ANALYTICAL PROCEEDINGS. OCTOBER 1989, VOL 26 329 International Symposium on Electroanalysis in Biomedical, Environmental and Industrial Sciences ~~ The following are summaries of three of the papers presented at a Joint Meeting of the Electroanalytical Group and the Electrochemistry Group of the Faraday Division held on April 1 Ith-I4th, 1989, in Loughborough University of Technology. Arrays of Electrodes for Multi-component Analysis W. E. van der Linden, M. Bos and A. Bos laboratory for Chemical Analysis-CT, University of Twente, NL-7500 AE Enschede, The Netherlands In a discussion of arrays of electrodes it is useful to differentiate between voltammetric or amperometric sensing techniques on the one hand and potentiometric techniques on the other hand, because the objectives for applying arrays in these two instances are often quite different. Voltammetry and Arnperometry Most of the recent publications on arrays of electrodes deal with applications in voltammetry and amperometry.In these applications the arrays almost always consist of assemblies of (ultra) microelectrodes, with diameters generally much less than 50 pm, which have interesting specific properties. Such microelectrodes have, for instance, the following advantages: firstly, high mass transfer rates resulting from hemispherical rather than linear diffusion; secondly, less dependence of the signal on the conductivity of the solution because of the very small currents passing through the electrodes; and thirdly, reduced flow dependence when used in streaming media.l-3 Furthermore, it has been demonstrated by Cope and Tallmanj that the replacement of a traditional macroelectrode by an array of microelectrodes results in an enhancement of the signal.This can be attributed partly to the so-called edge effect, which is based on the fact that for flat inlaid electrodes the Cottrellian current, due to planar diffusion, is augmented by a constant term proportional to the perimeter of the flat electrode.5 The contribution of this effect becomes increasingly important as the perimeter to area ratio of microelectrodes increases. Another contribution to signal enhancement in flowing media is based on a depletion layer recharge which occurs when depletion zones generated by upstream electrodes during their passage over insulating regions are replenished by diffusion.6 This signal enhancement, together with the dimin- ished flow dependence, leads to a decrease in the limit of detection.I n the arrays of electrodes used in voltammetry, discussed so far, the separate electrodes are interconnected and only one current is measured for the whole ensemble. Similar to the approach in photodiode arrays, where each single diode generates a separate signal, it is also possible to apply different voltages to each single electrode of the array and to register the corresponding individual currents. This would allow the registration of a complete voltammogram in an extremely short period of time allowing on-line applications such as in chromatography or flow injection analysis. In the literature only one such application has been reported yet,7 and some have announced that work in this field is in progress.8 With regard to the construction of these voltammetic array electrodes three main approaches have been described: 1, embedding of (carbon) fibres in a polymeric matrix’.’ or in gIass1().11; 2, etching holes in a resist coating, an extreme example of this approach being presented by Hepel and Osteryoungl’ who prepared an electrode assembly consisting of over 1 million active electrodes per cm’; 3, direct vapour deposition or sputtering of electrode material.13 Procedures 2 and 3 can be combined as is common practice in the microlithographic fabrication of all types of integrated circuits.Moreover, the sputtered material, e . g . , aluminium, can be electroplated by more noble metals such as platinum.14 Potentiometry As discussed above the use of array electrodes in voltammetry and amperometry is mostly associated with the specific behaviour of microelectrodes; this is not necessarily so for potentiometry.Of course, miniaturisation of single (ion-selec- tive) electrodes will allow the construction of more manageable arrays which can be used for different types of in vivo measurements. 15 The development of ion-selective field effect transistors (ISFETs) is important in this respect, but, unfortu- nately, the high expectations for these types of electrodes have not yet been completely realised. The major problems encoun- tered in the construction of suitable and robust ISFETs are of a purely technological nature such as, for instance, encapsula- tion of the devices.Hopefully, some solutions will be found in the near future because the basic principle of the ISFET, and the available integrated circuit technology for the large scale production of complex arrays, make them very attractive. Chemometric Approach The particular interest in combining ion-selective electrodes in arrays is related to the availability of simultaneous output that can be achieved rather than in miniaturisation as such. This has been recognised recently by Otto and Thomas,l6 Beebe ez al. 17 and Beebe and Kowalski. 18 The last two authors claim that the use of sparingly selective electrodes in combination with a chemometric approach can result in multi-component analyses that are superior, in many cases, to those obtained when the electrodes are highly selective.I t was also correctly mentioned by them that for the mathematical procedures to have some real applicability, the stability and reproducibility of the sensor is more important than its selectivity. The chemometric approach chosen by Otto and Thomas16 was based on multiple linear regression and partial least330 ANALYTICAL PROCEEDINGS. OCTOBER 1989. VOL 26 squares for the construction of a calibration model based on variation of the two parameters in the extended Nikolsky equation, i.e., the Eo, and Kjk,, the selectivity factor of the jth electrode for the kth analyte with respect to the lth interfering ion. Beebe et ~ 1 . 1 7 have proposed a non-linear regression procedure in which not only the Eo and the selectivity factors are allowed to vary, but also the a priori unknown slope is considered as a variable which has to be estimated simul- taneously.In a very recent paper Beebe and Kowalski’g pursued this line and introduced a non-parametric multivariate technique called “projection pursuit regression.” It allows the calibration of a system without a priori information about the functional relations between the responses and the concentra- tions. The method can provide both this functional relationship (logarithmic, linear, parabolic, etc.) and the values of the descriptive model parameters. Once calibrated, precise esti- mates for the unknown concentrations of multi-component samples can be calculated. In the last section an alternative approach is presented based on pattern recognition by means of a so called neural network.19 Fig. 1. Schematic model of one neuron: Oj is output or neuron j ; w,, is the weight factor associated with the output from neuronj and used as input for neuron i; neti is the input function; ai is the state of activation; F is the activation hnction;fis the output function. a* and O* refer to the new state of activation and output, respectively Application of a Neural Network A neural network consists of an assembly of interrelated neurons that can be defined as processing units. Within each single neuronj the total operation can be sub-divided into three steps, each characterised by a mathematical expression. Firstly, an input function, which allows the calculation of an effective stimulus or input, net,, of the neuron.Often such an input function is simply a weighted summation of the outputs of the activating neurons (Fig. 1). Secondly, an activation function that allows the calculation a new state of activation, a,(t), from the previous state, u,(t - 1) and the input, net,. This function determines the time dependent response of the neuron on changes in the input. Finally, an output function, O,, that calculates the output signal of the neuron given a certain state of activation. These three functions completely charac- terise the behaviour of a single neuron. However, the n n n w w W Fig. 2. one neuron. Note the layered structure Schematic diagram of neural network. Each circle represents operation of the network as an ensemble is not simply determined by the single neurons and the inputs, but particu- larly by the mutual arrangement of the neurons.It is, in fact, just this topology and the values of the weight factors associated with each input of the neurons that determine the basic properties and “knowledge” of the network. A schematic picture of a neural network is presented in Fig. 2. The set of weight factors is found by presenting experimental data to the network and applying learning rule heuristics. In this aspect the network can be considered to be an adaptive system. The main focus of the work carried out in our laboratory was on the application of complex networks, using as learning heuristics the so-called “backward error propagation rule,” also called the “generalised delta rule.” In this case the network is separated in several layers; neurons from one layer have only inputs from preceding layers, mostly even from one and the same preceding layer.The procedure consists of a repeated sequence of two passes through the network. In the first step, or forward pass, a certain pattern of inputs is supplied to the first layer of a network with arbitrary weight factors and propagated through it until the final outputs of the last layer are calculated. Then the generated output is compared with the desired output (target) and an error is calculated. In the second step, or backward pass, the weight factors associated with the inputs of the last layer are adapted to minimise this error. This adaptation is propagated layer by layer backwards through the network.When coming at the input layer the procedure is repeated with the same input pattern. This forward and backward propagation is continued until an acceptable set of weight factors is obtained. The procedure is consecutively repeated for all patterns until the network can generate adequate outputs for all of these patterns. The number of iterations required can be quite large and so the efficiency of the calculation procedure, i.e., the activation function, is very important. Because in the adaptation of the t t + 2 Fig. 3. Recurrent network used for the simulation of Nikolsky type of behaviour: U, is potential; U,, is standard potential; a, is activity in the solution for ion i; K,, is the selectivity factor for ion i with respect to interferent j : n, is the charge of the ion.t, t + 1 and t + 2 refer to consecutive steps in the iteration procedureANALYTICAL PROCEEDINGS. OCTOBER 1989. VOL 26 weight factors the derivative of the activating function is used, it is important to choose a function, the derivative of which can be easily calculated. A sigmoid function of the form upi = 1/{ 1 + exp[-netpi]} fulfils the requirement because the derivative can be simply expressed as dup,/dnetpi = up, (1 - upi) where the value of the activation state, api, has been calculated alrcady in the forward pass. In a later stage the activation function was slightly adapted by introducing a bias factor, 0, and a “temperature factor,” T , that can be used to vary the learning rate api = 1/{ 1 + exp[( -netp, + @)/TI} Preliminary results were obtained with an array consisting of a potassium, a calcium, a nitrate, a chloride and a pH electrode.For mixtures with concentrations of about 10-4 M the correct concentrations were found within about 5-10%. This is very acceptable because for single ion measurements accuracies, when expressed in concentration units, are generally not better than 4%. A neural network can also be used for parameter estima- tions. So, a recurrent network was used for simulating the Nikolsky equation (Fig. 3). By measuring the potential values of the various electrodes for mixtures of known concentrations (activities) the relevant values of the standard potential, U,,, the slope, S, and the selectivity factors, K , can be found. The advantage of this approach over existing procedures is that the information can be obtained from solutions that can contain all the ions at the same time.Therefore, the results may be considered to be more realistic. Conclusions Arrays of electrodes form an interesting field of research with great potentials for practical applications. More specifically the further development of chemometric approaches deserve more attention when the electrodes show some selectivity but still a strong interaction occurs: it was shown that neural networks have very interesting features in this respect. 33 1 Linkage of this development to another area of growing interest, i.e., process analytical chemistry, means that one other useful application of arrays has to be mentioned, that is, the redundancy that can be built in.The operational lifetime of the array device can thus be extended until the last electrode of the set of equivalent electrodes breaks down. This is of particular interest if the device has to be installed in an inaccessible place or has to be implanted in a living body. 1 . 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 1s. 16. 17. 18. 19. References Caudill. W. L.. Howell. J. 0.. and Wightrnan, R. M.. Anal. Chem., 1982. 54. 2532. Thorrnann. W.. van den Bosch. P., and Bond. A. M.. Anal. Chem., 1985. 57. 2764. Khoo, S. B., Gunashingharn. H.. Ang. K. P., and Tay, B. T.. J. Electroanal. Chem.. 1987. 216, 115. Cope. D. K.. and Tallman. D. E.. J. Electroanal. Chem., 1985. 188.21. Oldharn. K. B.. J . Electroanal. Chem.. 1981. 122, 1. Cope. D. K.. and Tallman.D. E.. J. Electroanal. Chem.. 1986, 205. 101. Matson, W. R.. Garnache. P. G.. Beal. M. F.. and Bird. E. D.. Life Sci.. 1987. 41, 905. DeAbreu. M.. and Purdy. W. C.. Anal. Chem.. 1987.59.204. Belal. F.. and Anderson. J. L.. Analysr. 1985. 110. 1493. Aoki. K., Akimoto, K.. Matsuda. H.. and Osteryoung. J.. J. Elecrroanal. Chem.. 1984. 171, 219. Bond, A. M., Fleischrnann, M.. and Robinson, J., J. Electro- unal. Chem.. 1984. 180. 257. Hepel, T.. and Osteryoung, J., J. Electrochem. SOC.. 1986. 133. 752. Kittlesen. G. P.. White. H. S.. and Wrighton. M. S., J. Am. Chem. SOC.. 1984, 106. 7389. Glavina. P. G.. and Harrison. D. J.. Can. J . Chem.. 1987.65. 1072. Wise. K. D.. Angell, J . B.. and Sturr. A.. IEEE Trans. Rio-Med. Eng.. 1970, 17, 238. Otto. M., and Thomas, J . D.R.. Anal. Chem., 1985,57,2647. Beebe. K., Uerz. D.. Sandifer. J., and Kowalski. B.. Anal. Chem.. 1988. 60. 66. Bccbe. K.. and Kowalski. B.. Anal. Chem.. 1988. 60. 2273. Bos. A.. MSc Thesis. University of Twente. The Netherlands, 1988. Polyethers as Potentiometric Sensors J. D. R. Thomas School of Chemistrv and Amlied Chemistrv, University of Wales College of Cardiff, P.O. Box 912, Cardiff CF1 3TB Polyethers owe their function as potentiometric sensors to their ability to crown, grasp and coil around analates. Such complexation is facilitated by these neutral organic molecules carrying a sequence of localised charges of sufficient energy to form ion - dipole bonds with appropriate cations. The conformation of the molecule permits a solvation type shell around the cation which effectively replaces the ion hydration shell.The charged cationic complex so formed is electrically balanced by anions. The impetus for studies on polyethers as ion sensors is attributable to the observation of Moore and Pressman’ that valinomycin is capable of actively transporting potassium across rat mitochondria membranes. This led to the very successful valinomycin sensor for potassium.?.3 It is also relevant that Pedersen,j in 1967, described the function of crown ethers in promoting the dissolution of salts in which they are otherwise insoluble. However. cation adducts of polyethylene glycols had already been known for several years,S-’ and it was this category of polyethers that was first exploited8 in ion-selective electrodes (ISEs). The special properties of polyethylene glycols in this application is attributable to the alkoxylate units (AOUs) in the polyalkoxylate complex assuming a tight helical conformation, of appropriate ring radius for holding the ion in a cage of oxygen atoms of the AOUs, through ion - dipole interaction.8 For the particular case of barium ions complexing with an NP, a nonylphenoxypolyethoxylate (Antarox CO-880 with 30 ethoxylate units or EOUs), 12 EOUs are involved in holding the Baz+ ions in a tight helical arrangement with a ring radius of about 1.3 A, the cage for the barium ions being formed by the 12 oxygens in two loops of 6 EOUs each.8 For the crown ethers, some of the earliest ISE studies were by Rechnitz and Eyaly o n dicyclohexyl- and dibenzo-18-crown- 6 ligands synthesised by Pedersen .4 Shortly afterwards, Petranek and Rybal(j.11 assessed various types of crown-6 and larger crown compounds for potassium ion-sensing and favoured dimethyldibenzo-30-crown-10 with dipentyl phthalate in PVC as the best of the series.Although these and332 ANALYTICAL PROCEEDINGS, OCTOBER 1989. VOL 26 later studies on crown ether sensors have indicated that the selectivity for potassium does not exceed that of valinomycin, researches on pol yethers as potentiometric sensors have produced many interesting features as well as new directions in sensing. In the UWCC laboratories, such studies exploit the ability to coil by polyalkoxylate to crown by various crown ethers, and to grasp by certain “half-crown” acyclic pol yether types. Potentiometric Sensing Based on Polyalkoxylates The NP, Antarox CO-880, forms the basis of a very good barium ISE.8-12 This is in accordance with the bulk extraction constant, K M , data for the barium complex obtained for extractions into the organic phase by both picrate13 and dipicrylamine.Within each polyalkoxylate metal ion group, the barium complex systems provide the best PVC based ISEs,l2-14 and this is in compliance with the solvent extraction data. For the alkali metal cations, it is interesting that rubidium and potassium are the most strongly extracted alkali metal - polyalkoxylate complexes.13 Thus, these interfere more strongly with barium ISE behaviour than does sodium.12-14 Although not predictable from solvent extraction data, good lithium ISEs can be obtained from some of the polyalkoxylate systems, especially for an ISE membrane consisting of Ba.(PPG 1025)o,69.(TPB)2 plus dioctylphenyl phosphonate in PVC.15 The possibility of such a lithium ISE was indicated by the high lithium interference of a calcium ISE based on the calcium propoxylate anomer16 but, in the event, the best lithium ISE qualities were obtained with the tetraphenylborate (TPB) - barium propoxylate complex as sensor.In relation to the association of potentiometric response with complex stability, the facile conversion of an Antarox CO-880 based barium ISE to a thallium(1) electrode17 is interesting. This is effected by soaking the barium electrode in thallium(1) nitrate for 24 h, such conversion being consistent with dipicrylamine solvent extraction data.14 Of course, there must also be favourable kinetics for the exchange of ions for such conversion and to provide the charge transfer mechanism for ion sensing.Barium ISEs find little application in terms of actual barium determinations, but they can be used for the indirect determi- nation of sulphate,l?.1”-2” especially for the titration stage of the oxygen flask determination of sulphur in organic com- pounds.12 They can also be used for the determination of lithium.*5 Apart from being set up as conventional PVC membrane ISEs with inner filling solutions, the barium ISEs based on Antarox CO-880 can be of all solid state design, whereby the PVC membrane with the sensing material is applied to the outer surface of an epoxy resin and the filling solution eliminated.?() Corresponding lead(I1) ISEs have also been made from complexes of the metal with polyalkoxylates, and the most selective was found to be based on Pb.Antarox CO-880.TPB sensor with 2-nitrophenyl phenyl ether as plasticising solvent mediator.” However, this is not as suitable for use as an indicating system in sulphate titrations, as the lead(I1) titrant system is susceptible to interference by oxidants, such as hydrogen peroxide.21 A large number of metal polyalkoxylates have been characterised in association with the ion-sensing studies.21.22 Sensing of Non-ionic Surfactants An important application of barium ISEs is in the analysis of non-ionic surfactants,23-25 including the measurement of their critical micelle concentration23~~4 (CMC).The observed poten- tiometric response is an increase in the emf of the barium ISE of up to about 100 mV, according to the amount of added alkoxylate in the 2 x to range.23 The response is linear with log[alkoxylate], but is characterised by a break in the linearity, which it attributed to the CMC of this class of non-ionic surfactants. The barium ISE, based on the barium complex with Antarox CO-430 (an NP with 4 EOUs), is superior in its response to non-ionic surfactants.This gave good recoveries in the analysis of Dobanol25-7, Lutensol A07, and Synperonic 7 (polyethoxylates with 7 EOUs and alkyl hydrophobes of C12H25 to C1sH31) in detergent powders’s with relative standard deviations of between 0.8 and 4.0%. With regard to CMC, the inflections observed in the emf versus log[alkoxylate] graphs are quite definite and the breaks’j compare favourably with literature CMC values26.’7 and with a derived formula26 h(l@ X CMC) = A n + B .. . . (1) where A and B are constants for a particular hyrophobic group and n is the average number of EOUs in the molecule. There is a correlation between complexing tendencies of polyalkoxylates and barium, as evinced by solvent extraction data, and electrode qualities with respect to both cationl3.14 and non-ionic surfactant14 sensing. Also, factors such as those impressed by the substituent tail attached to the end of the alkoxylate chain can be involved.14 Scanning electron micro- scopy, ESCA spectra and XRF data, as well as radiotracer experiments, indicate that there are barium - polyalkoxylate interactions between the membrane and solutions, thus point- ing to a mechanism for potentiometric response to non-ionic surfactants in solution.28 However, with regard to ion permea- tion through membranes, there is only limited permeation29 of radioactive barium ions compared with the high levels previ- ously observed for permeation through ion-exchanger type organophosphate-type membranes.3” For the permeation of radiotracers through membranes under applied potential, the barium polyalkoxylate complex breaks down as shown by the rapid drop in current flowing through the membrane.31 Current flow ceases after just one reversal of potential.31 Potentiometric Ion Sensing with Crown Ethers Researches on crown ether derivatives for ion sensing are now quite widespread.Thus, Pungor’s group32 have studied the synthesis and application of various mono- and bis-crown ether derivatives with urethane and urea linkages. In studies33-35 of dibenzo-30-crown-10 (DB30C10) and its derivatives as ISE sensors for diquat (DQT) and paraquat (PQT), the best PVC ISEs are based on DB30C10 with DQT.2TPB and 2-nitrophenyl phenyl ether. This work integrates the molecular recognition researches of Stoddart at Sheffield and structural studies of Williams at Imperial College, London, with the sensor studies of UWCC.36 On the fundamental side, the researches have sh0wn36.3~ that DB30C10 changes its shape dramatically to accommodate the guest molecules, while bis-paraphenylene-34-crown- 10 (BP34C10) is hardly modified in shape when it engulfs PQT.However, this feature is not carried through to improving the POP+ electrodes which are best based on PQT.2TPB without crown ether, and activated by a charge transfer interaction38 between PQP+ and TPB-. There is no advantage to be gained in replacing TPB- by an alternative ani0n.3~ The importance of guanidine in the biological and medical fields has led to guanidinium ISEs. The most suitable were previously39,40 found to be based on crown ethers with di be nzo-27-crown-9 with di bu t yl ph t hala te plasticising solvent mediator, but recent studies41 have shown this to be super- seded by bis-metaphenylene-26-crown-8 (BMP26C8). The facility of such sensors is provided by the ability of the guanidinium cation, [(HzN)&]+ to complex with crown ethers of between 18 and 33 ring cations with 27 members being the most selective.42-44 X-ray and NMR spectrometry data support the [NH - - .01 arrangement of hydrogen bonds to yield stable complexes of 1 mol host to 1 mol guest in most instances, and 2 to 3 in some instances, depending on the solvent and the guanidinium salt. There has been prospecting in the area of bis-crownANALYTICAL PROCEEDINGS. OCTOBER 1989. VOL 26 ethers.32.4-7 For example, Shono and C O - W O T ~ ~ T S ~ ~ , ~ ~ have shown good potassium ion selectivity for some bis- (15-crown-5) derivatives, and one containing a dodecyl link exhibits good lipophilicity and longer lifetime for the resulting ISEs.46 Further studies47 on bis[ (benzo- 12-crown-4)-2- ylmethyl]-2-dodecyl-2-methylmalonate and bis(benzo-15- crown-5)-15-ylmethyl pimelate show the former to be selective to sodium and the latter to potassium, both being best used with 2-nitrophenyl phenyl ether plasticising solvent mediator and potassium tetra-4-chlorophenylborate anion excluder in PVC matrices.Although these electrodes offer promising alternatives to glass electrodes for sodium and to valinomycin electrodes for potassium, the data for measurements of ion in blood serum indicate a need for further research in order to improve the correlations with flame photometric measure- men ts.47 With regard to smaller crowns, a very useful dodecylmethyl- 14-crown-4 shows good selectivity towards lithium and sodium.48 However, lithium ISEs have to function over a small concentration range around 1 mM [Li] in complicated blood and blood serum matrices.Only relatively few sensors approach the required selectivity specifications,j” and many new materials continue to fail the stringent tests.”) Potentiometric Ion Sensing with Diphenyl Ethers of Tetraethylene Glycol Two derivatives of these have been studied,41-s‘)-51 namely 1,l l-bis(2-hydroxy-5-formylphenoxy)-3,6,9-t~oxaundecane and its benzyl ether. These can be looked upon as “graspers” of cations in ion sensing. The performance of ISE types with the diphenol compound depends on the solvent mediator.5 However, the system with dioctylphenyl phosphonate as plasticising solvent mediator did not fulfil its promise51 as a lithium 1SE.W Indeed, despite all of the effort to achieve optimisation, the sensors based on the diphenyl compound and the benzyl ether are so far inferior to existing ones.51 Continuing studies on the materials have yielded greater success with guanidinium ion sensing.41 Thus, both are better than the previously recommended dibenzo-27-crown-9, but marginally less good than the now recommended41 BMP26C8 (see above).Conclusion Polyethers offer many prospects as selective potentiometric sensors. The considerable researches in the field have led to some useful systems, but it is clear that the mechanisms involved and modifying influence of the plasticising solvent mediator needs to be much better understood in order to exploit fully this interesting range of materials. The co-workers of the various references are thanked for their dedication and co-operation.Also, thc various sponsors who have provided financial support are thanked for their generos- ity and interest. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. References Moore, C.. and Pressman. B. C.. Biochem. Biophys. Res. Commun., 1954. 15. 62. Stefanac. Z.. and Simon. W.. Chimia. 1966. 20, 436. Stefanac. Z., and Simon. W.. Microchem. J.. 1967, 12. 125. Pedersen. C. J.. J. Am. Chem. SOC., 1967. 89, 7017. Neu. R.. Arzneim. Forsch.. 1959, 9, 585. Uno, T.. and Miyajima, K., Chem. Pharm. Buff. (Tokyo). 1963. 11, 75. Levins. R. J., and Ikeda, R. M.. Anal. Chem., 1965, 37. 671. Levins, R. J., Anal. Chem.. 1971, 43, 1045. Rechnitz. G. A., and Eyal, E., And. Chem., 1972. 44. 370. PetrAnek. J., and Ryba, 0.. “IUPAC International Sympo- sium on Selective lon-Sensitive Electrodes. UWIST.Cardiff. 9-12 April 1973.” Paper 13. 11. 12. 13. 14. 1s. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 333 Ryba, 0.. and Petranck. J., J. Efecrroanaf. Chem., 1973. 44. 425. Jaber, A. M. Y.. Moody. Ci. J.. andThomas, J . D. R.. Anafysr. 1976. 101, 179. Jaber, A. M. Y.. Moody. G. J.. and Thomas, J. D. R., J. Inorg. Nucf. Chem.. 1977. 39. 1689. Alexander, P. H . V.. Moody, G . J.. and Thomas. J . D. R., Anufysr. 1987. 112, 113. Gadzekpo. V. P. Y.. Moody. G . J., and Thomas, J . D. R.. Anufysr. 1985. 110. 1381. Jaber. A. M. Y.. Moody, (i. J., andThomas. J . D. R.. Anafysr. 1977. 102, 943. Lima. J. L. F. C., and Machado, A. A. S. C . , personal communication. Jones. D. L.. Moody, G . J..Thomas. J . D. R.. and Hangos. M.. Anafysr, 1979, 104. 973. Moody, G . J., and Thomas. J. D. R.. Lab. Prucr., 1979, 28. 125. Moody. G . J.. Thomas, J . D. R.. Lima, J . L. F. C.. and Machado. A . A . S. C., Anafysr. 1988. 113. 1023. Jaber, A. M. Y.. Moody, G. J., andThomas. J . D. R., Anafysr. 1988. 113. 1409. Dclduca. P. G . . Jaber. A. M. Y.. Moody. G. J.. and Thomas. J . D . R.. J. Inorg. Nucf. C‘hem., 1978. 40. 187. Moody. G . J., and Thomas, J . D. R.. in Cross, J.. Editor. “Non-ionic Surfactants: Chemical Analysis,” Marcel Dekker, New York, 1986, p. 117. Jones. D. L.. Moody, G . J., and Thomas. J. D. R.. Anufysr. 1981, 106. 439. Jones, D. L., Moody. G . J.. Thomas, J. D . R., and Birch, B. J., Anafysr. 1981. 106. 439. Hsiao. L., Dunning, H . N.. and Lorenz.P. B., J. Phys. Chem., 1956. 60, 657. Schick, M. J., Atlas. S. M., and Elrich. F. R., J. Phys. Chem.. 1962. 66, 1326. Alexander. P. H. V.. Moody, G . J.. Thomas. J. D. R.. and Birch, B., Analyst. 1987, 112, 849. Doyle, B., Moody. G . J.. and Thomas, J . D. R.. Tufanta. 1982. 29. 257. Jaber. A . M. Y.. Moody, G . J.. Thomas, J. D . R.. and Willcox, A.. Tafanra. 1977. 24. 655. Doyle. B.. Moody. G . J.. and Thomas, J. D. R., Tafanra. 1982. 29, 614. Lindner, E., Toth, K.. Orvath. M.. Pungor, E.. Agai, B., Bitter, I., Toke, L.. and Hell. Z.. Fresenius Z. Anal. Chem.. 1985. 322. 157. Moody, G . J., Owusu. R. K.. andThomas J. D. R.. Analyst, 1987. 112. 121. Moody. G. J.. Owusu. R . K.. and Thomas. J. D. R.. Analysr. 1987. 112, 1347. Moody, G. J., Owusu. R. K.. and Thomas, J.D. R., Analyst, 1988. 113, 65. Stoddart, J. F.. and Thomas, J . D. R., Chem. Sensors Club News, 1987, 5 , 6. Colquhoun, H. M.. Stoddart. J. F.. and Williams, D. J . , New Scienrisr. May lst, 1986, 44. Moody, G. J.. Owusu. R . K., Slawin. A. M. Z.. and Spencer. N.. Angew. Chem. Int. Ed. Engf.. 1987. 26, 890. Bochenska. M., and Biernat, J . F., Anal. Chim. Acra. 1984, 162. 369. Assubaie. F. N.. Moody, G . J.. and Thomas. J . D. R., Analyst. 1988, 113, 61. Assubaie, F. N.. Moody. G. J., andThomas. J . D. R.. Analyst. in the press. Kyba. E. P., Helgeson. R . C.. Madan. K.. Gokel, G. W.. Tarnowski, T. L., Moore, S. S.. and Cram, D. J.. J. Am. Chem. SOC.. 1977, 99,2564. Lehn, J. M., Vierling, P., and Hayward, R. C.. J. Chem. SOC., Chern. Commun., 1979. 296. Stolwijk, T.B., Grootenhuis, P. D. J., van dcr Wal, P. D . , Sudholter, J. W. H. M., and Kruise. L.. J. Org. Chem., 1986, 51, 4891. Kimura. K.. Mazeda. T.. Tamura, H . , and Shono. T., J. Efecrroanaf. Chem.. 1979. 95. 91. Kimura, K.. Tamura, H.. and Shono, T.. J. Chem. Soc., Chem. Commun.. 1983.492. Moody, G. J.. Saad. B. B.. and Thomas. J . D. R.. Anafysr. 1989, 114, 15.334 ANALYTICAL PROCEEDINGS. OCTOBER 1989. VOL 26 48. 49. Kitazawa. S.. Kirnura. K., Yano, H.. and Shono. T.. Analysr. 1985, 110. 295. Gadzekpo. V. P. Y.. Moody, G. J.. Thomas, J. D. R.. and Christian. G. D.. Ion-Sel. Electrode Rev.. 1986, 8. 173. 50. 51. Beswick, C. W., Moody, G. J.. and Thoma;, J . D. R.. Anal. Proc., 1989, 26. 2. Moody, G. J.. Saad, B. B.. Thomas. J . D. R.. Kohnkc. F. H..and Stoddart. J . F.. Analyst. 1988. 113. 1295. Amperometric Biosensors Jane E. Frew and Monika J. Green MediSense (UK), Inc., 14 Blacklands Way, Abingdon, Oxfordshire OX14 IDY ‘The concept of performing clinical analyses outside the centralised hospital laboratory has stimulated great interest in the development of rapid and simple analytical procedures. The emphasis is on the design of tests that may be carried out nearer to the patient or even by the patient. This might be on a hospital ward, in a casualty department, in a doctor’s office or even in the patient’s home. At present the clinical chemistry laboratory is required to monitor a very wide range of analytes, including naturally occurring low relative molecular mass species, drugs and enzymes. Many of the existing assays for such species are spectrophotometric or colorimetric and often involve extensive sample pre-treatment.The application of electrochemical techniques to assay design offers the opportunity for simplify- ing experimental protocols; for example, it is possible to perform measurements on whole blood rather than having to obtain serum or plasma. Dry -strip Technology In order to reduce the number of manipulative steps involved in carrying out a given amperometric assay it is necessary to move away from the use of the standard electrochemical cells and electrode configurations that would be employed in a research laboratory. The exploitation of screen printing technology allows the design of electrode systems in the form of a strip.’ Each strip consists of two electrodes screen printed on to a solid support: a silver - silver chloride reference electrode and a carbon-based working electrode which con- tains the assay reagents.Electrochemical measurements are made by placing a small volume (approximately 25 pl) of test sample on the electrode area. The strips are intended to be used once only and then discarded. Strategies for the Design of Amperometric Biosensors Two basic approaches have been adopted in the development of amperometric biosensors for the detection of clinical analytes. These involve: (i) the direct electrochemical monitor- ing of the product of an enzymatic reaction; (ii) the use of electron transfer mediators in the construction of enzyme electrodes. The simplest approach is undoubtedly the direct electro- chemical monitoring of the product of an enzymatic reaction, the analyte being a substrate for a specific enzyme.The configuration of such assays can be illustrated by tests for paracetamol (acetaminophen) and aspirin (via measurement of salicylate). These drugs are commonly used in overdose situations’ and under these circumstances it is essential to be able to discriminate which drug or combination of drugs have been ingested by the patient. The concentration of free drug in the blood must be known before an appropriate course of treatment can be prescribed. Par ace tamol In the assay for paracetamol (acetaminophen) the enzyme aryl acylamidase (EC 3.5.1.13) converts the drug to p-amino- phenol: NH2 NHCOCH3 OH OH Paracetamol pAminophenol The p-aminophenol is detected electrochemically by oxidation at +250 mV versus Ag/AgCI.This approach is more satisfac- tory than attempting to design an assay based on monitoring the drug directly. The oxidation of paracetamol requires much higher potentials, at which there will be considerable interfer- ence from other species in blood. Conjugated forms of the drug will also be oxidised; clinicians are interested only in the unconjugated form of paracetamol. Under appropriate condi- tions paracetamol can be monitored over the clinical range (0-3 mM) in whole blood. Salicylate Many derivatives of salicylate are available commercially, but the most important is acetylsalicylic acid or aspirin. The drug is hydrolysed rapidly to salicylic acid, which circulates in the blood in the ionised form.3 The assay uses the monooxygenase salicylate hydroxylase (EC 1.14.13.1) to convert salicylate to catechol in the presence of reduced nicotinamide adenine dinucleotide (NADH) and molecular oxygen-‘: Catechol Salicylate The size of the oxidation current generated by electrochemical oxidation of catechol formed in the enzyme reaction provides an estimate of analyte. Measurements are performed at a fixed potential of +300 mV versus Ag/AgCl and results are obtained within 1 min.Although analyses are carried out on blood, calibrations have been constructed in terms of plasma salicylate to aid interpretation of the data; serum or plasma are the accepted media for salicylate tests. The limitation of this system is that it cannot distinguish between different concentrations of plasma salicylate above approximately 4 mM.However, the calibration can be ex- tended by including benzoate as an assay reagent. Benzoate is a pseudosubstrate for salicylate hydroxylases; unlike the reac- tion involving salicylate there is no hydroxylation of the aromatic compound and all oxygen utilisation is diverted towards the production of hydrogen peroxide. With this configuration analyses can be performed up to a concentration of 7.2 mM plasma salicylate, thereby covering the range for both acute and chronic cases of salicylate intoxication.ANALYTICAL PROCEEDINGS. OCTOBER 1989. VOL 26 335 a- Amy lase Assays based upon the direct electrochemical monitoring of the product of an enzymatic reaction are also appropriate for detection of an analyte that is itself an enzyme.In this case the analytical system relies upon the conversion of a labelled substrate into an electroactive product. This approach has been utilised in the design of a new amperometric assay for the hydrolytic enzyme a-amylase.6 The determination of a-amylase (1,4-a-~-glucan gluco- hydrolase, EC 3.2.1.1) is important in the diagnosis of acute and chronic pancreatitis and obstruction of the pancreatic duct. The key to the success of the assay is the use of a soluble, well defined oligosaccharide substrate, 4-aminophenyl-c~-malto- pentaoside. In the presence of a-amylase this compound is hydrolysed to smaller sub-units which are themselves sub- strates for another enzyme, agfucosidase (EC 3.2.1.20). The final product of the reaction sequence is p-aminophenol, which can be monitored electrochemically by oxidation at + 155 mV versus Ag/AgCI.Under appropriate conditions the extent of formation of p-aminophenol is related to the concentration of a-amylase present in the test sample. The assay yields a linear calibration for a-amylase in blood over the clinical range of 50-1000 U I-’. Electron Transfer Mediators in Enzyme Electrodes The direct transfer of electrons between redox enzymes and electrodes is difficult to achieve and is dependent upon the nature of the electrode surface and the solution conditions.’ As a consequence, redox active mediators are often used to enhance the rate of electron transfer between enzyme and electrode. Ferrocenes as Mediators in Enzyme Electrodes The use of ferrocene derivatives as mediators of electron transfer has proved particularly successful in the design of enzyme electrodes.An assay for hydrogen peroxide exploits the ability of a ferrocene derivative to function as an electron donor.8 The fundamental reaction is the enzymatic reduction of hydrogen peroxide by horseradish peroxidase; the native enzyme is regenerated by subsequent electron transfer from an electrode to the peroxidase via the ferrocene mediator. This system provides not only a direct assay for hydrogen peroxide, but can also be a component of other analytical systems in which hydrogen peroxide is a product of one or more further chemical reactions. An electrochemical method for detection of cholesterol has been devised which incorporates the peroxidase coupled assay.9 Cholesterol esters in serum are associated with lipoprotein complexes and the first step of the assay is the liberation of the ester by the action of surfactant.Conversion to free cholesterol can then be achieved in the presence of cholesterol oxidase. Hydrogen peroxide is a product of this reaction and it can be determined by the peroxidase based route. The assay is effective for the determination of cholesterol in the millimolar concentration range. There are two other possible approaches to determining cholesterol which also utilise ferrocene derivatives as media- tors. The cholesterol can be oxidised by a specific flavoprotein oxidase coupled to ferricinium ion or oxidised by a dehydro- genase and the NADH generated can then be detected by a coupled reaction based on diaphorase.4-Methyl-o-quinone as a Mediator to NADH Another example of the use of mediators in enzyme electrodes is in an assay for 3-hydroxybutyrate.10 This compound is the major component of the ketone bodies, the others being acetoacetate and acetone. They are produced by incomplete fatty-acid metabolism in the liver in conditions involving the impaired utilisation or inadequate supply of carbohydrates.2 The determination of ketone bodies in plasma or serum is an extremely reliable guide to monitoring the efficacy of insulin therapy in the treatment of diabetic ketoacidosis. I n the assay, 3-hydroxybutyrate is converted to acetoacetate by the enzyme 3-hydroxybutyrate dehydrogenase (EC 1.1.1.30); NAD+ is required as a co-factor.NADH is a product of the enzymatic reaction. Although NADH formation can be detected by direct oxidation at an electrode it is more convenient to use a redox mediator. The NADH is oxidised by the mediator, 4-methyl-o-quinone, which is itself converted to 4-methyl catechol. The latter is then re-oxidised to the quinone at the electrode. Measurements can be performed at +350 mV versus Ag/AgCl over the clinical range of serum 3-hydroxy- butyrate (approximately 0-20 mM). Conclusion The analytical configurations described in this paper illustrate the fundamental principles underlying the design of ampero- metric biosensors for use in rapid diagnostics. The application of screen printing technology to the development of dry-strip electrode systems transforms the assays from labour-intensive, laboratory-based tests to a format that enables analyses to be performed both easily and rapidly. The commercial viability of strip-based assays with electrochemical detection has been demonstrated by the successful development of a disposable biosensor for glucose, the Exactech system. l 1 Dry-strip elec- trodes incorporating essential reagents are used in conjunction with a pen-sized meter to measure whole blood glucose in only 30 s. The advantages of biosensors compared with other ana- lytical devices are often quoted; they are small, portable, easy to use, accurate, precise and yield results rapidly. Now these advantages have been exploited in the design of diagnostic systems that can find application in the “real world,” not simply in the research laboratory. The authors wish to thank their colleagues at MediSense for their contributions to the work described in this paper. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. References Higgins. I. J., McCann, J . M., Davis, G.. Hill. H. A. O., Zwansiger, R., Treidl, B. L., Birket. N. N., and Plotkin, E. V.. Eur. Pat. Applic., Publication Number 127 958. Pesce. A. J., and Kaplan, L. A.. Editors, “Methods in Clinical Chemistry,” C. V. Mosby, St. Louis, MO, USA, 1987. Kwong, T. C., CRC Crit. Rev. Clin. Lab. Sci., 1987, 25. 137. Frew, J. E., Bayliff, S. W., Gibbs, P. N. B., and Green, M. J., Anal. Chim. Acta, in the press. White-Stevens, R. H., and Kamin. H.. J. Biol. Chem., 1972, 247, 2358. Batchelor. M. J.. Williams, S. C., and Green, M. J.. J. Electroanal. Chem., 1988, 246, 307. Frew, J. E., and Hill, H. A. O., Eur. J. Biochem., 1988, 172, 261. Frew. J. E.. Harmer, M. A., Hill, H. A. 0.. and Libor, S. I.. J. Electroanal. Chem., 1986, 201, 1. Ball. M. R., Frew, J . E., Green, M. J.. and Hill, H. A. O., Proc. Electrochem. Soc., 1986. 86-14, 16. Batchelor, M. J.. Green, M. J.. and Sketch, C. L., Anal. Chim. Acta. 1989, 221, 289. Matthews, D. R., Holman, R. R., Bown, E., Steemson, J., Watson, A., Hughes, S., and Scott, D.. Lancet, 1987, i, 778.