examination of Water and Wastewater, American Public Health Association, Washington DC, 18th edn., 1992, ch. 4. 4 K. Suzuki and G. Gordon, Anal. Chem., 1978, 50, 1596. 5 C. Adam and G. Gordon, Anal. Chem., 1995, 67, 535. 6 G. Peintler, I. Nagypal and I. R. Epstein, J. Phys. Chem., 1990, 94, 2954. 7 G. H. Jeffery, J. Bassett, J. Mendham and R. C. Denney, Vogel’s Textbook of Quantitative Chemical Analysis, Longman, Harlow, 5th edn., 1989, p. 396. 8 R. C. Brasted, Anal. Chem., 1951, 23, 980. 9 G. Charlot, Les Reactions Chimiques en Solutions. L’Analyse Qualitative Minerale, Masson, Paris, 5th edn., 1975. 10 L. C. Adam, I. Fabian, K. Suzuki and G. Gordon, Inorg. Chem., 1992, 31, 3534. 11 L. E. Vanatta and D. E. Coleman, J. Chromatogr. A, 1997, 770, 105. Paper 9/04562G Fig. 5 Chromatogram of a commercial product diluted 200-fold, obtained after the suppression of ClO2 with NO22. Chromatographic conditions are reported in Table 1. Table 4 Analysis of a hypochlorite commercial sample by IC analysis Ion Concentration/mol l21 NO32 (HClO) 0.559 ± 0.011 ClO22 0.00150 ± 0.00007 ClO32 0.049 ± 0.002 1442 Analyst, 1999, 124, 1439–1442

ISSN:0003-2654    

DOI:10.1039/a904562g

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Determination of equilibrium constant of alkylbenzenes binding to bovine serum albumin by solid phase microextraction Haodan Yuan,a Ravi Ranatunga,b Peter W. Carrb and Janusz Pawliszyn*a a The Guelph-Waterloo Center for Graduate Work in Chemistry and the Waterloo Center for Groundwater Research, Department of Chemistry, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1 b Department of Chemistry, University of Minnesota, Smith and Kolthoff Halls, 207 Pleasant Street Southeast, Minneapolis, MN 55455, USA Received 14th June 1999, Accepted 25th August 1999 Solid phase microextraction (SPME) coupled with GC has been applied to study the binding properties between bovine serum albumin (BSA) and volatile organic compounds such as benzene, toluene, ethylbenzene, propylbenzene and butylbenzene.Their protein–ligand equilibrium constants have been determined. The measurement of free and bound ligand concentrations in the aqueous solution was based on the equilibrium among the analyte in the fiber coating (Cf), headspace (Ch) and aqueous solution (Cs).The work demonstrated that SPME is a simple and effective method in the study of protein binding to measure the freely dissolved analyte concentration as well as the equilibrium constant. The theoretical aspect of the SPME applied to the equilibrium constant measurement in two-phase (liquid sample–fiber coating) and three-phase (liquid sample–headspace–fiber coating) systems has been thoroughly discussed.The results demonstrated that the interpretation of the calibration data is crucial to the determination of freely dissolved analyte concentration and the equilibrium constant especially when the sample volume is small. The error in the experimental system is discussed. It is demonstrated in this study that for the three-phase system the amount of the analyte partitioned in the headspace could be ignored only in certain circumstances, where the Henry’s law constant and the ratio between headspace volume and sample volume are sufficiently small.Introduction In toxicology, as well as in pharmacology, effects of a chemical are related to the target site concentration of the said chemical. In in vivo studies, concentrations are usually stated as either the nominal or the solvent extractable concentration. The disadvantage of these approaches is that the binding to the biological matrix itself is not taken into consideration.1 This can have severe implications for the quantitative description of dose-response relationship.In general, only the unbound drug is active and capable of diffusing across membranes.2 Conventional methods to determine freely available concentrations or to study the binding of hydrophobic chemicals to a biological matrix include equilibrium dialysis, ultrafiltration, and centrifugation.3 These methods are either labor intensive, time consuming, or not compatible with the sample matrix itself.Solid phase microextraction (SPME) is a novel sample preparation and sampling technique, which was first introduced in 1990.4 The method was initially developed for volatile organic compounds analysis, such as benzene and its derivatives. Now SPME has been applied to the determination of numerous compounds in different matrices. In SPME, a small fused-silica fiber coated with a polymeric stationary phase, which is often mounted in the needle of a syringe-like device for protection, is used for analyte extraction from a matrix.The stationary phase can be a viscous liquid [e.g., poly(demethylsiloxane), PDMS] or a porous solid (e.g., polydivinylbenzene; DVB). During the sampling process, analytes are absorbed into the fiber until equilibrium is reached within the system. The extraction amount will depend on the analyte partitioning between the sample and fiber coating. The SPME sampling can be carried out directly from the gaseous or liquid samples or from the headspace over liquid or solid phases.5 Since the amount of analyte extracted by SPME is so small that in most cases it will not disturb the equilibrium of the system, SPME has been successfully applied to the measurement of partition coefficient.6–8 Compared with some traditional methods, such as dialysis, ultradialysis and centrifugation, SPME is simpler and more efficient.Poerschmann et al.7 discussed the distribution of certain organic compounds binding to dissolved organic matter.The authors demonstrated that the amount of the analyte on the fiber could be so small that it can be neglected for semi-volatile and most volatile organic compounds. The amount of the analyte in the headspace was also neglected since the headspace volume is very small compared with sample volume. The total concentration was measured by isotopically labeled internal standard, while the freely dissolved fraction of the analyte in the solution was measured by the calibration of initial concentration versus the amount of the analyte on the fiber.Vaes et al.8 measured the freely available concentration of compounds and the lowering of the concentration due to binding to biological matrices. The calibration curves in which the total concentration spiked into the solution versus the amount of the analyte extracted by the fiber were established in buffered aqueous solutions of the test chemicals. The same calibration curves were used to measure the free concentrations of the compounds in the protein binding study when BSA was added.However, there are some cases where it is hard to prepare a very large volume of the protein solution with a significant headspace volume, or the compound is so volatile that the analyte partitioned in the headspace and fiber cannot be neglected even though the headspace volume is small. In these This journal is © The Royal Society of Chemistry 1999. Analyst, 1999, 124, 1443–1448 1443cases, the amount of the compounds in the headspace is not negligible and the free available concentration of the analyte should be measured through careful consideration of the distribution of the compound both in calibration and the experimental system.The aim of this study is to investigate the application of SPME in the determination of the freely dissolved analyte concentration as well as equilibrium constant under the above two circumstances.The errors introduced if the calibration curve of initial concentration versus amount of analyte extracted by the fiber is used under these circumstances and how to avoid them are discussed. In the experiment, alkylbenzenes (benzene, toluene, ethylbenzene, propylbenzene and butylbenzene) were used as the target compounds while BSA was used as the model protein. Theory For a three-phase system consisting of aqueous phase, dissolved pseudophase, which is dissolved protein in this study, and headspace above them, the initial amount of compounds in the sample is distributed among each phase according to their partition coefficients.With the fiber in the system, the fiber coating as another phase will be also partitioned with the compounds. The concentration of the compounds freely dissolved in aqueous phase (Cs), bound to the dissolved protein (Cb), partitioned in headspace (Ch) and on the fiber coating (Cf), and their interrelationships, are shown in Fig. 1. The distribution constant of the compounds between the fiber and headspace is Kfh, where Kfh = Cf/Ch. Similarly, the distribution of the compound between the headspace and sample solution, the fiber coating and the aqueous phase are defined as Khs = Ch/Cs, Kfs = Cf/Cs = KfhKhs,9 respectively. Obviously, Khs is the Henry’s law constant. Reversible interactions between a protein (P) and ligand (L) can be described by the following thermodynamic equilibrium: L + P LP K � æ æ Æ (1) K is the apparent equilibrium constant, expressing the affinity of the protein for a particular ligand.K can be calculated by eqn. (2): K = [LP] [L][P] (2) where [L], [P] and [LP] are the molar concentrations of the compound, the protein and bound analyte (bound protein) respectively. The total concentration of the ligand and protein is known as the initial concentration spiked into the system. In order to calibratehe absolute amount of the compound extracted by the fiber, nf, the response factor of the GC detector should be obtained.This could be achieved by the liquid injection. In the calibration step, in which the protein is absent, the amount of an analyte extracted by the SPME fiber from headspace of a three-phase system n0f is described by the following relationship:9 n K K V C V K V K V V f 0 0 = + + fh hs f s fh f hs h s (3) where C0 is the initial concentration of the analyte in the matrix; Vf, Vs and Vh are the volumes of the coating, the matrix and the headspace, respectively.The volumes of the fiber coating, the headspace and the matrix can be assumed constant. The distribution ratios are also constant if the temperature is kept constant. Therefore, the amount of an analyte extracted by the SPME fiber has a linear relationship with the initial concentration of the analyte in the sample. This relationship is what people usually used as a calibration curve to determine the free concentration in the binding study.In the three-phase system, the mass balance is: n n n n total f h s 0 0 0 0 = + + (4) where n0f , n0h and n0s are the amounts of the compounds partitioned on the fiber coating, headspace and aqueous phase, respectively. An equation could be derived to show the relationship between the equilibrium concentration (C0s ) in the sample solution and the amount extracted by the SPME fiber by substituting the distributions ratios into eqn. (4): C n n V V K n K V s total f s h hs f fs f 0 0 0 0 = - + = (5) From eqn.(5) the equilibrium concentration has a linear relationship with the amount of the analyte extracted by the SPME fiber. However, the initial concentration C0 is not the equilibrium concentration in the sample solution. It should be emphasized that in the equilibrium constant determination, the equilibrium concentration of free analyte in the sample must be used in the preparation of the calibration curves rather than the initial concentration in the sample matrix.In other words, the linear relationship represented by eqn. (5) should be used as a calibration curve to measure the free concentration in the protein binding study rather than eqn. (3). Once the free concentration is obtained from the calibration curve represented by eqn. (5), the bound concentration can be obtained through the mass balance in a four-phase system. ¢ = ¢ + ¢ + ¢ + ¢ n n n n n total f h s b (6) where nA b is the amount of the analyte bound to protein.The bound concentration equals: ¢ = ¢ - ¢ - + Ê Ë Á � � � ¢ C n n V K V V C b total f s hs h s s 1 (7) The equilibrium constant can be calculated by substituting the values of CAsand CAs into eqn. (2). Experimental Apparatus and reagents The target alkylbenzene compounds (benzene, toluene, ethylbenzene, propylbenzene and butylbenzene) and bovine serum Fig. 1 Partition of alkylbenzenes in a four-phase (fiber coating– headspace–aqueous phase–dissolved protein) system. 1444 Analyst, 1999, 124, 1443–1448albumin (BSA, 98% purity) were purchased from Sigma (St. Louis, MO, USA). The SPME holder and fibers coated with 30 mm PDMS were purchased from Supelco (Bellefonte, PA, USA). These devices were employed for experiments with all five alkylbenzene compounds. Fibers were conditioned for 2 h at 250 °C under a flowing helium stream in the GC injector. The buffer was prepared by combining 100 mM disodium hydrogen orthophosphate and 100 mM sodium dihydrogen orthophosphate solution to pH 7.This buffer solution was used to prepare BSA solutions (4 mg mL21). Alkylbenzene stock standard solutions (1 mg mL21) were prepared by adding the required amount of alkylbenzene into methanol; 0.1, 0.01 and 0.001 mg mL21 stock standard solutions were prepared by 10 times dilution of the more concentrated standard. Standard solutions for the five alkylbenzenes individually and with all five compounds in one solution were prepared for experimental convenience. A Varian (Palo Alto, CA, USA) 3400cx GC was used for the experiments, as was an SPB-5 30 m 3 0.25 mm column with 0.25 mm film thickness.The injector was held at 250 °C for fiber desorption during the analysis. For the syringe injection, the injector was temperature programmed as: initial temperature, 43 °C, with a hold for 1 min, then the injector temperature was increased to 250 °C with the 250 °C min21 ramp. This temperature was held for the rest of the run.The column temperature was programmed as follows: for SPME fibre analysis, the initial column temperature was 60 °C, this temperature was held for 1 min, then the temperature was increased to 120 °C at a rate of 20 °C min21. This temperature was held until the end of the run. The total analysis time for each run required 6 min. Fig. 2 shows the chromatogram and temperature program. For the syringe injection, the initial column temperature was 40 °C, held for 1 min. Then the temperature was increased to 120 °C at 10 °C min21.This temperature was held for 1 min. The total analysis time for each run was 10 min. Helium (UHP grade) was used as the carrier gas. The head pressure during the analysis was 25 psi. 15 mL glass vials (Supelco, Bellefonte, PA, USA) were used for the experiment. Experimental procedure Syringe injection was used to calibrate the absolute mass of the analytes injected into the GC. By comparing the area counts of the syringe and fiber injections, we know the absolute mass that was injected into the GC injector in fiber injection.All the extractions described below were performed at ambient temperature under 23 °C. For the calibration measurement, 8 mL of buffer solution was transferred to a 15 mL vial. The actual volume of the vial was determined by measuring the volume of the water required to completely fill the vial. The headspace volume of the vial was determined by subtracting the solution volume from the total vial volume.An amount of analyte stock solution was spiked into the buffer solution to give the desired concentration of analyte solution. The fiber was inserted into the headspace of the vial while the sample was agitated with a digital stirrer/hot plate and magnetic stir bar. The extraction required 3 min to reach equilibrium. After the extraction, the fiber was introduced into the GC injector port for the thermal desorption.The desorption time was controlled for 1.5 min. No carryover was found. The linear range of the analytes was from 50 to 2000 ppb. For each concentration, three replicates were performed. In the protein binding analysis, 8 mL of BSA solution was added into the vial (4 mg mL21 BSA in buffer solution). The experimental procedure was the same as that used in the calibration curve measurement. Three concentrations of alkylbenzenes (200 ppb, 500 ppb and 1000 ppb) were investigated.The K value was calculated according to the method described in the theory section. Results and discussion The calibration equations and their regression coefficients for the five test compounds are listed in Table 1. These calibration curves were used to calculate the freely dissolved analyte concentration (CAs) in the presence of BSA. The concentration of alkylbenzenes bound to CAb can be subsequently calculated from eqn. (7). The values of CAs and CAb calculated by using 200 ppb sample (8 mL solution) are presented in Table 2.The mean values of logK calculated from the measurements of three alkylbenzene concentrations (200 ppb, 500 ppb and 1000 ppb) are shown in Table 3. The results obtained from headspace GC analysis and SPME/GC10 analysis are listed and compared in Table 3. The relative difference of the two methods for the test Fig. 2 The chromatogram and the temperature program for the alkylbenzenes binding to bovine serum albumin.Table 1 Calibration curve of analyte concentration in water (ng ml21, y axis) versus analyte mass loaded on fiber (ng, x axis) Chemical Regression equation r2 Benzene y = 59.364x + 0.2051 1.0000 Toluene y = 19.429x + 2.5477 1.0000 Ethylbenzene y = 6.7254x + 4.4276 1.0001 Propylbenzene y = 2.0572x + 2.8099 1.0003 Butylbenzene y = 0.+ 7.5732 0.9996 Table 2 Summary of analyte fiber mass loading and concentrations freely dissolved in water and bound to protein (initial ligand sample concentration was 200 ppb) Chemical nf/ng Cs/ng mL21 Cb/ng mL21 Benzene 2.59 153.77 16.69 Toluene 6.02 119.51 52.25 Ethylbenzene 11.64 82.71 93.50 Propylbenzene 24.77 53.77 125.39 Butylbenzene 26.23 22.52 116.07 Table 3 Comparison of logK values obtained from the SPME method and from headspace GC analysis Chemical LogKSPME LogKheadspace GC Relative difference (%)a Benzene 3.51 3.53 0.57 Toluene 3.71 3.83 3.23 Ethylbenzene 4.15 4.16 0.24 Propylbenzene 4.49 4.42 1.55 Butylbenzene 4.87 — — a Relative difference (%) = abs[(logKSPME 2 logKheadspace GC)/ logKSPME)]3100 Analyst, 1999, 124, 1443–1448 1445compounds was less than 4%.Comparable results obtained from another independent analytical method experimentally proved that SPME was a valid method in the equilibrium constant measurement. Effect from the headspace and fiber coating The effect of the headspace and fiber coating on the equilibrium constant measurement can be clarified by the following discussions.Equations are derived to describe the freely available analyte concentration in different systems. In a two-phase system with aqueous solution and dissolved protein the mass balance is: ¢ = ¢ + ¢ n n n total s b (8) Since CAp = Cp 2 CAb, where Cp is the initial protein concentration, from eqn. (2), a few simple rearrangements yield the expression describing the bound concentration at equilibrium: ¢ = ¢ ¢ × C KC KC C b s s p 1+ (9) The expression of the concentration of freely available analyte can be subsequently obtained by substituting eqn.(9) in eqn. (8): ¢ = ¢ + + ¢ C n V KC V KC s total s p s s /( ) 1 (10) In a three-phase system the mass balance is: ¢ = + + = + ¢ + ¢ n n n n V K V C C V total s h b s hs h s b s ( ) (11) Similarly, an equation describing the freely dissolved analyte concentration in such a system can be obtained. The term in the denominator (KhsVh) shows the presence of headspace. ¢ = ¢ + + + ¢ C n V K V KC V KC s total s hs h p s s ( ) /( ) 1 (12) However, after the fiber was inserted into this three-phase system, a four-phase system was established.The mass balance for such a system is: ¢ = ¢ + + ¢ + ¢ + ¢ n n V K V C C C V total f s hs h s s b s ( ) (13) The equation to describe the freely dissolved analyte concentration at equilibrium is: ¢ = ¢ + + + + ¢ C n V K V K V KC V KC s total s hs h fs f p s s ( ) /( ) 1 (14) The term KfsVf in the denominator in eqn. (14) shows the presence of fiber in the system.It is obvious from eqns. (10), (12) and (14) that the concentrations of the freely dissolved analyte are dependent on the different systems. In other words, even with the same sample volume and the same concentrations of dissolved protein and compounds, the free analyte concentration in the solution is different in the presence of headspace and fiber due the equilibrium distribution of the analyte among these matrices. Fig. 3 illustrates this phenomenon. Toluene is used as the example compound to give these curves.The following parameters have been used to calculate the course of the curves: K = 5128 mol21·L, Khs = 0.23, Kfs = 189,9 Vs = 8 mL, Vh = 7.8 mL, Cp = 5.97 3 1025 mol L21. Curves (a), (b) and (c) correspond to eqns. (10), (12) and (14). In eqns. (12) and (14), for volatile compounds, Khs is usually close to 1,9 which means that headspace volume can be neglected only when it is very small with respect to the sample volume.Semi-volatile compounds have much smaller Khs, the KhsVh term may be negligibly small in a relatively large headspace volume. However, in such applications, the assumption that the amount of the analyte in the headspace can be neglected should always be verified. The term KfsVf in the denominator of eqn. (14) represents the effects from the fiber extraction. The fiber volume Vf usually is a very small value; for 30 mm PDMS fibre, the volume of the liquid coating is 0.132 3 1026 L.This indicates that the value of Kfs can increase up to 106 in order to have a significant effect. Apparently, headspace plays a more important role relative to the SPME fiber. The second column of Table 4 shows the percentage of the analyte in the headspace relative to the amount of the analyte in water when the volume of headspace and that of sample solution are identical. These numbers are calculated from Henry’s law constant in column 1. It is clearly shown that these values range from 23.1% (benzene) to 55.5% (butylbenzene) and they are sufficiently large not to be ignored. Therefore, the amount of the analyte in the headspace is too large to be neglected in this experiment, thus the calibration curve of the total concentration versus the amount of the fiber could not be used.The amounts of the analyte partitioned in the headspace and the fiber have to be considered. The calibration curve of equilibrium free concentration in aqueous solution (C0s ) and the amount of the analyte extracted by the fiber (n0f ) should be used to measure the free concentration.Calibration methods As described in the Introduction, in the previous studies the initial concentration versus the amount of the analyte parti- Fig. 3 Freely available analyte concentration in (a) two-phase system (aqueous phase–dissolved protein) (b) three-phase system (headspace– aqueous phase–dissolved protein) (c) four-phase system (fiber–headspace– aqueous phase–dissolved protein).Table 4 Henry’s law constant for the target analytes (column 2) and the percentage of the total mass of analyte in the headspace relative to the mass in the solution at equilibrium when the volume of the headspace and sample solution are the same (25 °C) (column 3) Chemical KH/ L atm mol21 Percentage in headspacea Benzene 5.62 23.1% Toluene 6.76 27.8% Ethylbenzene 8.51 35.0% Propylbenzene 9.77 40.0% Butylbenzene 13.49 55.5% a The value in this column is calculated from Henry’s law constant listed in column 2. 1446 Analyst, 1999, 124, 1443–1448tioned onto the fiber was plotted as the calibration curve since it was thought that the amount of the analyte partitioned in the headspace and on the fiber was negligible. Since the initial concentration can be considered as the concentration when the vial is full with the aqueous solution, it is likely to be mistakenly regarded that the concentration measured is the free available concentration when the vial is full of protein solution. This thinking is wrong in that the partitioning of analyte into the headspace and fiber will cause the shift of the equilibrium between freely dissolved analyte and protein to the freely dissolved analyte side [eqn.(1)]. In this way, the freely dissolved analyte concentration measured is the sum of freely dissolved analyte concentration before the extraction and analyte concentration from the dissociation of the bound analyte.From eqns. (10) and (14), suppose KCAs < < 1 and KC0s , which is satisfied in the experiment, we will have: ¢ = ¢ + + + = + + + + ¢ n K V n V K V K V KC V K V V KC V V K V K V KC V C f fs f total s hs h fs f p s fs f s p s s hs h fs f p s s ) ( ) ( ) (15) However, the dependence of the true freely dissolved analyte concentration and the amount of the analyte on the fiber can be easily obtained from eqn. (5): n K V V V K V K V C f fs f s s hs h fs f s 0 0 = + + (16) The difference between eqn.(15) and eqn. (16) with a large range of K values is illustrated in Fig. 4 [curve (a) for eqn. (15) and curve (b) for eqn. (16)]. It is clear that the difference becomes more significant with the increase of equilibrium constant, K. It should also be noted that when KCAs and KC0s become comparable with 1, the system will be non-linear. Conclusion The amount of the analyte extracted by the fiber normally is so small that it can be neglected in the free available concentration, thus the equilibrium constant, measurement.When the volume of headspace is insignificant compared with the sample volume or the Henry’s law constant is very small, the amount of the analyte partitioned into the headspace can be neglected. In this case, the calibration curve of initial concentration versus the amount of the analyte extracted by the fiber can be employed to calculate the free available concentration. In the other cases where the headspace volume is significantly large compared with the sample volume or the Henry’s law constant is big, the amount of analyte in the headspace should be considered in establishing the calibration curves. In this situation, using the calibration curve of the initial concentration versus the amount of the analyte on the fiber to calculate the freely dissolved analyte concentration will cause systematic experimental errors. The amount of the error will depend on the equilibrium constant K. Acknowledgements This work was supported by the Natural Sciences and Engineering Research Council of Canada, National Institutes of Health, Supelco Inc. and Varian. References 1 R. L. Fisher, A. J. Candolfi, I. G. Sipes and K. Brendel, Drug Chem. Toxicol., 1993, 16, 321. 2 J. K. Seydel and K.-J. Schaper, Pharmacol. Ther., 1982, 15, 131. Fig. 4 The difference between eqn. (15) (a) and eqn. (16) (b) with different K values (see the text for detail). Analyst, 1999, 124, 1443–1448 14473 J. Oravcová, B. Böhs and W. Lindner, J. Chromatogr. B, 1996, 677, 1. 4 C. L. Arthur and J. Pawliszyn, Anal. Chem., 1990, 62, 2145. 5 Z. Zhang and J. Pawliszyn, Anal. Chem., 1993, 65, 1843. 6 J. R. Dean, W. R. Tomlinson, V. Makovskaya, R. Cumming, M. Hetheridge and M. Comber, Anal. Chem., 1996, 68, 130. 7 J. Poerschmann, Z. Zhang, F.-D. Kopinke and J. Pawliszyn, Anal. Chem., 1997, 69, 597. 8 W. H. J. Vaes, E. U. Ramos, H. J. M. Verhaar, W. Seinen and J. L. M. Hermens, Anal. Chem., 1996, 68, 4463. 9 J. Pawliszyn, Solid Phase Microextraction: Theory and Practice, Wiley-VCH, New York, USA, 1997 (pp. 146 and 156). 10 J. Li and P. W. Carr, Anal. Chem., 1993, 65, 1443. Paper 9/04723I 1448 Analyst, 1999, 124, 1443–1448

ISSN:0003-2654    

DOI:10.1039/a904723i

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Development of a methodology for the determination of carbon monoxide using a quartz crystal microbalance M. Teresa S. R. Gomes,* P. Sérgio T. Nogueira, A. C. Duarte and João A. B. P. Oliveira Department of Chemistry, University of Aveiro, 3810-193 Aveiro, Portugal Received 7th June 1999, Accepted 27th August 1999 Starting from a method published in the literature for the determination of CO by a quartz crystal microbalance, important changes were made in order to solve the principal analytical problems found, as well to simplify the analytical procedure in order to become a general methodology.The new methodology allows quantitative determination of both CO and CO2. Sensitivity for CO is governed by its conversion to CO2 by silver oxide, which depends on the dimension of the Ag2O column percolated by the gas. Detection limit for CO, using a crystal coated with tetramethylammonium fluoride with an amount corresponding to a frequency decrease of 18.7 kHz was 7.7 mL, and for CO2 a detection limit of 19.2 mL was found, using a crystal with a coating amount that produced a frequency decrease of 13.6 kHz.A sample of the exhaust gas of a car was analysed and the results compared with the ones obtained by non-dispersive infrared spectrometry. Introduction Carbon monoxide is a toxic gas, and due to the lack of colour and smell it is difficult to be detected by man. Its high toxicity is related to the fact that blood haemoglobin possesses an affinity 200 times higher for CO than for O2.The presence of CO in closed spaces produces headache and sleepiness and can finally lead to death, if half of the haemoglobin molecules have been complexed with CO. The general concern and environmental protection policies demands a control of CO emissions. Carbon monoxide exists mainly as a result of incomplete oxidation of carbons and hydrocarbons, and besides environmental concern, economic aspects also demand an effective control of the efficiency of combustion processes.Sensors for carbon monoxide, based on quartz crystal microbalances (QCM), have already been reported.1,2 They were either based on the reaction of CO with HgO, and consequent production of mercury vapour that amalgamates on the gold electrodes of the piezoelectric crystal,1 or they took advantage of the temperature coefficient of the crystal.2 The first sensor has the drawback of a continuous mercury background, while the latter was also sensitive to inflammable gases, like H2 or isobutane.Starting from the sensor proposed by Ho et al.,1 changes in the methodology have been performed, and a new sensor based on the reaction between CO and Ag2O is proposed. The experimental steps that lead to the ultimate choice are here displayed, as they elucidate the principal problems associated with alternative arrangements, and show the advantages of the method now presented. A real sample from the gases emitted by a car with engine problems was analysed, with respect to CO and CO2, both by the proposed method and by non-dispersive infrared spectrometry. Experimental Apparatus Three methodologies were compared in this work.The first methodology, which involves HgO and Hg(0) detection, was previously described by Ho et al.,1 while the others are new, and will be the only ones for which the apparatus will be described. Both new methodologies use quartz piezoelectric crystals coated using a spraying device described elsewhere.3 Methodology 2: involving HgO and carbon dioxide detection.Fig. 1 shows the experimental layout used to determine both CO and CO2 standards in the first experiments. A quartz crystal coated with tetramethylammonium fluoride tetrahydrate (TMAF) was used to detect CO2 4 when a sample was injected through an Omnifit septum injector. The syringes were SGE (Scientific Glass Engineering, Palo Alto, CA, USA) gas tight, with valve.CO was detected as CO2 after reaction with HgO. An oven was used to increase the temperature at which CO reacts with HgO. Before reaching the reaction chamber, the gas stream was heated during the passage through a 2 m long copper coil, also placed inside the oven. After the reaction chamber, the gas flows through an acidic KMnO4 solution, that prevented mercury vapour from reaching the crystal, and then through a Nafion membrane. The gaseous stream is finally divided in two streams, each one directed to one of the coated quartz crystal electrodes.The gas flow was controlled with a variable area flowmeter (Cole Parmer). Fig. 1 Experimental layout for the determination of CO by methodology 2, using HgO. (P: power supply; O: oscillator; F: frequency meter; X: crystal cell; K: acidic KMnO4 solution; I: sample introduction; R: flowmeter). This journal is © The Royal Society of Chemistry 1999. Analyst, 1999, 124, 1449–1453 1449Methodology 3: involving Ag2O and carbon dioxide detection.Fig. 2 shows the ultimate layout for the methodology now proposed. The injected sample, carried by the nitrogen flow, passes through two absorbent tubes placed before the crystal cell. For the determination of CO in samples, the first tube contained soda lime and the second one Ag2O, while for the determination of CO2 there is just one tube with molecular sieve RbA (MS RbA). The piezoelectric quartz crystals were 9 MHz (SIWARD), its frequency was monitored with a Universal Counter Board (Keithley MetraByte Corporation, Taunton, USA), and the data were stored in an ASCII file format.Oscillator and power supply were both laboratory made. A 9 MHz quartz crystal with gold electrodes (ICMInternational Crystal Manufacturing Co, Inc., Oklahoma, USA) was used for the evaluation of mercury background. A gaseous sample was collected into a sampling bag (Cole Parmer, Vernon Hills, USA). For comparison, the sample was analysed both by the proposed QCM methodology and by a single beam non-dispersive infrared spectrometer ADC, model RF558, with a rotating filter (The Analytical Development Co, Ltd., Hoddesdon, UK).Reagents TMAF (Aldrich 10,721-2; Madrid, Spain) was dissolved in ethanol (Merck 11727; Darmstadt, Germany). SO2 was generated by the addition of hydrochloric acid (Riedel-de-Häen 30721; Hannover, Germany) to sodium sulfite (Merck 6652) and CO was generated by the addition of formic acid (Merck 264) to sulfuric acid (Merck 100731).NOx was generated by the addition of HNO3 (Carlo Erba 408025; Milan, Italy) to brass. All the generated gases were dried with silica gel. CO2 was N45 grade and both N2 and Air were Alphagaz 1, all from ‘ArLíquido’ (Lisbon, Portugal). A molecular sieve with a different pore size was prepared by mixing molecular sieves 3A (BDH 33162, Poole, UK) with rubidium chloride (Aldrich 21,527-9). Procedure As before, just the two new methodologies will be described.In both of them, quartz crystals were coated on both sides by spraying a 1% solution of TMAF in absolute ethanol. Crystals were immediately introduced into the crystal cell, where ethanol evaporated under a nitrogen flow. Stabilisation in the frequency of the crystal is achieved after the solvent evaporation. Known volumes of pure CO, or CO2, or of a gaseous sample, were injected into a constant nitrogen flow, and after passing through the reactants that convert CO into CO2, and also through appropriate absorbents to remove interferents, they reach the quartz crystal coated with TMAF.TMAF interacts reversibly with CO2,4 and the observed frequency change is proportional to the CO2 that impinges the coated crystal. Methodology 2: involving HgO and carbon dioxide detection. Monitoring of CO. CO is converted to CO2 after reacting with HgO, at a temperature higher than 200 °C. As mercury vapour is also produced in the reaction, the gaseous stream needs to pass through an acidic KMnO4 solution, to prevent it from reaching the Ag coated electrodes of the quartz crystal where it would produce a frequency change, irreversible at room temperature.As TMAF is hygroscopic, the gaseous stream needs to be dried after passing through the aqueous solution, and this is the reason why a Nafion membrane, with an external counterflow of air, was placed before the crystal cell. As the methodology was used just with pure CO, and it was abandoned during the sensor development stage, no absorbents were used, either to absorb CO2 (to allow selective CO determination), or to eliminate gaseous interferents.Monitoring of CO2. CO2 can also be monitored, if the gas does not pass through the Hg2O tube, and conversion of CO into CO2 does not occur. Standards of pure CO2 were injected. Methodology 3: involving Ag2O and carbon dioxide detection. Monitoring of CO. In the final arrangement, CO is converted into CO2 after reacting with Ag2O, at room temperature.The detection system is the same as before, and neither absorbent solution, nor drying of the standards of gas is needed. As the coated crystal responds to CO2, independently from its origin, in order to sense just the CO2 that has its source in the CO present in the sample, a soda lime absorbent was used to remove CO2 from the gas stream. Soda lime is a very efficient absorbent which removes most interferent gases, as well as water from the sample.It is important to keep the order of the soda lime and Ag2O tubes in the experimental layout for CO monitoring, as water deactivates Ag2O.1 Monitoring of CO2. Removing both the soda lime absorbent tube, and the one with the converting reagent (A2O), the CO2 originally present in the sample can be monitored. In the analysis of CO2, where soda lime can not be present, molecular sieves with such a porosity and polarity that allows CO2 to pass and remove SO2 and NOx, were used instead. Water, if present in the samples, is also removed by the molecular sieves.5 Tests showed that all the following zeolites 3A, 4A, 5A and 13X, retained CO2.A specific zeolite that allowed CO2 to pass through, and capable of discriminating between SO2 and CO2 was prepared from 3A zeolites, exchanging the intrazeolitic cation K+ by Rb+, which due to its larger radius reduces the pore size. The preparation consisted in maintaining during 24 h at 60 °C, a mixture of the 3A zeolite reduced to powder, with RbCl in water, with constant stirring.The product is then removed by filtration. The procedure was repeated three times, after which the zeolite was pressed into pellets. Monitoring of both CO and CO2 in a real sample. Calibration lines for CO and for CO2 were obtained injecting known volumes of the corresponding pure gas. A sample to be analysed was pumped from the exhaust pipe of a car into a gas sampling Fig. 2 Experimental layout for the proposed method (methodology 3) to analyse CO or CO2. (P: power supply; O: oscillator; F: frequency meter; X: crystal cell; I: sample introduction; R: flowmeter). 1450 Analyst, 1999, 124, 1449–1453bag, from where a volume which gives a frequency signal near the centroid of each of the calibration lines was withdrawn and injected into the injection port of the analysis system. For the analysis of the same sample by infrared spectrometry, the sample was pumped directly from the sampling bag into the analyser.Results and discussion Methodology 1: involving HgO and mercury vapour detection Previous methodology1 was based on the reaction between CO and HgO: HgO(s) + CO(g) ? Hg(g) + CO2(g) which produces mercury vapour that amalgamates on the gold electrodes of a quartz crystal microbalance. The crystal was cleaned by thermal desorption. As the reaction rate of this transformation is too slow at room temperature to permit quantitative analysis, temperatures of 210 °C have been recommended.1 However, HgO decomposes on heating, giving a continuous mercury background: 2HgO(s) ? 2Hg(g) + O2(g) This mercury background must be considered, and must be known.The detector was zeroed passing a reference stream, with no mercury. Fig. 3 shows several plots of the frequency of the same crystal with gold electrodes when a constant flow of 50 cm3 min21 of nitrogen is passing through heated HgO. The different plots show that the slope of the frequency decrease versus time depends on the initial frequency of the crystal.Therefore, mercury background signal becomes highly dependent on the efficiency of the crystal cleaning process, and must be evaluated for each experiment. Besides the dependence of mercury background on the ability to reproduce experimental conditions, including the efficiency of the crystal cleaning process, Fig. 3 shows that although the slopes of each of the recorded frequency decrease plots are, in general, constant, there are a few irregularities difficult to be predicted or eliminated.Methodology 2: involving HgO and carbon dioxide detection The problem of mercury background can be overcome if mercury is prevented from reaching the crystal and if, instead of mercury, CO2 is the monitored gas. This was accomplished using a quartz crystal coated with TMAF, and passing the gaseous stream through an acidic KMnO4 solution, as shown on Fig. 1. The method allows both CO and CO2 determination, and although crystal coating is necessary, reversibility is faster, as in the Guilbault method1 mercury could only be removed from the Fig. 5 Frequency changes observed after the injection of known volumes of pure CO or CO2 following the proposed methodology, in the arrangement shown in Fig. 2, always without the first tube of soda lime or molecular sieves, but with (a) one Ag2O tube, or (b) two Ag2O tubes. Fig. 3 Several frequency plots of the same quartz crystal with gold electrodes, exposed to mercury background.Fig. 4 Frequency changes observed after the injection of known volumes of pure CO or CO2 in the arrangement shown in Fig. 1. From these results it can be concluded that a 100% conversion of CO to CO2 was not achieved with HgO, under the applied conditions. Analyst, 1999, 124, 1449–1453 1451gold electrodes after heating the crystal at temperatures of 170 °C. It is also known that thermal shocks promotes crystal ageing.6,7 The presence of the trap solution of KMnO4 forces the inclusion of a gas drying unit, but has the advantage of elimination of interferences from SO2 and other reducing gases.The conversion rate of CO into CO2 is dependent on the temperature, and so, besides the oven that was set at 230 °C, a heating tape was placed around the HgO container. Nitrogen flow was maintained at 50 cm3 min21. Fig. 4 shows the frequency changes observed after the injection of known volumes of pure CO or CO2.From these results it can be concluded that complete conversion of CO to CO2 was not achieved. Monitoring CO in the presence of CO2 would require a trap of soda lime. Methodology 3: involving Ag2O and carbon dioxide detection Although the latter methodology allowed CO and CO2 determination, the replacement of HgO with Ag2O would obviate the heating unit, the trap solution, as no mercury is produced in the reaction Ag2O(s) + CO(g) ? 2Ag(s) + CO2(g) and also the drying membrane.The conversion step depends now on the contact between silver oxide and the carbon monoxide stream. In order to evaluate this dependency, an experiment was performed injecting known volumes of CO in the experimental arrangement as displayed in Fig. 2, but without the soda lime tube, and also when a second silver oxide tube was placed after the previous one. In order to evaluate conversion, known volumes of CO2 were also injected.Fig. 5(a) and (b) show that although complete conversion was not reached neither with one nor with two silver oxide tubes, conversion increased when the gas was forced through two silver oxide columns. SO2, which reacts with TMAF,4 is a most likely interferent. A volume of 500.0 mL of SO2 produced a signal of 84 Hz on a crystal with a frequency decrease due to coating of 16.4 kHz. Molecular sieves specially prepared, by exchange of K+ from molecular sieve 3A for Rb+, efficiently eliminated the interference, and in fact, when 2.0 mL of SO2 were allowed to pass through it, no frequency signal was observed.In the analysis of CO, the interference of SO2 is not a problem as soda lime efficiently eliminates it. NOx also interferes, as 1.0 mL gives rise to a 340 Hz signal on a crystal with a frequency decrease due to coating of 25.0 kHz. The interference is not a problem in CO determination, as soda lime reduces this signal to 17 Hz, a signal of the same magnitude as the one obtained for 30.0 mL of CO2, and mainly because, with soda lime, an injection of 300.0 mL of NOx, an amount above the one that can be found in the exhaust gas of an old car of 1970,8 gives no signal on a crystal coated with 19.2 kHz.Concerning CO determination, 1.0 mL of NOx produces no change in the frequency of a crystal with a frequency decrease due to coating of 17.3 kHz, when the gas passes through a tube with the molecular sieve RbA, while on the same crystal, an amount of 10.0 mL of CO2 produced a noticeable frequency shift.A sample of the exhaust gas from a car was collected into a sampling bag in a car workshop. CO2 and CO content were determined both by the proposed method and by non-dispersive infrared spectrometry. The results obtained by the proposed QCM method are displayed in Table 1, and according to them, the sample was 1.43 ± 0.06% in CO, and 2.15 ± 0.09% in CO2. The analysis by non-dispersive infrared spectrometry gave for CO and CO2 1.15% and 2.3%, respectively. The results obtained by both methods are of the same order of magnitude, with a good precision for the QCM analysis.Although it is not possible to give the precision of the results obtained with the infrared analyser, it was known, from previous experiments, that the result of 2.3% was expected to be affected by an imprecision of ±0.1%. The detection limits (3s) of the proposed method are 7.7 mL for CO, with a crystal with a frequency decrease due to TMAF of 18.7 kHz, and of 19.2 mL for CO2, with a coated crystal with a frequency decrease due to TMAF of 13.6 kHz.The methodology can also be applied for analysis in a continuous stream of gas. For comparison between experiments performed by the injection of the sample, and when it was admitted continuously, a quartz crystal was coated with an amount of TMAF corresponding to a frequency decrease of 13.2 kHz. Known volumes of CO2 were injected as before, and a series of experiments was also performed, with a stream of gas with known % volume of CO2 in N2, flowing continuously through the crystal cell.The total flow was maintained at 50 cm3 min21, and proportions between CO2 and N2 were fixed with a gas proportion flowmeter. The equation of the calibration line for the injection method was 0.219 3 CO2(mL) + 8.85 (r2 = 0.992), while for the method with the continuous flow was 9.12 3CO2(%v/v) + 159 (r2 = 0.997).The analysis of a sample Table 1 Results of replicate analyses of a sample of the exhaust gases of a car, performed by the proposed method CO signal/ Hz (5.0 mL injected) CO2 signal/ Hz (8.6 mL injected) CO calibration line (7 standards) CO2 calibration line (7 standards) 56 59 55 59 DF = 0.952 3 DF = 0.376 3 54 58 CO(mL) 2 13.0 CO2(mL) 2 11.3 56 58 r2 = 0.996 r2 = 0.997 54 58 Table 2 Comparison between three methodologies, based on QCM, to monitor CO Methodology Advantages Disadvantages Method 1— With HgO and detection of Hg(0) No coating Conversion depends on temperature Continuous mercury background Crystal submitted to thermal shocks during cleaning Reducing gases interfere CO2 cannot be detected Method 2— With HgO and detection of CO2 Mercury background does not reach the crystal Interaction with coating is reversible CO2 can also be monitored SO2 and reducing gases do not intefere Coating Conversion depends on temperature Trap solution is needed Drying of the gas is necessary, even with dried samples Solid trap for CO2, during CO monitoring Method 3— With Ag2O and detection of CO2 No heating No background Interaction with coating is reversible No trap solution CO2 can also be monitored SO2, or NOx do not intefere.Traps included in the methodology can also dry the sample. Coating Solid traps 1452 Analyst, 1999, 124, 1449–1453with a concentration corresponding to the centroid of the calibration line for the method where samples are admitted continuously (10.6% CO2), could be performed by injecting 5.6 mL of the sample, which corresponded to a CO2 amount at the centroid of the calibration line for the methodology where samples are injected into the nitrogen flow.The concentration range of interest for CO analysis depends on the sample, and although in the environment the ‘safe’ limit of concentration is placed at 1000 ppm,8 which corresponds to an injection of 7.7 mL of the sample to attain the limit of detection of the proposed method, the concentration in cigarette smoking is 4 3 104 ppm8 which is easily analysed injecting a volume sample of at least 193 mL, and CO analysis in unpolluted and in ocean areas, which is 0.1 to 0.2 ppm, can only be achieved with a preconcentration stage.The amount of CO in gaseous fuels is much higher and depends on its origin. It ranges from a few percent 4–18% (v/v) in coal gas, 27% in blast furnace gas, and around 29% in producer gas, to 30.5–41% in water gas.9 CO concentration in gaseous fuels is thus able to be determined by the proposed method, by adjusting the injected volume to the quantity of CO within the range of the calibration curve, that is injecting as much as tens to hundreds of mL of the gaseous sample. Conclusions Starting from a published methodology for CO monitoring, two new approaches were tested, based on a QCM. Based on the comparison between the three methodologies shown inTable 2, the third methodology is highly recommended. Acknowledgements The authors acknowledge Eng. L. Tarelho for the analysis by non-dispersive infrared spectrometry. References 1 M. H. Ho, G. G. Guilbault and E. P. Scheide, Anal. Chem., 1982, 54, 1998. 2 N. Miura, H. Minamoto, G. Sakai and N. Yamazoe, Sens. Actuators B, 1991, 5, 211. 3 M. T. S. R. Gomes, A. C. Duarte and J. A. B. P. Oliveira, Anal. Chim. Acta, 1995, 300, 329. 4 M. T. S. R. Gomes, T. A. Rocha, A. C. Duarte and J. A. B. P. Oliveira, Anal. Chim. Acta, 1996, 335, 235. 5 Y. Yan and T. Bein, J. Am. Chem. Soc., 1995, 117, 9990. 6 A. W. Warner, D. B. Fraser and C. D. Stockbridge, IEEE Trans. Sonics Ultrasonics, 1965, June, 52. 7 E. A. Gerber, Proc. IEEE, 1966, 54, 103. 8 B. M. McCormac, Introduction to the scientific study of atmospheric pollution, D. Reidel Publishing Company, Dordrech, 1971, pp. 34–40. 9 H. M. Spiers, Technical data on fuel, The British National Committee World Power Conference, London, 1962, p. 255. Paper 9/04494I Analyst, 1999, 124, 1449–1453 1453

ISSN:0003-2654    

DOI:10.1039/a904494i

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Comparison of formats for the development of fiber-optic biosensors utilizing sol–gel derived materials entrapping fluorescently-labelled protein Kulwinder Flora and John D. Brennan* Department of Chemistry, McMaster University, Hamilton, Ontario, Canada L8S 4M1. E-mail: brennanj@mcmail.cis.mcmaster.ca Received 20th June 1999, Accepted 17th August 1999 The development of fiber-optic biosensors requires that a biorecognition element and a fluorescent reporter group be immobilized at or near the surface of an optical element such as a planar waveguide or optical fiber.In this study, we examined a model biorecognition element–reporter group couple consisting of human serum albumin that was site-selectively labelled at Cys 34 with iodoacetoxy-nitrobenzoxadiazole (HSA–NBD). The labelled protein was encapsulated into sol–gel derived materials that were prepared either as monoliths, as beads that were formed at the distal tip of a fused silica optical fiber, or as thin films that were dipcast along the length of a glass slide or optical fiber.For fiber-based studies, the entrapped protein was excited using a helium–cadmium laser that was launched into a single optical fiber, and emission was separated from the incident radiation using a perforated mirror beam-splitter, and detected using a monochromator–photomultiplier tube assembly. Changes in fluorescence intensity were generated by denaturant-induced conformational changes in the protein or by iodide quenching. The analytical parameters of merit for the different encapsulation formats, including minimum protein loading level, response time and limit-of-detection, were examined, as were factors such as protein accessibility, leaching and photobleaching. Overall, the results indicated that both beads and films were suitable for biosensor development.In both formats, a substantial fraction of the entrapped protein remained accessible, and the entrapped protein retained a large degree of conformational flexibility.Thin films showed the most rapid response times, and provided good detection limits for a model analyte. However, the entrapment of proteins into beads at the distal tip of fibers provided better signal-to-noise and signal-to-background ratios, and required less protein for preparation. Hence, beads appear to be the most viable method for interfacing of proteins to optical fibers. Introduction In the past few years there has been a significant increase in the number of reports describing the encapsulation of biological components into inorganic silicate matrices formed by the sol– gel processing method.1–3 In many cases, the biologicallydoped glasses (or biogels) have been able to maintain the activity of entrapped compounds such as enzymes3 and antibodies.4 Numerous studies have been reported regarding the function,5 structure,6 dynamics,7 accessibility,6,8 reaction kinetics,6a,9 initial stability10 and long-term stability11 of entrapped proteins.These studies have established that, in many cases, entrapped biological molecules retain their characteristic biochemical functionality and remain stable over periods of months.1,3b Given the stability imparted to entrapped proteins, and the advantages inherent in the use of a silica matrix (optical transparency, chemical inertness, processibility), biomaterials based on sol–gel entrapped proteins should be ideal for the development of fiber-optic biosensors.However, while there have been many reports of sensing strategies using sol–gel entrapped proteins, the majority of these studies involved glasses that were formed either as blocks (often referred to as monoliths) or thick films. The use of such formats is impractical for biosensor development for several reasons. Both monoliths and thick films age slowly over a period of several weeks or even months, resulting in alterations in the analytical response characteristics over time.4a,12 In addition, the long diffusional path for entry of analytes into the monoliths produces long response times, ranging from several minutes up to hours.6a Monoliths are also fragile and are prone to cracking due to hydration stress.13 Finally, it is difficult to interface monoliths to devices based on optical fibers in order to allow remote analysis to be done.To overcome these obstacles, it is necessary to entrap proteins into sol–gel derived glasses that are placed onto optical fibers.Sol–gel processed materials can be placed as beads at the distal tip of a fiber.14 Alternatively, protein-loaded thin films, of the order of 1 mm or less in thickness, have been spin-coated,15 dipcast16,17 or sprayed18 along planar substrates. However, only sol–gel derived films without proteins have been dipcast onto the distal end of an optical fiber.19 No reports currently exist describing the development of viable fiber-optic biosensors using dipcast thin films of protein-loaded glasses.In this study, a model biorecognition element–reporter group couple, consisting of human serum albumin (HSA) that was site-selectively labelled at Cys-34 with iodoacetoxynitrobenzoxadiazole (NBD), was entrapped into a variety of different sol–gel derived structures. The HSA–NBD system was chosen because the fluorescence signal is known to be sensitive to several factors, including conformational motions of the protein (allowing for denaturation studies)20 and the presence of external quenchers, which allows accessibility studies to be done.6,8 In addition, fluorescently-labelled HSA and bovine serum albumin have previously been characterized when entrapped in sol–gel derived monoliths, and these studies have indicated that the fluorescence of the entrapped protein is able to report on the local microenvironment within the glass.7 The labelled protein was encapsulated into three different types of structures that were each derived by a two-step sol–gel processing method: monoliths, beads that were placed at the distal tip of an optical fiber, or thin films that were cast onto This journal is © The Royal Society of Chemistry 1999.Analyst, 1999, 124, 1455–1462 1455either glass slides or a 2 cm length at the distal tip of a fused silica optical fiber that was part of a fiber-optic fluorimeter. In each case, the minimum amount of protein that was required to generate a fluorescence signal was determined.Factors such as protein leaching, photobleaching, accessibility and conformational flexibility were also examined. The analytical parameters of merit for the model fiber-optic biosensor, including response times and limit-of-detection for model analytes, are also reported. Experimental Chemicals Tetraethylorthosilicate (TEOS, 99+%) and human serum albumin (HSA) were purchased from Sigma (St. Louis, MO, USA).Urea (99.9%) and potassium iodide (99.9%) were purchased from Aldrich (Mississauga, ON, Canada). Iodoacetoxynitrobenzoxadiazole (IANBD) was obtained from Molecular Probes (Eugene, OR, USA). Guanidine hydrochloride (sequanol grade) was obtained from Pierce (Rockford, IL, USA). The Sephadex G-25 fine powder was supplied by Pharmacia Biotech (Uppsala, Sweden). All water was distilled and deionized using a Milli-Q 5-stage water purification system. All other chemicals were of analytical grade and were used as received.Instrumentation A schematic of the laser-based fiber-optic fluorimeter is shown in Fig. 1. The excitation source was a Liconix 4200 NB helium– cadmium laser operating at 28.5 mW with CW output at 441.6 nm. The laser radiation was passed through a 441.6 nm interference filter (3 nm FWHM, Andover Corporation, Salem, NH, USA) to eliminate extraneous plasma discharge. The light was then passed through a 1.8 OD neutral density filter to reduce the power to 45 mW, followed by passage through the rear of an UV-enhanced perforated mirror, which acted as a wavelength-independent beam-splitter.21 The laser radiation was then passed through a fused silica lens (50 mm diameter, 50 mm focal length, Melles Griot, Irvine, CA, USA) and focused into a fused silica optical fiber [400 mm core, 100 mm cladding, 0.40 numerical aperture (NA), Fiberguide Industries SPC- 400/500R, Stirling, NJ, USA], which was positioned using a XY- Z micrometer controlled translation stage (Newport, Santa Ana, CA, USA).The laser radiation was passed through the optical fiber, exciting the fluorescent sample that was located within a sol–gel derived bead or thin film that was located at the distal end of the fiber. Fluorescence emission originating at the distal end of the fiber was captured by the fiber and the fluorescence exiting the proximal end of the fiber was collimated by passage through the fused silica lens to produce a spot of ~ 1 cm diameter.The fluorescent spot was reflected from the front side of the UV enhanced mirror through a 475 nm long pass filter and then focused onto the entrance slit of a monochromator (ScienceTech 9010, 200 mm focal length, f/3.5) using a second lens (60 mm focal length, 30 mm diameter, Melles Griot), which produced a spot 1.2 mm in diameter at the slit. Fluorescence was detected using a Hamamatsu R928 photomultiplier tube located in an analog housing (ScienceTech PMM-02, Bridgewater, NJ, USA) operated at 950 V.The fluorescence signal was collected using a PC with software supplied by the manufacturer (SciSpec version 2.0 for Windows). Procedures Purification and labelling of protein. Human serum albumin was purified using Sephadex G-25 and then labelled according to standard procedures.20 The labelling efficiency was determined to be 90 ± 3% using e277 = 19 000 M21 cm21 and e472 = 23 000 M21 cm21 for NBD,22 and e277 = 36 000 M21 cm21 for HSA.23 The labelled protein was dissolved into 10 mm phosphate buffer containing 100 mm KCl at pH 7.2 for preparation of monoliths, or into 10 mm Tris buffer containing 100 mm KCl (pH 7.2) to make beads and thin films.The concentration of protein in the buffer was varied from 0.5 to 25 mm to examine the effect of protein level on the signal-to-noise levels for the various formats. Entrapment of proteins. Solutions of hydrolyzed TEOS were prepared as described elsewhere using a H2O:TEOS molar ratio of 4.0 for hydrolysis.6a Entrapment of protein into sol–gel derived monoliths was done as described previously, producing blocks with dimensions of 20 3 8 3 0.2 mm.6a These samples were aged for 15 d in air at 4 °C before testing.Deposition of sol–gel entrapped proteins onto optical fibers utilized fibers that had 2 cm of cladding stripped from the distal end using hexane. Both fibers and glass slides (32 3 8 3 1 mm) were cleaned by soaking in a 49% solution of HF for 1 min, followed by treatment with 1.0 m NaOH for 5 min and copious washing with distilled water.Preparation of sol–gel derived thin films was carried out by mixing 70 mL of the hydrolyzed TEOS with 70 mL of the HSA– NBD solution in an Eppendorf tube and then transferring the mixture to a casting well which consisted of a cylindrical hole drilled in a Teflon block (3 mm wide 3 25 mm deep) for fibers, or a rectangular well (25 310 33 mm) for glass slides.A 2 cm length of the fiber or glass slide was rapidly cast into the solution immediately after mixing, and was withdrawn at a speed of 4 mm min21 to form the film. The dipcast films were allowed to dry for between 30 min and 3 d before use. Film thickness was determined using an Alpha-step 500 profilometer. For sol–gel derived beads, 25 mL of hydrolyzed TEOS were mixed with 25 mL of the HSA–NBD solution in an Eppendorf tube whose lid had been previously cut in half and separated from the rest of the tube.The clean fiber tip was set into the sol–gel mixture immediately following mixing and was left for 24 h to allow the bead to harden around the tip of the fiber. The bead aged in air at room temperature for a further 1–3 d before removal from the Eppendorf tube, resulting in a protein-doped silicate bead attached to the end of the fiber. Fluorescence spectroscopy. Fluorescence spectra of proteins entrapped in sol–gel derived monoliths or films on glass slides were obtained using an SLM 8100 steady state fluorimeter (Spectronic Instruments, Westbury, NY, USA), as described in detail elsewhere.6a Fluorescence measurements of sol–gel derived beads and thin films cast onto optical fibers were made exclusively with the fiber-optic fluorimeter.In all cases samples were excited at 441.6 nm and emission was Fig. 1 Schematic of the laser-based fiber-optic fluorimeter used for examination of protein-doped sol–gel derived beads and thin films. 1456 Analyst, 1999, 124, 1455–1462collected from 480 to 650 nm in 1 nm increments using a 4 nm bandpass on the emission monochromator and a 0.1 s integration time per point. Appropriate blanks were collected for all samples tested, and these were subtracted from the sample emission spectra to remove scattering artifacts. Emission spectra were not corrected for distortions introduced by the emission monochromator throughput and PMT spectral response.Protein leaching and photobleaching. Protein leaching was examined by incubating the sample in 1 mL of a fresh buffer solution for a period of 24 h, after which the sample was removed and the fluorescence emissions from both the sample and the buffer solution were tested and compared to the initial values. Photobleaching of entrapped proteins was tested by exciting samples at 441.6 nm and recording the decrease in intensity at the emission wavelength maximum over a period of 240 s.The results are reported as per cent. loss of signal per minute. Response times and analyte detection limits. Response times were obtained for the interaction of analytes with proteins that were in solution or entrapped in monoliths, beads or thin films. Three different analytes were examined: urea, which is a neutral protein denaturant, guanidine hydrochloride, which is a positively charged denaturant, and iodide, which is a negatively charged species that quenches NBD fluorescence.A 10 mm concentration of protein was present in the buffer used to entrap the protein into monoliths and beads, while solutions containing 25 mm protein were used to form thin films. The fluorescence changes in the free and entrapped proteins were monitored over a period of several hours after introducing 7.0 m GdHCl, 10.0 m urea or 1.0 m iodide solution to the sample. In the case of iodide, the limit of detection was also determined for both free and entrapped proteins.The samples were titrated with 1.0 m iodide and emission spectra were collected after each addition. The emission spectra were integrated and the per cent. change in intensity was compared to the relative noise on each spectrum. Samples containing free proteins had the intensity values corrected for dilution factors. The LOD was based on the minimum amount of iodide required to cause a decrease in fluorescence intensity greater than 3 standard deviations in the baseline noise.Protein accessibility. The iodide quenching data were analyzed to determine the fraction of accessible protein and the Stern–Volmer quenching constant24 using the following equation: 8a F f F F f K Q k Q 0 0 1 1 1 ( ) ( ) [ ] [ ] - - = + = + i i sv q 0 t (1) where F0 is the fluorescence intensity in the absence of the quencher, F is the fluorescence intensity at a given molar concentration of quencher [Q], fi is the fraction of protein which is inaccessible to quencher, KSV is the Stern–Volmer quenching constant (m21), kq is the bimolecular quenching rate constant (m21 s21), and t0 is the fluorescence lifetime in the absence of quencher.Protein unfolding experiments. Both the free and entrapped proteins were subjected to chemical denaturation using guanidine hydrochloride to determine the effects of entrapment on protein conformational freedom. A 5 mm concentration of protein was used for denaturation studies of proteins in solution, and the experiment was performed by titrating the sample with GdHCl, and collecting spectra after each addition, as described elsewhere.8a,20 Denaturation of entrapped protein samples was carried out using 10 mm protein for monoliths or beads and 25 mm for thin films.Emission spectra were collected as a function of increasing levels of GdHCl, corrected for background contributions and then integrated over the emission band. Intensity was plotted against denaturant concentration to generate an unfolding curve, which was then used to ascertain protein conformational flexibility.Results and discussion Preparation of sol–gel entrapped proteins Characteristics of sol–gel derived beads and films. Preparation of samples in either bead or thin film formats required careful control of gelation and drying of the sol–gel derived biogels. Previous work for our group and others has demonstrated that both pH and buffer type and content affect gelation times.13,25 Increasing pH toward more alkaline conditions was found to decrease the gelation time.Both the level and type of buffer were also found to affect gelation times, with phosphate buffers resulting in substantially shorter gelation times as compared with Tris or PIPES buffers. It was determined that solutions that had longer gelation times produced superior films, since the viscosity of the solution was low enough to ensure thin films (500 ± 70 nm thick as determined by profilometry).Solutions which gelled in less than 20 min were more viscous and produced thicker films ( > 1 mm), which cracked extensively and peeled off the surface of fibers or slides. It was also noted that beads and thin films prepared with a Tris buffer system did not crack upon rehydration after aging, while all materials prepared with PBS did. For these reasons, all film and bead preparations were done with a 10 mm Tris buffer to achieve a gelation time on the order of 30 min, thereby avoiding large changes in solution viscosity during dipcasting.The other major factors affecting the quality of films were surface treatment prior to film casting and the casting speed. A series of different surface cleaning procedures were examined; however, it was determined that treatment with a 49% HF solution was required to obtain a surface to which the thin films easily adhered. Other treatments, such as activation with chromic acid or KOH, resulted in films that peeled from the surface of the slides and fibers.The casting speed also affected film thickness, and hence film quality. It is well known that slower casting speeds result in thinner films.26 In the present study, withdrawal speeds ranging from 1 mm min21 to 10 mm min21 were examined. A withdrawal speed of 4 mm min21 was found to be optimal since this was slow enough to produce thin films that remained on the fiber, but fast enough (ca. 7 min) to avoid large changes in solution viscosity during film casting.Films produced at other withdrawal speeds either cracked extensively during rehydration or peeled off the surface of the substrate. Films prepared at 4 mm min21 were easy to rehydrate, and could be used within 1 h of preparation and for up to 10 d after preparation. For sol–gel derived beads, the surface treatment did not have as significant an effect on the ability of the bead to adhere to the fiber tip. However, it was critical that the bead be aged at least 24 h before use so that immersing the bead between different solutions would not cause it to dislodge from the fiber.For most studies described below, the bead was aged for at least 3 d to increase the durability of the bead, and to avoid leaching of entrapped proteins. Characteristics of Fiber-Optic Instrument Fluorescence Spectra of Free and Entrapped Proteins. Fig. 2 shows the raw spectra obtained from a 10 mm solution of HSA–NBD and a blank using the fiber-optic spectrofluorimeter.The spectra obtained using the fiber-optic system showed a Analyst, 1999, 124, 1455–1462 1457relatively high background signal, and had a lower signal-tonoise (S/N) ratio, than spectra obtained using the commercial fluorimeter. The S/N ratio could be improved by increasing the laser power or by increasing the spectral collection time, but this method had the disadvantage of increasing the rate of photobleaching. The signal-to-background (S/B) ratio was independent of the laser intensity, indicating that the high background was likely due to scattering and/or emission from the same component in the fiber-optic system.A detailed examination of the various components of the fiber-optic fluorimeter indicated that it was the fiber that produced the background contribution. This contribution could be minimized by ensuring that the proximal and distal tips of the fiber were flat and smooth, and by optimizing the alignment of the system.Decreasing the length of the fiber also resulted in improvements in the S/B ratio; however, it was not possible to remove this contribution completely. It is possible that changing the type of fiber used may reduce this problem significantly, and we are currently examining a variety of different types of fiber for this purpose. Fig. 3 shows an overlay of the processed spectra obtained using the commercial and fiber-optic fluorimeters.Processing involved subtraction of the background contributions and normalization of the resulting spectra. In this case, the spectral contours are virtually identical, and the only significant differences are a lower S/N ratio and a slight alteration of the rising edge (blue region) of the spectrum obtained from the fiber-optic fluorimeter. The spectra obtained by the fiber-optic fluorimeter were reproducible in terms of emission wavelength maximum and spectral FWHM, and showed only a slight decrease in intensity over multiple sequential runs owing to a minor amount of photobleaching.The ability to acquire undistorted emission spectra is one of the key advantages to using the perforated mirror beamsplitter, since the UV enhanced mirror has a reflectivity that is wavelength independent in the visible range.21 The slight shift to longer wavelengths for the blue end of the spectrum is likely due to the presence of the 475 nm long-pass filter used in the emission path of the fluorimeter.Otherwise, the excellent correspondence between the two spectra indicates that the fiber-optic fluorimeter is well suited for spectral studies of species that absorb in the blue region of the spectrum. Table 1 compares the spectral information obtained for HSA– NBD in solution and when entrapped into sol–gel derived beads and thin films using both the commercial and the fiber-optic fluorimeter. Several interesting features are apparent.First, the emission maxima obtained for the protein in a given environment is the same for both instruments (within error). However, the spectral maximum of the protein in solution or beads is similar, while the spectral maxima of proteins entrapped in thin sol–gel derived films are slightly blue-shifted, consistent with previous observations that show blue shifts in the emission maxima of entrapped proteins.8a Second, the spectral fullwidth- at-half-maximum (FWHM) is in general a few nanometers narrower for spectra obtained using the fiber-optic fluorimeter, again showing the effects of the 475 nm long-pass filter.Third, the S/B ratio decreases on going from solution to a bead (or monolith) to a thin film. This is partially due to having a lower total amount of protein in monoliths and films, as compared to solution, and may also be due to increased scattering, particularly for monoliths and thin films in the commercial fluorimeter. Fourth, the S/N ratio, obtained at the emission maximum of 524 nm, decreases by about a factor of 2 on going from solution to a bead (or monolith), and by a further factor of 2 on going to the thin film format.In addition, the S/N for all formats is approximately a factor of 3 lower for the fiberbased instrument as compared to the commercial fluorimeter. In the worst case of thin films present on optical fibers, a S/N ratio of 8 is obtained when 10 mm of protein are entrapped, suggesting that experiments done in this format should use a higher level of entrapped protein.It should be noted that the relatively poor S/N ratios for both instruments are, in part, the result of short integration times. As described below, continuous excitation caused problems with photobleaching, particularly for entrapped samples. Thus, it was important to collect the spectra as quickly as possible while Fig. 2 Raw emission spectrum and background of HSA–NBD in solution obtained on the fiber-optic fluorimeter.Fig. 3 Overlay of processed spectra (sample minus background, normalized) for the commercial and fiber-optic fluorimeters. Table 1 Spectral characteristics of free and entrapped proteins at 10 mm concentration obtained on SLM 8100 and fiber-optic fluorimeter Sample Emission maximuma/ nm FWHMa/nm Protein LODb/mm Peak S/Nc Peak S/Bc SLM Solution 523 67 0.3 98 750 Monolith 522 70 0.6 47 100 Thin film 521 70 1.2 24 3 Fiber Solution 524 65 1.1 27 15 Bead 524 67 1.7 17 13 Thin film 521 64 3.8 8 5 a All emission maximum and FWHM values at ±1 nm for SLM, and ±2 nm for the fiber-optic fluorimeter.b LOD is based on a S/N ratio of 3. Errors in LOD values are ± 0.1 mm. c Peak S/N and S/B refers to signal-to-noise or signal-to-background ratios obtained at the emission maximum using a 0.1 s integration time and a 4 nm bandpass on the emission monochromator. 1458 Analyst, 1999, 124, 1455–1462maintaining an acceptable S/N ratio.Using the 0.1 s integration time it was possible to collect spectra in ~ 20 s. Use of a 1 s integration time (n.b. the ScienceTech PMT did not allow for integration times between 0.1 and 1 s) resulted in an unacceptable amount of photobleaching during the course of an experiment. Characteristics of entrapped proteins for sensing Protein leaching and photobleaching. Reliable quantitation of analyte-dependent fluorescence signals requires that both leaching of entrapped proteins and photobleaching be minimized or eliminated.Protein leaching was examined at various times after the gelation of beads and films, ranging from 1 h up to 3 d. Leaching did not occur from films, even when only 1 h was allowed for ageing. However, beads had to be aged for a minimum of 1 d in air to eliminate leaching. The present results refer to leaching of a medium sized protein (HSA, MW 67 000). It is possible that smaller proteins may show a higher degree of leaching, and hence may require longer ageing times to eliminate leaching.6a Photobleaching of entrapped protein samples was a more serious problem, and was approximately twice as great using the fiber-optic system as compared with the commercial fluorimeter.Beads and thin films showed similar levels of photobleaching and did not show changes in photobleaching efficiency with ageing time. The placement of neutral density filters into the path of the laser beam substantially reduced this problem, but even with 1.8 OD of filtering (the maximum OD to obtain an acceptable S/N ratio) the intensity decreased by ca. 1% per min. Hence, it was important to minimize the amount of time that the sample was exposed to the laser beam, resulting in short integration times during spectral collection. It should be noted that the NBD probe is not particularly photostable,27 and hence these results may represent a worst case scenario. The use of newer photostable probes28 should significantly reduce this problem, and may allow higher integration times and thus improve the S/N ratio as well. Protein accessibility.The entrapment of proteins into sol– gel derived glass materials often results in a fraction of protein being located in regions that are not accessible to analytes.6,8 To examine the accessibility of the HSA–NBD entrapped into monoliths, beads and thin films, the ability of iodide to quench NBD fluorescence was examined, and was compared to the quenching of the free protein.Table 2 shows the results of Stern–Volmer (SV) analysis of the quenching results obtained using equation (1). Surprisingly, all of the SV plots showed downward curvature (data not shown), resulting in the need to account for an inaccessible fraction of probe. Even when the protein was in solution, the best fit to the quenching data was obtained by setting the inaccessible fraction to 0.38. This unexpected result may be due to the presence of a mixture of protein conformations (i.e., native and denatured states), or may reflect electrostatic interactions between the negatively charged quencher and the charged residues in the vicinity of the NBD label.Previous studies using intrinsic protein fluorescence (from Trp 214) and CD spectra suggest that the labelled protein remains in a native state.20 Hence, electrostatic factors are most likely the cause of the downward curvature in the SV plots. Narazaki et al.29 have studied the interaction of charged quenchers with HSA that was labelled with acrylodan at site 34.Their results indicated that iodide was electrostatically attracted to residues lining the pocket surrounding cys-34. Hence, the local concentration of I2 during the initial stages of quenching may be increased owing to preconcentration, while at later stages the higher I2 concentration would result in increased charge screening and less electrostatic interactions, reducing the degree of quenching.Such a process would result in a downward curvature in a Stern–Volmer plot, resulting in the need to fit an inaccessible fraction of protein. One can compare the differences in fractional accessibility for free and entrapped proteins by taking the ratio of accessible fractions for each. This can be calculated from: f f f a i sample i solution (1 = - - ( ) ) 1 (2) where fa is the accessible fraction of entrapped protein relative to that in solution. These values are also listed in Table 2.The results indicate that the accessibility of the protein in both monoliths and beads is similar to that in solution, within experimental error. The lower degree of quenching for the protein in sol–gel derived beads (48% versus 80% at 1 m KI) is the result of a smaller KSV value. Decreases in Stern–Volmer quenching constants upon entrapment of protein have previously been reported,6a and are proposed to be the result of a higher viscosity for entrapped solvents.The accessibility of proteins in thin films is also decreased relative to the values in solution. This result suggests that some of the pores within the thin sol–gel derived films are too small to allow passage of even small analytes, such as KI. Response times and reversibility. The responses for interaction of neutral (urea), positively charged (GdHCl) and negatively charged (I2) species were examined using HSA– NBD both in solution and when entrapped into sol–gel derived monoliths, beads and thin films.Fig. 4 shows representative response curves obtained upon repeatedly moving a fiber, containing HSA–NBD entrapped in beads or thin films, between solutions containing 1.0 M I2 and fresh buffer. In both cases the signal is fully reversible, decreasing in the presence of I2 and recovering in the presence of fresh buffer. This result shows that the fiber-optic sensor is re-usable, which is a major advantage as compared to solution based assays.The results also show that the detection of I2 is highly reproducible, showing no significant deviation over 3 exposures to the I2 solution. The response time for samples entrapped in beads (full panel) is much longer than that for samples in thin films (inset panel). Table 3 shows the average response times (defined as the time to reach 95% of full signal) for all sample–analyte combinations. Two trends are apparent. First, for all entrapped species Table 2 Stern–Volmer analysis of iodide quenching for HSA–NBD in solution and when entrapped in sol–gel derived monoliths, beads and thin films Sample Decrease in intensity with 1 m KI (%) KSV/m21 fi a fa b Constant r2 Solution 80 14.6 ± 0.6 0.38 ± 0.03 1.00 ± 0.11 1.00 0.993 Monolith 80 13.6 ± 1.2 0.36 ± 0.04 1.03 ± 0.14 0.98 0.950 Bead 48 9.5 ± 0.2 0.42 ± 0.02 0.93 ± 0.08 1.02 0.995 Thin filmc 35 12.9 ± 1.0 0.54 ± 0.04 0.74 ± 0.08 1.01 0.970 a From equation (1).b From equation (2).c The thin film values reported are those obtained using the fiber-optic fluorimeter. Similar values were obtained for films cast onto glass slides. Analyst, 1999, 124, 1455–1462 1459the interaction of positively charged species is generally faster than is obtained from either neutral or negatively charged species. This is expected given that there is a net negative charge on the glass at neutral pH, which may help in preconcentrating positively charged analytes.6a Second, the response times tend to decrease as the thickness of the sol–gel derived matrix decreases, being slowest for monoliths and fastest for thin films.The average response time for films is always under 1 min (generally comparable to solution response times), suggesting that this format should be able to follow rapid changes in analyte concentration. The response time for beads was always on the order of 10 min, which is too slow for rapid monitoring applications but suitable for taking average readings over extended time periods.These slow response times are due to diffusion limitations when using larger silica samples.6a It is likely that the response times of bead can be improved by preparing smaller beads, or beads of a different shape. Alternatively, it may be possible to manipulate the pore sizes of the sol–gel matrix to increase analyte diffusion rates and hence response times.30 Work in these areas is currently underway in our laboratory.Detection limits and sensitivity. The limit-of-detection for iodide quenching was tested to provide a quantitative comparison of the utility of the various entrapment formats for sensor development. Table 3 shows the LOD values. These values were obtained by determining the amount of quencher required to produce a decrease in signal that was 3-fold greater than the standard deviation in the baseline noise. In all cases the entrapped proteins showed LOD values that were poorer than those obtained for the free protein.This is to be expected given that: (a) the signal measured has a lower S/N when using entrapped proteins as compared to proteins in solution; (b) some of the entrapped protein is not accessible; and (c) the silicate matrix is negatively charged, which may have reduced the ability of the analyte to enter the glass. Increasing the amount of protein entrapped, or improving the collection parameters to increase the S/N ratio, should provide improvements in the LOD.It should be noted that the LOD values for entrapped proteins were in reasonable agreement with one another (within error), with the thin film format being best and the bead format having a slightly poorer LOD than was obtained using monoliths. LOD experiments were not done with the other analytes since these act by denaturing the protein, and thus do not have a significant affect on protein fluorescence until a concentration of 1 m or higher is reached, even for free proteins in solution.Conformational freedom of entrapped proteins. A final consideration for entrapped proteins is whether they retain the ability to undergo large-scale conformational motions. This question is particularly important since many proteins must undergo conformational changes in order to bind an analyte,31 and in many cases the conformational change itself may be used to generate an analytical signal.32 To examine the conformational motions of HSA–NBD, both the free and entrapped proteins were subjected to chemical denaturation using guanidine hydrochloride.All samples showed changes in intensity with increasing levels of denaturant that were consistent with protein unfolding (data not shown). The spectral changes on protein unfolding are listed in Table 4. Several points merit further attention. First, the results show that it is possible to monitor the denaturant induced conformational changes of HSA using an NBD label at position 34, in agreement with other reports which used labels at position 34 to monitor HSA denaturation.20 Second, the unfolding curves clearly indicate that both free and entrapped proteins respond to the presence of denaturants, with the curves showing a clear decrease in intensity that corresponds to protein unfolding.Hence, the entrapped proteins retain some conformational flexibility, regardless of entrapment format. Third, the spectral characteristics for entrapped proteins are markedly different that those for the free protein.For example, the entrapped proteins show smaller changes in intensity, smaller red-shifts in emission Fig. 4 Changes in fluorescence intensity with time resulting from repeated movement of the fiber-optic sensor between 1.0 m I2 and fresh buffer solutions. In both cases, 10 mm of HSA–NBD is entrapped in sol–gel derived matrixes at the end of the optical fiber. The full panel shows the response obtained using sol–gel derived beads, with the fiber starting in buffer, moving to I2, and then moving back to buffer.The inset panel shows the response obtained using thin films, with the fiber starting in fresh buffer solution and moving between samples three times. Table 3 Response characteristics of free and entrapped proteins at 10 mm concentration obtained on the commercial and fiber-optic fluorimetera Sample GdHCl response time Urea response time KI response time LOD I2/mb Solution 4 min < 10 s < 10 s 0.010 ± 0.001 Monolith 17 ± 2 min 12 ± 2 min > 1 h 0.09 ± 0.01 Bead 6 ± 2 min 10 ± 2 min 8 ± 2 min 0.14 ± 0.03 Thin filmc < 10 s < 1 min < 10 s 0.08 ± 0.01 a Samples tested were 7.0 m GdHCl, 10.0 m urea and 1.0 m KI.b The limit of detection is defined as the minimum amount of KI that produced a decrease in intensity 3-fold greater that the standard deviation in the baseline noise. c The thin film values reported are those obtained using the fiber-optic fluorimeter.Similar values were obtained for films cast onto glass slides. Table 4 Spectral data for GdHCl denaturation of HSA–NBD in solution and when entrapped into sol–gel derived monoliths, beads or thin films Sample Emission maximum (native) /nm Emission maximum (denatured) a/nm Intensity decreasea (%) FWHMb (native) /nm FWHM (denatured)/ nm Solution 523c 546 84% 67 91 Monolith 522 532 73% 70 88 Bead 524 532 41% 67 73 Thin film 521 532 48% 64 71 a Total change in emission wavelength or intensity between 0 and 6.0 m GdHCl.b FWHM is full-width-at-half-maximum. c The error in solution and monolith emission maximum and FWHM values is ±1 nm, while that for beads and thin films is ±2 nm. 1460 Analyst, 1999, 124, 1455–1462maxima, and smaller increases in FWHM values on denaturation, particularly for proteins entrapped into beads or thin films. It is well known that the internal environment of sol–gel derived matrixes is significantly different than that of bulk solution,33 and that this can lead to changes in the spectral characteristics6a, 8a and dynamics7 of entrapped proteins.Overall, the results show that proteins do retain a degree of conformational freedom when entrapped into sol–gel derived materials, but that spectroscopic changes on unfolding are smaller for entrapped proteins. Potential of beads and thin films for fiber-optic sensor development Overall, the results show that entrapment of proteins into either sol–gel derived beads or thin films can provide a system which is amenable to biosensor development.Undistorted emission spectra could be obtained from samples located at the distal end of the fiber, providing information on both intensity and emission wavelength for entrapped species. Samples located in both beads and thin films were susceptible to photobleaching, but reliable quantitative data could still be obtained by using a rapid spectral collection time. Photobleaching problems may be overcome by using a more photostable probe, or by shifting to a pulsed source such as a laser diode or a nitrogen laser pumping a dye module.Entrapment of proteins in beads located at the distal tip of optical fibers resulted in better S/N and S/B ratios, and provided a somewhat higher proportion of accessible protein. In this case, less protein was required to obtain a useful signal, making the technique more amenable to cases where expensive protein reagents are used.On the other hand, entrapment of proteins into thin films resulted in faster response times, and in the case of iodide, provided detection limits which were comparable with those obtained for beads. The key problems with the use of thin films were that this format required higher concentrations of protein to obtain useful signals, and that the preparation technique that was employed was quite wasteful of protein. Generally, a total of 70 mL of a 25 mm protein solution was used to fill the casting well. However, only a fraction of this material was actually transferred onto the optical fiber, and the protein remaining in the casting well was discarded. Improved methods for forming thin films on irregular surfaces, such as that of an optical fiber, must be developed to overcome this problem.The major advantage of the use of entrapped proteins was the fact that they provided a fully reversible signal, and could be reused multiple times with no significant change in response characteristics.Solution-based assays can be carried out only once, and are not reversible. In the case of sensors using thin films, the response times were also on a par with solution assays. The only significant disadvantage was the lower LOD, which can likely be improved by simply increasing the amount of entrapped protein so as to increase the S/N ratio for signal acquisition. In cases where monitoring of intensity is done, one could also remove the monochromator and use filters to increase the light throughput, thereby increasing the S/N ratio and improving the LOD.A key observation was that proteins that were entrapped into either beads or thin films retained a substantial degree of conformational freedom. This finding is extremely important since in many cases, the binding of an analyte to a protein, particularly regulatory proteins, requires that the protein be able to undergo conformational changes.31,32 Our results demonstrate that entrapment of fluorescently-labelled regulatory proteins into either films or beads should be able to provide a platform for the development of reagentless biosensors for a range of analytes.Our research group is currently pursuing studies in this area. Conclusions A fiber-optic spectrofluorimeter using a He–Cd laser as a source and a perforated mirror as a beam-splitter was constructed. The instrument used a single fused silica optical fiber for delivery of excitation from the source to the sample, and emission from the sample to the detector.While the instrument produced higher background signals and more photobleaching than was obtained from a commercial cuvette-based fluorimeter, it was still capable of collecting useful emission spectra from samples located at the distal end of a single optical fiber. A model biomolecule–reporter group couple was produced by siteselectively labelling the single free cysteine (Cys-34) of HSA with the NBD probe.HSA–NBD was immobilized onto the distal end of the fiber via entrapment into sol–gel processed beads or thin films, resulting in a prototype of a reagentless fiber-optic biosensor. Both entrapment formats provided good analytical characteristics; however, entrapment into beads gave superior spectral characteristics and required less protein to be used during preparation, while films provided faster response times and slightly improved detection limits for a model analyte.Further work is required in the development of both beads and films, and we are currently examining the effect of polymer additives on the porosity and durability of both beads and films. The results will be described in a future manuscript. Acknowledgements The authors thank the Natural Sciences and Engineering Research Council of Canada and Research Corporation (Cottrell College Science Award) for funding of this work.References 1 J. D. Brennan, Appl. Spectrosc., 1999, 53, 106A. 2 (a) L. M. Ellerby, C. R. Nishida, F. Nishida, S. A. Yamanaka, B. Dunn, J. Selverstone Valentine and J. I. Zink, Science (Washington, D.C.), 1992, 255, 1113; (b) B. C. Dave, B. Dunn, J. S. Valentine and J. I. Zink, Anal. Chem., 1994, 66, 1120A. 3 (a) S. Braun, S. Rappoport, R. Zusman, D. Avnir and M. Ottolenghi, Mater. Lett., 1990, 10, 1; (b ) D. Avnir, S. Braun, O. Lev and M. Ottolenghi, Chem. Mater., 1994, 6, 1605. 4 (a) R. Wang, U. Narang, P. N. Prasad and F. V. Bright, Anal. Chem., 1993, 65, 2671; (b) A. Turniansky, D. Avnir, A. Bronshtein, N. Aharonson and M. Altstein, J. Sol-Gel Sci. Technol., 1996, 7, 135; (c) J. Livage, C. Roux, J. M. Da Costa, I. Desportes and J. F. Quinson, J. Sol-Gel Sci. Technol., 1996, 7, 45; (d) A. Bronshtein, N. Aharonson, D. Avnir, A. Turniansky and M. Altstein, Chem. Mater., 1997, 9, 2632. 5 F. Nishida, J. M. McKiernan, B. Dunn, J. I. Zink, C.J. Brinker and A. J. Hurd, J. Am. Ceram. Soc., 1995, 78, 1640. 6 (a) L. Zheng, W. R. Reid and J. D. Brennan, Anal. Chem., 1997, 69, 3940; (b) P. L. Edmiston, C. L. Wambolt, M. K. Smith and S. S. Saavedra, J. Collord. Interface Sci., 1994, 163, 395. 7 J. D. Jordan, R. A. Dunbar and F. V. Bright, Anal. Chem., 1995, 67, 2436. 8 (a) L. Zheng and J. D. Brennan, Analyst, 1998, 123, 1735; (b) C. L. Wambolt and S. S. Saavedra, J. Sol-Gel Sci. Technol., 1996, 7, 53. 9 C. Shen and N. M.Kostic, J. Am. Chem. Soc., 1997, 119, 1304. 10 (a) J. M. Miller, B. Dunn, J. S. Valentine and J. I. Zink, J. Non-Cryst. Solids, 1996, 220, 279; (b) Z. Chen, D. L. Kaplan, K. Yang, J. Kumar, K. A. Marx and S. K. Tripathy, J. Sol-Gel Sci. Technol., 1996, 7, 99. 11 (a) S. Braun, S. Shtelzer, S. Rappoport, D. Avnir and M. Ottolenghi, J. Non-Cryst. Solids, 1992, 147, 739; (b) S. Wu, L. M. Ellerby, J. S. Cohan, B. Dunn, M. A. El-Sayed, J. S. Valentine and J. I. Zink, Chem. Mater., 1993, 5, 115; (c) U.Narang, P. N. Prasad, F. V. Bright, K. Ramanathan, N. D. Kumar, B. D. Malhotra, M. N. Kamalasanan and S. Chandra, Anal. Chem., 1994, 66, 3139; (d) U. Narang, P. N. Prasad, Analyst, 1999, 124, 1455–1462 1461F. V. Bright, K. Kumar, N. D. Kumar, B. D. Malhotra, M. N. Kamalasanan and S. Chandra, Chem. Mater., 1994, 6, 1596; (e) S. A. Yamanaka, B. Dunn, J. S. Valentine and J. I. Zink, J. Am. Chem. Soc., 1995, 117, 9095; (f) D. J. Blyth, J. W. Aylott, D.J. Richardson and D. A. Russell, Analyst, 1995, 120, 2725; (g) J. S. Lundgren and F. V. Bright, Anal. Chem., 1996, 68, 3377; (h) J. W. Aylott, D. J. Richardson and D. A. Russell, Analyst, 1997, 122, 77; (i) F. Akbarian, A. Lin, B. Dunn, J. S. Valentine and J. I. Zink, J. Sol-Gel Sci. Technol., 1997, 8, 1067; (j) S. A. Yamanaka, F. Nishida, L. M. Ellerby, C. R. Nishida, B. Dunn, J. S. Valentine and J. I. Zink, Chem. Mater., 1992, 4, 495. 12 K. Flora and J. D. Brennan, Anal. Chem., 1998, 70, 4505. 13 K. Flora, M. A. Dabrowski, S. P. Musson and J. D. Brennan, Can. J. Chem., 1999, in the press. 14 (a) A. Navas Diaz and M. C. Ramos Peinado, Sens. Actuators B, 1997, 38-39, 426; (b) B. D. MacCraith, C. McDonagh, A. K. McEvoy, T. Butler, G. O’Keeffe and V. Murphy, J. Sol-Gel Sci. Technol., 1997, 8, 1053. 15 (a) J. W. Aylott, D. J. Richardson and D. A. Russell, Chem. Mater., 1997, 9, 2261; (b) U. Narang, P. N. Prasad and F. V. Bright, Chem. Mater., 1994, 6, 1596. 16 B. C. Dave, H. Soyez, J. M. Miller, B. Dunn, J. S. Valentine and J. I. Zink, Chem. Mater., 1995, 7, 1431. 17 B. C. Dave, J. M. Miller, B. Dunn, J. S. Valentine and J. I. Zink, J. Sol-Gel Sci. Technol., 1997, 8, 629. 18 J. D. Jordan, R. A. Dunbar and F. V. Bright, Anal. Chim. Acta, 1996, 332, 83. 19 B. D. MacCraith, Sens. Actuators B, 1993, 11, 29. 20 K. Flora, J. D. Brennan, G. A. Baker, M. A. Doody and F. V. Bright, Biophys. J., 1998, 75, 1084. 21 R. S. Brown, J. D. Brennan and U. J. Krull, Microchem. J., 1994, 50, 337. 22 S. Lin and W. S. Struve, Photochem. Photobiol., 1991, 54, 361. 23 G. A. Pico, Int. J. Biol. Macromol., 1997, 20, 63. 24 (a) M. R. Eftink and C. A. Ghiron, Biochemistry, 1976, 15, 672; (b) M. R. Eftink and C. A. Ghiron, Biochemistry, 1977, 16, 5546; (c) M. R. Eftink and C. A. Ghiron, Biochim. Biophys. Acta, 1987, 916, 343. 25 J. D. Brennan, J. S. Hartman, E. I. Ilnicki and M. Rakic, Chem. Mater., 1999, in the press. 26 M. Guglielmi, P. Colombo, F. Peron and L. Mancinelli Degli Esposti, J. Mater. Sci., 1992, 27, 5052. 27 R. Peters and K. Beck, Proc. Natl. Acad. Sci., 1983, 80, 7183. 28 R. P. Haugland, Molecular Probes Catalog, Molecular, Probes Inc., Eugene, OR, 1992, ch. 2. 29 R. Narazaki, T. Maruyama and M. Otagiri, Biochim. Biophys. Acta, 1997, 1338, 275. 30 C. J. Brinker and G. W. Scherer, Sol-Gel Science, Academic Press, New York, USA, 1989. 31 M. Gerstein, A. M. Lesk and C. Chothia, Biochemistry, 1994, 33, 6739. 32 (a) J. D. Brennan, J. Fluores., 1999, in the press; (b) H. W. Hellinga, and J. S. Marvin, Trends Biotechnol., 1998, 16, 183; (c) K. A. Giuliano, P. L. Post, K. M. Hahn and D. L. Taylor, Annu. Rev. Biophys. Biomol. Struct., 1995, 24, 405. 33 (a) S. Xu, L. Ballard, Y. J. Kim and J. Jonas, J. Phys. Chem., 1995, 99, 5787; (b) J.-P. Korb, A. Delville, S. Xu, G. Demeulenaere, P. Costa and J. Johas, J. Chem. Phys., 1994, 101, 7074. Paper 9/06308K 1462 Analyst, 1999, 124, 1455–1462

ISSN:0003-2654    

DOI:10.1039/a906308k

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Electropolymerised platinum porphyrin polymers for dissolved oxygen sensing A. Sheila Holmes-Smith,* Alan Hamill,† Michael Campbell and Mahesh Uttamlal Department of Physical Sciences, Glasgow Caledonian University, Cowcaddens Road, Glasgow, UK G4 0BA Received 10th June 1999, Accepted 16th August 1999 Fast response optical sensors for oxygen, based on luminescent electropolymerised porphyrin polymers, are described. The sensitivity of the electropolymerised polymers for oxygen detection was found to be dependent on the potential at which the polymers were formed.Results are presented for polymers formed at two different oxidation potentials and for two platinum porphyrins: Pt tetraphenylporphyrin (Pt-TPP) and Pt octaethylporphyrin (Pt-OEP). Sensitivities of the poly-Pt-TPP and poly-Pt-OEP films to dissolved oxygen, given by the Stern–Volmer quenching coefficients, were obtained from luminescence lifetime measurements and were 1.12 (mg l21)21 and 2.07 (mg l21)21, respectively.The response time, which was determined from luminescence intensity measurements, for the poly-Pt-OEP to a step change in oxygen concentration of 4.3 mg l21 to 38.6 mg l21 was 8 ms. The limit of detection obtained for the Pt-OEP based sensor was 0.06 mg l21. Introduction The determination of molecular oxygen in real time and in situ in both the gas and solution phase is of great importance in medical, environmental and industrial applications. In particular, the measurement of dissolved oxygen is used in industrial applications as a measure of water quality, a decrease in oxygen generally indicating the presence of organic waste.Ideally, oxygen sensors should not consume oxygen, should have a fast response time and have a sensitivity which is appropriate to the application area. The classical method for dissolved oxygen determination is using the Clark electrode.1 This system operates by measuring the current at an electrode surface by the selective reduction of O2.The electrodes tend to be fragile, consume oxygen and are susceptible to stirring effects. More recently optical oxygen sensors have been proposed most of which are based on quenching of the excited state of a suitable luminescent indicator trapped within a polymer matrix on the surface of an optical fibre. The measurement process is based on either the change in the luminescence intensity or lifetime of the indicator molecule as a function of oxygen concentration.In recent years the advantages of the latter method have come to be appreciated, being easier to calibrate as it is independent of indicator concentration and being less susceptible to changes in signal intensity due to instrumental fluctuations. Suitable indicators for application to oxygen sensing are transition metal complexes which have been studied extensively over the past few years. In particular, complexes based on Ru2+ 2–10 and metalloporphyrins11–20 have been reported.These indicators have a high quantum yield, a large Stokes shift and a long luminescent lifetime which makes them attractive for employment as oxygen sensitive indicators. The indicator support for preparing the sensor should have a high permeability to oxygen, good mechanical and chemical stability and should not interfere with the fluorescence measurement. The matrices which have been studied to date have been polyvinylchloride (PVC),12,14,17 silicone rubber,2–5,11,14–16 polystyrene, 12,14,15,17 sol-gel,8,13 poly(methyl methacrylate) (PMMA)9,14,18 and cellulose acetate butyrate (CAB).9,18 PVC, PMMA and CAB films, in general, have a plasticizing agent to aid the permeability of the polymers to oxygen.9,12 Unfortunately this type of sensing film has the tendency to leach the constituent components into the surrounding medium, particularly if the elements are soluble in this medium, thus making these devices only really suitable for sensor fabrication in a gaseous environment.Silicone rubber, which has a high permeability to oxygen, has been, to date, the preferred host matrix. However, Ru2+ complexes, being ionic, have a poor solubility in silicones and thus require modification before entrapment in a silicone matrix.2,4 The response time for oxygen sensors depends on the matrix in which the indicator molecule is embedded and also whether they are measuring gaseous or dissolved oxygen.In a gaseous environment response times are typically less than 1 s2,3,6 and for dissolved oxygen response times have been reported ranging from 5 s to 10 min.3,7 Oxygen sensitivity of these devices, which again depends on the matrix, is determined by the slope of the Stern–Volmer plot. In many cases non-linear Stern–Volmer plots are reported21 which are attributed to inhomogeneous indicator distribution in the polymer and can be analysed using modified Stern–Volmer equations.In this work optical sensors for oxygen are prepared by the electropolymerisation of metalloporphyrins. These polymers have the potential to be coated directly onto the end of an optical fibre coated with a thin conducting film. The films produced using this method are optically clear and robust. This paper discusses the advantages of using the electropolymerising technique for optical oxygen sensors based on the quenching of the luminescence lifetime of the metalloporphyrins Pt tetraphenylporphyrin (Pt-TPP) and Pt octaethylporphyrin (Pt- OEP).Experimental Pt-TPP and Pt-OEP were purchased from Midcentury Chemicals (Posen, IL, USA) and used without any further purification. All other chemicals were used as received from Aldrich (Gillingham, Dorset, UK) unless otherwise stated. All solvents were of reagent grade and deionised water was used throughout. † Present address: European Molecular Biology Laboratory, Heidelberg, Germany. This journal is © The Royal Society of Chemistry 1999.Analyst, 1999, 124, 1463–1466 1463An Oxford Instruments (Abington, Berks., UK) potentiostat was used in conjunction with an ABB Goerz SE790 X–Y chart recorder (Belmont Instruments, Glasgow, UK) for the electropolymerisation process which was performed using a conventional three electrode set-up. The working electrode was an indium (tin) oxide (ITO) glass slide approximately 1 3 2 cm (Balzers, Berkhamsted, Herts., UK), the reference electrode was Ag/Ag+ and the counter electrode platinum. The monomer stock solution contained 1 mM metalloporphyrin and 0.1 M tetrabutylammonium perchlorate (TBAP) in dry acetonitrile.The Pt-TPP polymers were produced by cycling the potential at a scan rate of 200 mV s21 for 10 cycles between 0.1 V and either 1.0 V or 1.2 V vs. Ag/ Ag+ with a final voltage of 0 V applied to the film for 2 min. To prepare the Pt-OEP polymers 0.05 M 2,6-di-tert-butylpyridine (DBUP), which is believed to act as a catalyst for the electropolymerisation process, was required to be added to the solution.The DBUP is thought to co-ordinate with the peripheral pyrrole groups thus catalysing the electropolymerisation process.22 The potential was scanned at 200 mV s21 for 10 cycles between 0.2 and 1.2 V before holding the potential at 1.2 V for 60 s (this is intended to force any oligomers or dimers in the vicinity of the electrode to polymerise) then returning to 0 V for 2 min.All solutions were degassed by bubbling with nitrogen both before and during the polymerisation process. The slides were washed in acetonitrile then deionised water, dried in nitrogen and stored in the dark prior to being used. Luminescence lifetime measurements were made using a laboratory-built system. It consisted of a iLEE UV100 nitrogen pumped dye laser (Laser 2000, Northampton, UK) as the excitation source with a pulse rate of 1 Hz. The dyes employed were Coumarin 515 and 540 (Exciton Inc., OH, USA) with wavelengths of 515 nm and 540 nm for excitation of Pt-TPP and Pt-OEP, respectively.A silica bifurcated fibre optic bundle (1500 mm) was used to transmit light to and from the sensing polymer which was attached at the common end of the bundle. The detector was a Hamamatsu (Enfield, London, UK) R928 side window photomultiplier tube with a 550 nm cut-off filter positioned in front of the window. Data acquisition was achieved using a Hewlett Packard (Bracknell, Berks., UK) 54540 digitising oscilloscope connected to a PC via a IEEE-488 interface.A dynamic data transfer software incorporating C++ as the high level language and a commercial software package, Testpoint™ (CEC, Keithley Instruments, Reading, Berks., UK), was developed to link programmes and facilitate analysis of the data. Steady state absorption and emission spectra were obtained using Perkin Elmer (Bucks., UK) Lambda II and LS-50B spectrometers respectively.Calibration of the sensors was achieved by mixing oxygen and nitrogen using flow meters and a mixing tube of 1 m length. The gas mixture was then bubbled through the deionised water containing the sensors for at least 10 min prior to each measurement. Response times were obtained using luminescence intensity measurements as the lifetime instrumentation used in this study cannot give a fast enough response due to the laser repetition rate and the averaging process used to obtain t.Here the pulsed laser source was replaced with an ultra-bright green LED, of peak wavelength 525 nm and intensity 6000 mcd at this wavelength (RS Components Ltd., Northants, UK). A short pass filter of cut-off wavelength 540 nm was inserted between the LED and fibre. The luminescence emission was detected using the same detection system as above but in this case the luminescence intensity was scanned with time. Deionised water saturated with either oxygen or nitrogen was fed via a syringe and plunger mechanism to the sensor situated in a chamber.The thickness of the films was determined using a laser scanning confocal microscope (LSCM) (Odyssey; Noran Instruments, Bicester, Oxon., UK). The laser source was a tuneable dye laser with a selected output wavelength of 515 nm. Results and discussion Fig. 1 shows the cyclic voltammograms for the formation of poly-Pt-TPP produced at the first oxidation wave (1.00 V) and the poly-Pt-OEP.The changing amplitude of the oxidation and reduction waves for each successive scan indicated that a film was forming on the electrode.23,24 Films formed in this way were optically clear, uniform and had reproducible spectroscopic properties. The films are also chemically robust, the polymers being insoluble when soaked in a variety of polar and non-polar solvents for several hours. The absorption and emission spectra of Pt-TPP both in solution and in polymeric form are given in Figs. 2A and 2B, respectively. The spectra show a small (2–3 nm) red shift in the spectra for the polymers. Similar spectral shifts were observed for Pt-OEP. No major change in the shape of the spectra indicates that there has been no change in the molecular structure of the porphyrin during the electropolymerisation process. Table 1 gives a summary of the absorption peak used for excitation and the emission peak observed from the polymers. Film thickness was determined using LSCM. The film thickness for 10 cycles was less than the optical resolution of the microscope which was ±0.5 mm.However, the thickness of films formed from 100 and 200 cycles was measured and a linear relationship between the number of scans and film thickness was assumed to estimate a film thickness of 0.2 mm for 10 potential cycles. Fig. 1 Cyclic voltammograms for the electropolymerisation of (A) Pt-TPP and (B) Pt-OEP. The arrow indicates the direction of the scan. 1464 Analyst, 1999, 124, 1463–1466The luminescence lifetime was determined by fitting an exponential function to the data: I t A t i i i ( ) exp( / ) = - Â t (1) where ti is the luminescence lifetime of the sample, Ai is a preexponential factor and i indicates the number of decay components required.The luminescence lifetime values obtained for the polymers in an aqueous environment are given in Table 1. A single exponential function was sufficient to fit the data in the absence of oxygen for both polymers and also in the presence of dissolved oxygen in the case of the Pt-TPP polymer.A biexponential function was required to describe the decay kinetics for poly-Pt-OEP in the presence of dissolved oxygen. From this analysis a mean weighted lifetime was calculated for use in the Stern–Volmer equation. The mean lifetime for poly- Pt-OEP in the presence of saturated dissolved oxygen is 1.75 ms. The sensitivity of the quenching process can be determined from the Stern–Volmer relationship:25 t t 0 2 1 = + Kq[ ] O (2) where t0 and t are the luminescence lifetime in the absence and presence of oxygen, respectively, [O2] is the oxygen concentration and Kq the Stern–Volmer quenching coefficient.Fig. 3 shows the Stern–Volmer plots obtained for the Pt-TPP films formed at different electropolymerisation potentials and it can be seen that an increase in sensitivity is observed for the films which were formed at the lower electropolymerisation potential.Fitting the data to eqn. (2) gives a good fit with r2 values of 0.998 and 0.997 and Stern–Volmer quenching coefficients of 1.12 (mg l21)21 and 0.79 (mg l21)21 for the polymers formed at the first and second oxidation potential, respectively. This would suggest that the polymers which are formed at the different oxidation potentials have different morphologies with the polymer formed at the lower oxidation potential having a higher permeability to oxygen.The Stern–Volmer plot for the poly-Pt-OEP is shown in Fig. 3. The figure clearly indicates a non-linear response and suggests a two site model is operating which would not be surprising given that a biexponential decay function was required to fit the luminescence decay in the presence of oxygen. In this case a modified Stern–Volmer plot can be used:25 t t t 0 0 2 1 1 - = + f K f q O [ ] (3) where f is the maximum mole fraction of dye molecules which are accessible to oxygen.The analysis of data using eqn. (3) gives a good agreement to the theory (r2 = 0.999) with a value for f Å 1 indicating that all the excited Pt-OEP molecules are equally likely to be quenched by oxygen. A similar trend was found for the same porphyrin in PVC, silicone and polystyrene matrices.13 However, analysing the data to 9.03 mg l21 oxygen (air saturation at 20 °C) gives a good fit to eqn. (2) (r2 = 0.998) with a value for Kq of 2.07 (mg l21)21 which is in reasonable agreement with the Kq value obtained when fitting the whole data set to eqn.(3) where the value for Kq is 1.94 (mg l21)21. The limit of detection for the Pt-OEP sensor proposed here is 0.06 mg l21 based on three times the standard deviation at zero dissolved oxygen. The response times for the Pt-TPP and Pt-OEP polymers for a step change in oxygen concentration are given in Table 2. The response time for the Pt-OEP electropolymerised films to changes in dissolved oxygen concentration is shown in Fig. 4. The values obtained for the response times are at the limit of the switching mechanism used to purge the sample container with Fig. 2 A: Absorption spectra for Pt-TPP; (a) toluene (b) polymer. B: Emission spectra for Pt-TPP; (a) toluene (b) polymer. Table 1 Absorption peak used for excitation, peak emission wavelength and lifetime values obtained for poly-Pt-TPP and poly-Pt-OEP Polymer labs/nm lem/nm t0/ms t1/ms (O2 saturated) t2/ms (O2 saturated) A1 (%) Pt-TPP 514 648 68.8 ± 5.0 1.2 ± 0.5 — 100 Pt-OEP 540 644 109.8 ± 4.0 2.7 ± 0.2 0.76 ± 0.03 37.4 Fig. 3 Stern–Volmer plot for poly-Pt-TPP: oxidation potential /, 1.00 V; -, 1.20 V and poly-Pt-OEP, :. Table 2 Stern–Volmer quenching constant and response times of the poly-porphyrins to dissolved oxygen. t—: Time taken for step change in oxygen concentration from 4.3 mg l21 to 38.6 mg l21; t–: time taken for step change in oxygen concentration from 38.6 mg l21 to 4.3 mg l21 Polymer Kq/(mg l21)21 t—/ms t–/ms Pt-TPP 1.12 20 80 Pt-OEP 2.09 8 19 Analyst, 1999, 124, 1463–1466 1465the saturated solutions.Response times of Ru2+ based sensors for dissolved oxygen range from seconds to minutes.3,7 The electropolymerised systems therefore represent some of the fastest response sensors for dissolved oxygen. The main reason for this is the thickness of the sensors proposed here, which is only 0.2 mm, and the high permeability of the polymers to oxygen. Porphyrins have been reported to suffer from photodegradation when employed as indicator molecules.13,14,16,17 The photostability of the electropolymerised polymers was examined under ‘normal’ operating conditions.Both polymers were exposed to the excitation light for Å 8 h per week over a 6 month period (total exposure time to excitation light Å 192 h). Between testing the polymers were stored under dry conditions and in the dark at room temperature. No change in the optical properties of the polymers was observed with the same response to oxygen within the limits of uncertainty being observed over this period.Over a 18 month period ( Å 580 h illumination) a reduction in the luminescence lifetime of 2.2% and 4.8% for the Pt-TPP and the Pt-OEP polymers, respectively, was observed. This is in contrast to that observed for platinum porphyrin molecules in other sensor matrices13,14,16,17 and suggests that the electropolymerisation process stabilises the porphyrin such that little photodegradation is evident.Conclusion Electropolymerisation of metalloporphyrins to give optical oxygen sensors has been demonstrated. Sensitivities of these films have been shown to be comparable to other optical sensors proposed for dissolved oxygen in the literature but have the potential advantage of being able to be bound directly onto the end of an optical fibre. The response times of the sensors are faster than already reported and this is a consequence of the thin polymer structure produced. Work is underway to immobilise the molecules directly onto the fibre tip before assessing the sensor in a real environmental application.Acknowledgements AH wishes to thank Glasgow Caledonian University for assistance in the form of a University funded studentship. Also, thanks to Dr Craig Daly, Institute of Biological and Life Sciences, Glasgow University for the LSCM measurements. References 1 L. C. Clark, Trans. Am. Soc.Artificial Internal Organs, 1956, 2, 41. 2 I. Klimant and O. S. Wolfbeis, Anal. Chem., 1995, 67, 3160. 3 I. Klimant, P. Belser and O. S. Wolfbeis, Talanta, 1994, 41, 985. 4 J. R. Bacon and J. N. Demas, Anal. Chem., 1987, 59, 2780. 5 M. E. Lippitsch, J. Pusterhofer, M. J. P. Leiner and O. S. Wolfbeis, Anal. Chim. Acta, 1988, 205, 1. 6 A. Mills and M. Thomas, Analyst, 1997, 122, 63. 7 C. Preininger, I. Klimant and O. S. Wolfbeis, Anal. Chem., 1994, 66, 1841. 8 A. K. McEvoy, C. M.McDonagh and B. D. MacCraith, Analyst, 1996, 121, 785. 9 A. Mills and M. D. Thomas, Analyst, 1998, 123, 1135. 10 Z. Rosenzweig and R. Kopelman, Anal. Chem., 1995, 67, 2650. 11 W. Xu, K. A. Kneas, J. N. Demas and B. A. DeGraff, Anal. Chem., 1996, 68, 2605. 12 P. Hartmann and W. Trettnak, Anal. Chem., 1996, 68, 2615. 13 S.-K. Lee and I. Okura, Analyst, 1997, 122, 81. 14 P. M. Gewehr and D. T. Delpy, Med. Biol. Eng. Comput., 1993, 31, 11. 15 D. B. Papkovsky, J. Olah, I. V. Troyanovsky, N. A. Sadovsky, V. D. Rumyantseva, A. F. Mironov, A. I. Yaropolov and A. P. Savitsky, Biosens. Bioelectron., 1991, 7, 199. 16 W. W.-S. Lee, K.-Y. Wong, X.-M. Li, Y.-B. Leung, C.-S. Chan and K. S. Chan, J. Mater. Chem., 1993, 3, 1031. 17 S.-K. Lee and I. Okura, Spectrochim. Acta, Part A, 1998, 54, 91. 18 A. Mills and A. Lepre, Anal. Chem., 1997, 69, 4653. 19 I. Klimant, M. Kühl, R. N. Glud and G. Holst, Sens. Actuators, B, 1997, 38–39, 29. 20 D. B. Papkovsky, Sens. Actuators, B, 1993, 11, 293. 21 J. N. Demas, B. A. DeGraff and W. Xu, Anal. Chem., 1995, 67, 1377. 22 T. J. Savenije, R. B. M. Koehorst and T. J. Schaafsma, J. Phys. Chem. B, 1997, 101, 720. 23 K. A. Macor and T. G. Spiro, J. Am. Chem. Soc., 1983, 105, 5601. 24 K. A. Macor and T. G. Spiro, J. Electroanal. Chem., 1984, 163, 223. 25 J. R. Lakowicz, Principles of Fluorescence Spectroscopy, Plenum Press, New York, 1986, ch. 9, pp. 257–301. Paper 9/04634H Fig. 4 Response of the electropolymerised Pt-OEP polymer to changes in dissolved oxygen concentration by switching between nitrogen and oxygen saturated aqueous environments. 1466 Analyst, 1999, 124, 1463–1466

ISSN:0003-2654    

DOI:10.1039/a904634h

出版商:RSC    

年代:1999

数据来源: RSC