MINIREVIEW Piezoelectric sensors based on molecular imprinted polymers for detection of low molecular mass analytes Yildiz Uludag ˘ 1,2 , Sergey A.Piletsky 1 , Anthony P. F. Turner 1 and Matthew A. Cooper 2 1 Cranfield Health, Cranfield University, Silsoe, UK 2 Akubio Ltd, Cambridge, UK Introduction Biosensors are analytical devices that comprise a sam- ple-delivery mechanism with a biological recognition element and a suitable transducer, usually coupled to an appropriate data-processing system (Fig. 1). The biological recognition element is typically an enzyme, microorganism, cell, tissue or other bioligand [1] and the transducer is required to convert the physico-chem- ical change resulting from the interaction of molecules with the receptor into an electrical signal. Over the past decade the benefits of label-free analysis have begun to gain a foothold as a mainstream research tool in many laboratories [2,3]. These techniques do not require the use of detection labels (fluorescent, radio or colorimetric) to facilitate measurement; hence detailed information on an interaction can be obtained during analysis while minimizing sample processing requirements and assay run times [4]. Unlike label- and reporter-based technologies that simply confirm the presence of the detector molecule, label-free tech- niques can provide direct information on analyte bind- ing to target molecules typically in the form of mass addition or depletion from the surface of the sensor substrate [5]. Of the various label-free detection modalities, piezo- electric sensing has become popular with researchers because of the low barriers to entry, inherent simp- licity, ease-of-use, low cost, and speed to result. However, there are relatively few examples of small- molecule detection using traditional immunoassay formats: niacinamide detection has been achieved in serum and urine [6], the endogenous cofactors NAD + and NADP + have been titrated against the enzyme Keywords drug; hapten; label-free; molecularly imprinted polymer; QCRS; quantification; quartz crystal microbalance; screening; small molecule Correspondence M. A. Cooper, Akubio Ltd, 181 Cambridge Science Park, Cambridge CB4 0GJ, UK Fax: +44 1223 225 336 Tel: +44 1223 225 326 E-mail: mcooper@akubio.com (Received 6 July 2007, accepted 24 August 2007) doi:10.1111/j.1742-4658.2007.06079.x Biomimetic recognition elements employed for the detection of analytes are commonly based on proteinaceous affibodies, immunoglobulins, single- chain and single-domain antibody fragments or aptamers. The alternative supra-molecular approach using a molecularly imprinted polymer now has proven utility in numerous applications ranging from liquid chromatogra- phy to bioassays. Despite inherent advantages compared with biochemi- cal ⁄ biological recognition (which include robustness, storage endurance and lower costs) there are few contributions that describe quantitative ana- lytical applications of molecularly imprinted polymers for relevant small molecular mass compounds in real-world samples. There is, however, sig- nificant literature describing the use of low-power, portable piezoelectric transducers to detect analytes in environmental monitoring and other appli- cation areas. Here we review the combination of molecularly imprinted polymers as recognition elements with piezoelectric biosensors for quantita- tive detection of small molecules. Analytes are classified by type and sample matrix presentation and various molecularly imprinted polymer synthetic fabrication strategies are also reviewed. Abbreviations MIP, molecularly imprinted polymer; QCM, quartz crystal microbalance. FEBS Journal 274 (2007) 5471–5480 ª 2007 Akubio Ltd. Journal compilation ª 2007 FEBS 5471 glucose dehydrogenase and rank order binding to the enzyme has been determined [7], and biotin has been detected with a high-frequency microfluidic acoustic biosensor using an anti-biotin serum [8]. Real-time detection of 4-aminobutyrate (one of the main inhibi- tory neurotransmitters) was achieved with an anti- (4-aminobutyrate) serum with a minimum detection limit of 38 lm [9]. Kurosowa et al. [10] reported a portable dioxin sensor able to detect 2,3,7,8-tetrachlorodibenzo- p-dioxin. Dioxin is well-known as a highly toxic com- pound that poses a threat to the environment. The quartz crystal microbalance (QCM) sensor surface was prepared by immobilizing anti-(2,3,7,8-tetrachlo- rodibenzo-p-dioxin) serum and a linear calibration curve was created using 100–0.1 ngÆmL )1 2,3,7,8-tetra- chlorodibenzo-p-dioxin before detection of the com- pound in fly ash samples. Hence mass-sensitive acoustic immunoassays can provide a label-free method for detecting and analysing molecules. However, biological materials are expensive, sensitive to harsh conditions and their shelf life on the sensor surface can be limited. The use of molecularly imprinted polymers (MIPs) as synthetic receptors pro- vides an attractive alternative to biological receptors. MIP recognition elements provide sensor surfaces that have a long shelf life, are robust and simple to prepare, and provide a 3D interfacial matrix layer with high binding capacity for analytes. Such binding capacity is crucial when the molecular mass of the analyte is < 500 Da. This review summarizes the key approaches to incorporating imprinted polymers as recognition ele- ments in piezoelectric sensors for detection of small molecules. Acoustic biophysics Acoustic biosensors allow the label-free detection of molecules and the analysis of binding events. In gen- eral, they are based on quartz crystal resonators, which are found in electronic devices such as watches, com- puters and televisions, with over one billion units mass-produced each year. Quartz crystal is a piezoelec- tric material which mechanically oscillates if an alter- nating voltage is applied. A QCM consists of a thin quartz disc sandwiched between a pair of electrodes. The mode of oscillation depends on the cut and geom- etry of the quartz crystal. If mass is applied to the sur- face of the quartz resonator, the frequency of the oscillation decreases. By measuring the change of fre- quency, it is possible to determine the change in mass. Measurement of mass using quartz crystal resonators was first examined by Sauerbrey [11], who showed that the frequency change of the crystal resonator is a linear function of the mass per area m s , or absolute mass Dm: Df m ¼À f 2 0 F q q q m s ¼À f 2 0 F q q q Dm s A el ð1Þ where f 0 is the resonance frequency of the unperturbed quartz resonator, F q the frequency constant of the crystal (F q ¼ f 0 .d q ), d q the thickness, q q the mass den- sity, and A el the electrode size of the crystal resonator. Equation (1) is valid only for thin, solid layers depos- ited on the resonator. Initially, the QCM system was used for dry measure- ments, later when suitable oscillator circuits were developed, it was possible to carry out measurements under liquid conditions [12]. This led to the use of QCM systems as biosensors to detect molecular inter- actions (Fig. 2). A new equation was derived by Kanazawa and Gordon to explain the relationship between density (q l ) and viscosity (g l ) of the liquid and the frequency of the quartz crystal resonator: Df ¼Àf 3=2 q ffiffiffiffiffiffiffiffiffiffiffiffi q 1 g 1 pq q l q r ð2Þ where q q and l q are the quartz density and shear mod- ulus, respectively [13]. In a two-layer system these frequency shifts simply add up to an overall shift: Df ¼ Df m þ Df l ¼Àf 2 0 Dm s F q q q A el þ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi g 1 q 1 f 0 pl q q q r ! ð3Þ In addition to the frequency shift, there also exists a dampening of the resonator caused by the viscous Fig. 2. Schematic representation of quartz crystal resonance sensing. Antibody/Protein Enzyme Microorganism Cell Recognition elements Transducer Optical Piezoelectric Electrochemical Calorimetric Electric signal Analyte Antibody/Protein Enzyme Microorganism Cell Recognition elements Transducer Optical Piezoelectric Electrochemical Calorimetric Electric signal Analyte Fig. 1. Schematic representation of a biosensor. Detection of low molecular mass analytes Y. Uludag ˘ et al. 5472 FEBS Journal 274 (2007) 5471–5480 ª 2007 Akubio Ltd. Journal compilation ª 2007 FEBS liquid layer. There are numerous examples in the liter- ature on the applications of Eqns (2) and (3) and more sophisticated algorithms that incorporate the measure- ment of complex shear modulus in addition to mass, viscosity and density, but a detailed analysis of this approach lies beyond the scope of this review. How- ever, it is possible to further optimize sensor sensitivity by appropriate matching of the interfacial polymer chemistry physical properties with the acoustic sensor design, and a few seminal examples in this regard are noted below. It can been seen from the above that a quartz crystal resonator is thus sensitive not only to mass, but also to viscosity and density changes on the resonator sur- face. Therefore the term quartz crystal microbalance is not strictly accurate for all applications. The device is also known as thickness-shear mode resonator or a bulk acoustic wave sensor, because the bulk of the crystal oscillates at a resonance frequency in a thick- ness shear mode of vibration. Surface acoustic wave sensors are also based on the piezoelectric properties of quartz crystal. In this case only a surface wave is generated by the electrodes and the frequency of the surface waves is $ 100 MHz to 1GHz [14]. These frequencies are much higher than thickness-shear mode resonators, and this is the reason for the higher sensitivity of surface acoustic wave sen- sors. However, higher sensitivity also means higher response to viscosity changes and this problem causes difficulties when surface acoustic wave sensors are used in liquids [15]. By monitoring the change in resonant frequency and motional resistance that occurs upon adsorption of a ligand to the surface, quartz crystal resonators can be used to characterize interactions with peptides [16], proteins and immunoassay markers [17], oligonucleotides [18], viruses [19], bacteria [20] and cells [21]. The technology can thus be applied to an extremely wide range of biological and chemical entities with a molecular mass range from < 200 Da through to an entire cell. Application of acoustic sensors to small molecule detection The detection limit of many affinity biosensors is clo- sely linked to the molecular mass of the analyte. Many researchers prefer to immobilize the small molecule analyte on the sensor surface and measure the binding of a larger molecule [22] or to perform a competitive displacement assay with a hapten–carrier conjugate [23]. Another option is to conjugate the small molecule to a bead or other additional mass load to increase the molecular mass of the complex detected [24]. Other approaches utilize a coupled assay format in which direct binding then results in capture of an enzyme that can convert a soluble substrate to a precipitate to effect signal amplification. For example, organophos- phorous and carbamate pesticides have been detected using a two-enzyme system to produce peroxide, which combined with peroxidase and benzidine formed an insoluble product that absorbed to the sensor surface [25]. Nonionic surfactants were reported to enhance the surface deposition of suspended precipitate enabling detection of carbaryl and dichlorvos pesti- cides down to 1 p.p.m. This group also published an assay for 4-aminophenol which involved precipitation of indophenol in an amount proportional to the 4-aminophenol concentration in the sample [25]. In addition to the above strategies, the intrinsic sen- sitivity of QCM to shear modulus, viscosity and den- sity changes manifested at the surface interfacial layer allows for the development of novel small-molecule detection assays. In this case, the binding of a small molecule that induces conformational changes in the interfacial layer results in a modulated shear modulus (related to rigidity and ⁄ or flexibility). Such effects can significantly amplify the response due to mass binding alone. For example, Carmon et al. [26] immobilized a glucose ⁄ galactose receptor on a QCM sensor surface and exposed the receptor to 180 Da sugars. A repro- ducible frequency change was observed which was ascribed to the conformational change of the receptor upon ligand binding. Similarly conformational changes have been invoked in the binding of ions and peptides to calmodulin due to ion or peptide binding [27], and the insertion of an Ad-2a model peptide onto glyco- lipid monolayers [28]. In the latter case, the 2a-helix structure of the peptide in the bulk solution is known to convert to a b-structure upon association with a lipid monolayer. This conformational change was man- ifested as a frequency decrease for the piezoelectric sensor. This approach has been extended further in a dynamic electropolymerization study in which imped- ance data were acquired during polymerization at the fundamental and third harmonic modes of a 10 MHz thickness shear mode resonator [29]. At a critical thickness, the system exhibited mechanical resonance, a special condition in which the mechanical shear deformation across the polymer film corresponded to one quarter of the acoustic wavelength. At this point, the resonant frequency and admittance data showed dramatic changes with polymer coverage. Several groups have also extended full impedance analysis incorporating shear modulus modelling to protein films [30] and phage binding [31]. Y. Uludag ˘ et al. Detection of low molecular mass analytes FEBS Journal 274 (2007) 5471–5480 ª 2007 Akubio Ltd. Journal compilation ª 2007 FEBS 5473 For all the examples cited above, the binding capacity of the sensor surface is critical to maximize the sensitivity of the sensor for the detection of small molecules. Here the synergies between robust, stable MIPs with a mass and shear-modulus sensitive sensor become apparent. The meso ⁄ microporous 3D matrix structures formed by MIPs, not only increase the total number of binding sites acoustically coupled to the sensor, but also result in additional frequency changes manifested in analyte- and dose-dependent modulation of the surface-associated shear modulus of the poly- mer layer in the polymeric structure. In other words, binding of analyte to a MIP can result a larger change in the total acoustic load on the sensor and hence enable more robust detection of small mole- cules. It is possible to model this phenomena as a finite viscoelastic layer (of thickness h f , density r f and shear modulus G ¼ G ¢ + jG¢¢ [32]. The latter value, G, is the complex shear modulus where G¢ and G¢¢ are the layer storage and layer loss moduli, respectively. These layers are then exposed to a bulk liquid (of vis- cosity g L and density q L ). For these layered compo- nents, it is possible to derive the surface mechanical impedances: Z M ¼ jxq s ð4Þ for an ideal ⁄ rigid mass layer, where r s is the mass per area contributed by the interfacial layer; Z F ¼ ffiffiffiffiffiffiffiffi q f G q tanhðch f Þð5Þ for a viscoelastic film (MIP), where c is the shear wave propagation constant (c ¼ j2pf o (r f ⁄ G) 1 ⁄ 2 ) and j ¼ Ö-1; Z L ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2pf 0 q L g L 2 r ð1 þ jÞð6Þ for a liquid layer (semi-infinite Newtonian liquid), and finally; Z ¼ j2pf 0 ffiffiffiffiffiffiffiffi q f G q Z L cosh hðch f Þþ ffiffiffiffiffiffiffiffi q f G p sinhðch f Þ ffiffiffiffiffiffiffiffi q f G p cosh hðch f ÞþZ L sinhðch f Þ ð7Þ where Z is the impedance for the composite system. Note that the impedance measured by the piezoelectric sensor is not simply the sum of those for individual layers, as for each layer there will be an acoustic phase shift, which causes a transformation of the impedance contributed by layers more distant from the resonator. In addition, this model does not accommodate the typ- ically inhomogeneous layers that exist in reality in MIPs exposed to biological systems, or changes in dis- tribution of mass induced by varying matrix conditions and analyte binding. When layers of a material of differing density and viscosity to the liquid (such as a MIP) are deposited at the surface–liquid interface, there is a contribution from both the bulk liquid and the added material that has displaced liquid and mechanically coupled to the interface. In this case, the penetration depth can be defined as: d ¼ð1 À v p Þ ffiffiffiffiffiffiffiffiffiffiffi g L pf 0 qL r þ v p ffiffiffiffiffiffiffiffiffiffiffi g L pf 0 q f s ð8Þ where g p and q p are the liquid viscosity and density, respectively, and v p is the fraction of the volume within the penetration depth occupied by protein. This could be extended to encompass both a receptor layer and analyte layer if necessary. Integrating the mole fraction of MIP ⁄ water in combination with the defini- tion of a composite impedance above, we can derive: Z ¼ v p j2pf 0 ffiffiffiffiffiffiffiffi q f G q ð1 À v p ÞZ L cosh hðch f Þþ ffiffiffiffiffiffiffiffi q f G p sinhðch f Þ ffiffiffiffiffiffiffiffi q f G p cosh hðch f Þþð1 À v p ÞZ L sinhðch f Þ ð9Þ Molecularly imprinted polymers The history of MIPs can be traced to experiments per- formed by Polyakov and his group in 1931. Silica gels prepared by Polyakov’s group showed selective binding towards one of the solvents used for the gel prepara- tion. Later, studies by Wulf [33] and Haupt and Mos- bach [34] helped to establish this technique in relation to organic polymers. Initially, MIPs were used as sta- tionary phases for chromatographic methods. Later, the application of imprinted polymers was extended to the biosensors area, where MIPs have been used as recognition elements as an alternative to biological materials such as antibodies and proteins. Similarly, synthetic receptors formed by molecular imprinting can be used to recognize biological or nonbiological molecules on QCM sensors. Here the imprint of a tem- plate molecule is formed on a synthetic polymer that has cavities resembling the geometric shape of the tem- plate and also has binding sites for template recogni- tion [33]. MIPs as synthetic receptors have several advantages over biological receptors [35]. The main advantage of MIPs is their stability to harsh condi- tions, in contrast to natural biomolecules that are sen- sitive to environmental changes and can denature easily. Because of the robust nature of MIPs, biosen- sors that use MIP surfaces in general have a longer shelf life than analogous biological sensors. MIPs are simple to prepare, and their adaptation to a variety of Detection of low molecular mass analytes Y. Uludag ˘ et al. 5474 FEBS Journal 274 (2007) 5471–5480 ª 2007 Akubio Ltd. Journal compilation ª 2007 FEBS practical applications has been widely demonstrated. In addition, molecular interaction studies with MIPs can be performed in organic solvents as well as aque- ous solvents. Template molecules can be imprinted to a polymer with covalent- [36], noncovalent- [37] or metal-ion- mediated [38] interactions, followed by appropriate cross-linking agents. Imprinting consists of three steps; first, one or several functional monomers are mixed with the template molecule in a solvent. Second, poly- merization of monomers occurs in the presence of a cross-linker. Third, the template is removed from the polymer using basic, acidic or detergent solutions (Fig. 3). The performance and selectivity of MIPs towards a target molecule depend on many factors. These include the molecular diversity of the monomer units employed, the geometry of the imprinted cavity, the rigidity of the cavity and associated implications for enthalpy ⁄ entropy compensation. These factors all govern the affinity and selectivity of the MIP recogni- tion elements towards the analyte of interest. A discus- sion of the importance of hydrogen bonds, van der Waal’s forces, ionic interactions, ion dipole interac- tions and hydrophobic interactions and other molecu- lar phenomena between the template and the monomer lies beyond the scope of this minireview and the reader is referred to recent reviews and books in this area [35,39]. To prepare MIPs with good affinity and selectivity, polymerization conditions and the selection of mono- mer and cross-linking agent are important parameters for optimization (Table 1). There is no general proce- dure for MIP preparation; therefore, depending on the application, procedures need to be examined thor- oughly before a decision is made. After the selection of the polymerization procedure, the variables of the pro- cess should be optimized. Template design, monomer selection, solvent selection and polymerization condi- tions all require attention. In general, the performance of MIPs in aqueous solutions is poor, therefore, if water-soluble templates need to be used the polymeri- zation method needs to be carefully considered. High nonspecific binding and heterogeneity of binding sites needs to be addressed for successful application. If, after polymerization, there are still embedded template molecules remaining in the polymer, this will reduce the capacity and invalidate analysis. Therefore extra care needs to be taken to remove the template from the polymer and the 3D structure of the polymer should allow for easy regeneration of the template from the polymer for repeated bindings. Reproducible fabrication of MIPs is essential for gathering reliable results from each assay. MIP–QCM sensors Every year many new studies are published involving MIP–QCM sensors. In these applications, MIP synthe- sis is performed either in situ on the sensor surface or via preprepared MIP particles ⁄ beads that are immobi- lized on the sensor surface using a PVC matrix. The thickness of the imprinted polymers varies between 18 nm and 5 lm. To obtain a reproducible and reli- able MIP–QCM sensor, it is essential to control the thickness and properties of the polymer coating on the sensor surface. Monomers and target Target & monomers complex Polymerisation -T arget + Target Imprinted polyme r Monomers and target Target & monomers complex Polymerisation -T arget + Target Imprinted polyme r Fig. 3. Schematic representation of molecular imprinting. Table 1. Variables that need to be optimized for the preparation imprinted polymers. Imprinting mechanism Polymerization format Monomer selection Medium selection Polymerization conditions Covalent Bulk polymerization Combinatorial screening Organic solvent Cross-linking agent selection Noncovalent MIP beads Thermodynamic calculations Aqueous solvent Ratio of template ⁄ monomer ⁄ cross-linking agent Metal-mediated Films on bead ⁄ particles or sensor surface Computational methods Temperature Pressure Time Y. Uludag ˘ et al. Detection of low molecular mass analytes FEBS Journal 274 (2007) 5471–5480 ª 2007 Akubio Ltd. Journal compilation ª 2007 FEBS 5475 In situ polymerization Two approaches to synthesize in situ imprinted poly- mers for QCM sensing have been reported. In one approach, the gold surface of the QCM was treated with a thiolated molecule to create active groups on the sensor surface and improve the adhesion of the imprinted polymer on the gold electrode (Table 2). Allyl mercaptan [40,41] (N-Acr-l-Cys-NHBn) 2 [42], mercaptoethanol [43], thioctic acid-modified glycidyl methacrylate [44], and mercaptoundecanoic acid [45] have been used to activate the gold electrode. In the other approach, polymer synthesis was performed directly on to the gold surface without any activation. MIPs could be deposited by surface grafting, spin coating, sandwich casting, or electro-polymerization methods. Surface grafting method Although it is difficult to control MIP film thickness during polymer synthesis, in a sensing device it is essential to reduce batch-to-batch variation. For this purpose, Piacham et al. investigated a possible route to prepare ultra-thin MIP films (< 50 nm) specific for (S)-propranolol [45]. The imprinting process was per- formed directly on to the quartz surface after coating the gold-coated crystal surface with mercaptoundeca- noic acid. The carboxyl groups of mercaptoundecanoic acid were then activated with initiators, 2-ethyl-5-phen- ylisoxazolium-3-sulfonate and 2,2-azobis(2-amidino- propane) hydrochloride. The sensor was dipped into a solution containing template, monomer (methacrylic acid) and cross-linker (trimethylolpropane trimethacry- late). UV irradiation resulted in a surface-grafted poly- mer film on the quartz resonator. Although the detection limit of this MIP–QCM sensor was too high for practical application, Piacham et al. succeeded in producing sensors that generated $ 30 Hz response on the injection of 19 mm (S)-propranolol. Sandwich casting method Alternatively, a sandwich casting method can be used either after surface activation of the QCM sensor sur- face [40,46,47], or directly on to the sensor surface [17,40,48,49]. In this method, a polymerization mixture is dropped on to the quartz crystal and a microcover glass is pressed on to the sensor while UV irradiation is applied. The aim is to distribute the polymer Table 2. Some examples of MIP-QCM studies. 2,2¢-azobis, (2 amidinopropane) hydrochloride; ABAH, 2-ethyl-5-phenylisoxazolium-3-sulfonate; AIBN, azobis-(isobutyronitrile); AMVN, 2,2¢-azobis (dimethylvaleronitrile); EGDMA, ethylene glycol dimethacrylate; GMA, glycidyl methacry- late; HEMA, 2-hydroxyethyl methacrylate; MAH, methacrylamidohistidine; MUDA, mercaptoundecanoic acid; NBAA, N-benzylacrylamide; TRIM, trimethylolpropane trimethacrylate; 4-Vpy, 4-vinylpyridine. Template Surface activation Polymerization method Initiator Cross-linker Monomer(s) Ref Glucose Methacryloyl Surface grafted AIBN EGDMA MAH-Cu(II) ⁄ – D-glucose [56] Nanopeptide (N-Acr- L-Cys-NHBn) 2 Surface grafted – – NBAA, acrylic acid, acrylamide [42] (S)-Propranolol Mercaptoundecanoic acid Surface grafted ABAH TRIM Methacrylic acid [45] Bilirubin Allyl mercaptan Surface grafted Benzophenone Divinylbenzene 4-Vpy [41] Indole-3-acetic acid Allyl mercaptan and 1-butanethiol Sandwich casting AIBN EGDMA N,N-dimethylaminoethyl methacrylate [40] Sialic acid Allyl mercaptan Sandwich casting – EGDMA 4-Vpy and AMVN [40] Dansylphenylalanine Thioctic acid-modified GMA and thioctic acid dodecane ester Sandwich casting AIBN EGDMA Methacrylic acid, 4-Vpy [46] L-Tryptophan Thioctic acid-modified GMA Sandwich casting – TRIM Acrylamide [44] Sialic acid – Sandwich casting – EGDMA N,N,N-trimethylaminoethyl methacrylate, HEMA and AMVN [40] L-serine – Sandwich casting AIBN EGDMA Methacrylic acid [17] L-menthol – Sandwich casting AIBN EGDMA Methacrylic acid [48] Sorbitol – Electro-polymerization AIBN – m-Aminophenol [51] Tegafur – Electro-polymerization – – m-Aminophenol [52] Benzene, toluene and xylene – Spin coating AIBN Divinylbenzene Styrene [62] Detection of low molecular mass analytes Y. Uludag ˘ et al. 5476 FEBS Journal 274 (2007) 5471–5480 ª 2007 Akubio Ltd. Journal compilation ª 2007 FEBS solution evenly on the quartz and thus obtain a uni- form polymer layer. Kugimiya and Takeuchi synthesized a MIP on a quartz crystal sensor surface using a plant hormone, indole-3-acetic acid, as a template [40]. Platinum coated 9 MHz quartz crystals were treated with allyl mercaptan and 1-butanethiol to introduce vinyl groups to the sensor surface. The monomers and the template were dissolved in chloroform and dropped onto the QCM sensor and held with a glass microcover. Poly- merization was initiated by UV irradiation. Assays for indole-3-acetic acid binding to the MIP-coated surface were performed using three crystals and a linear rela- tionship was obtained from 10 to 200 nmol indole- 3-acetic acid. The coefficient of variation between the three sensors was 9.0%. Spin coating method The spin coating approach was used by Ling et al. [43] to prepare MIPs for catecholamines. Ling et al. first activated the gold-coated crystal with mercaptoethanol, and then the homogeneous phase MIP solution was spin coated. Using this method, dopamine-, epinephrine- and norepinephrine-imprinted resonators were prepared and binding assays were performed. The results show that dopamine-specific MIP-coated resonators have bet- ter selectivity than the other MIPs prepared (relative binding selectivity of dopamine–MIP is 1 for dopamine and, 0.03 for norepinephrine and 0.02 for epinephrine). Electro-polymerization method An electro-polymerization method was applied to pre- pare imprinted polymers on quartz crystal sensor sur- faces to detect sorbitol, poly(o-phenylenediamine), tegafur and nucleotides [50–52]. Feng et al. synthesized an o-phenylenediamine film for sorbitol on a QCM crystal by cyclic voltammetry [51]. After MIP deposi- tion the binding assays with sorbitol, glucose, fructose, mannitol and glycerol were performed using a QCM device. Glycerol could bind to the sorbitol-imprinted surface, however, the binding of other compounds was very limited. The detection limit of sorbitol binding was found to be 1 mm. Polymerization prior to sensor coating MIPs have be prepared using a bulk polymerization method; after grinding and sieving, the resulting parti- cles are mixed with PVC and coated on the sensor surface with spin coating. Imprinted polymers for microcystin-LR, nandrolone, phenacetin, nicotine and paracetamol were prepared using this method on QCM sensor surfaces [17,53–55]. Application areas of MIP–QCM biosensors The three most common application areas for MIP– QCM sensors are clinical diagnosis, environmental monitoring and control of enantiomeric separation. Most studies describe detection in buffer or organic solvents indicating the early stage in development of these devices with respect to real applications [17,40,43,51,52,56,57]. Although it is possible to detect small molecules in buffered or organic solutions, it is also important to determine the amount of a particular drug or any other chemical in body fluids. For instance, Tan et al. determined the amount of nicotine and paracetamol in human serum and urine [55,58]. The detection limit of nicotine was found to be 25 nm using an imprinted polymer-coated sensor. Wu and colleagues fabricated bilirubin specific imprinted poly- mers on a QCM sensor using a photo-graft poly- merization method [41]. Bilirubin concentration is considered an important index to identify liver dis- eases. A bilirubin-specific sensor was challenged with both bilirubin and its analogue biliverdin, cross-reac- tivity of bilirubin versus biliverdin binding was 31 : 20. Yan et al. [59] developed a MIP–QCM sensor for daminozide (a potentially carcinogenic chemical the detection of which is important in food safety) with a detection limit of 50 pgÆmL )1 . MIP–QCM sensors for acetaldehyde, monoterpenes and bisphenol A have been prepared for environmental pollution detection by different groups [48,60,61]. Synthesis of enantiomer- ically pure organic compounds is very important for industrial production. MIP–QCM sensors capable of enantiomeric discrimination are very useful tools for process end-point analysis and various groups have discriminated between R- and S-propranolol, l- and d-tryptophan, l- and d-serine and l- and d-dansylphe- nylalanine enantiomers [17,44,49]. Summary The inherent robustness, ease of manufacture and high capacity of MIPs make them a potentially useful alternative for small molecule detection using piezo- electric biosensors. Although the majority of applica- tions involve the use of buffered pure solutions rather than real clinical or environmental samples for detec- tion, this perhaps simply reflects the early stage of development of the technology. Selectivity is still a significant issue for imprinted polymers and this can hamper specific, sensitive detection of analytes in Y. Uludag ˘ et al. Detection of low molecular mass analytes FEBS Journal 274 (2007) 5471–5480 ª 2007 Akubio Ltd. Journal compilation ª 2007 FEBS 5477 complex fluids. Polymerization methods are critical determinants of selectivity and overall assay perfor- mance of MIPs and as there is no general procedure for MIP preparation, each template requires optimiza- tion of several parameters to fabricate reproducible, high-performance sensor surfaces. It is expected that the improvements to polymerization techniques should greatly enhance the selectivity and binding capacity of the MIP–QCM sensors. In many ways, the stage of development of MIP interfaces is reflected in the state of development of robust piezoelectric biosensors compared with analo- gous robust electrochemical and optical biosensors that have benefited from more than two decades and several billion dollars of research and development. Many researchers rely on piezoelectric devices built in-house that are generally very sensitive to artefacts including temperature drift, humidity and pressure effects, resulting in less reliable and less reproducible measurements. Commercially available systems that minimize these effects with physical compensation on in-line referencing are appearing, but they are still far from the idealized portable device for near patient testing, point of care or remote environmental moni- toring in a handheld device. However, the fundamen- tal elements of acoustic sensors, i.e. cost of goods, simplicity and ease of manufacture, clearly have the potential to be exploited in an integrated portable device for use in quantitative detection of low molec- ular mass analytes. It is expected that recent advances in understanding acoustic biosensing technology will generate such devices that can be combined with the benefits of MIPs as the recognition element to pro- vide a new generation of sensors for use in routine measurements outside the controlled laboratory envi- ronment. References 1 Turner APF (2007) Encyclopedia of Chemical Technology. Wiley, New York, NY. 2 Myszka DG, Jonsen MD & Graves BJ (1998) Equilib- rium analysis of high affinity interactions using BIA- CORE. Anal Biochem 265, 326–330. 3 Myszka DG & Rich RL (2000) Implementing surface plasmon resonance biosensors in drug discovery. Pharm Sci Techn Today 3, 310–317. 4 Cooper MA (2002) Optical biosensors in drug discov- ery. Nat Rev Drug Discov 1, 515–528. 5 Richter RP, Him JLK, Tessier B, Tessier C & Brisson AR (2005) On the kinetics of adsorption and two-dimensional self-assembly of annexin A5 on sup- ported lipid bilayers. Biophys J 89, 3372–3385. 6 Long YM, Li WF, Nie LH & Yao SZ (2001) Ion-selective piezoelectric sensor for niacinamide assay in serum and urine. J Pharmaceut Biomed Anal 24, 361–369. 7 Godber B, Thompson KS, Rehak M, Uludag Y, Kelling S, Sleptsov A, Frogley M, Wiehler K, Whalen C & Cooper MA (2005) Direct quantification of analyte con- centration by resonant acoustic profiling. Clin Chem 51, 1962–1972. 8 Li X, Thompson KS, Godber B & Cooper MA (2006) Quantification of small molecule-receptor affinities and kinetics by acoustic profiling. Assay Drug Dev Technol 4, 565–573. 9 Zhou AH & Muthuswamy J (2004) Acoustic biosensor for monitoring antibody immobilization and neurotrans- mitter GABA in real-time. Sens Actuators B Chem 101, 8–19. 10 Kurosawa S, Park JW, Aizawa H, Wakida SI, Tao H & Ishihara K (2006) Quartz crystal microbalance immuno- sensors for environmental monitoring. Biosens Bioelec- tron 22, 473–481. 11 Sauerbrey G (1959) Verwendung von Schwingquarzen zur Wagung dunner Schichten und zur Microwagang. Z Phys 155, 206–212. 12 Nomura T & Okuhara M (1982) Frequency-shifts of piezoelectric quartz crystals immersed in organic liquids. Anal Chim Acta 142, 281–284. 13 Kanazawa KK (1997) Mechanical behaviour of films on the quartz microbalance. Faraday Discussions, 77–90. 14 Dickert FL, Forth P, Bulst WE, Fischerauer G & Knauer U (1998) SAW devices-sensitivity enhancement in going from 80 MHz to 1 GHz. Sens Actuators B Chem 46, 120–125. 15 Janshoff A, Galla HJ & Steinem C (2000) Piezoelectric mass-sensing devices as biosensors – an alternative to optical biosensors? Angew Chem Int Ed 39, 4004–4032. 16 Furtado LM, Su HB, Thompson M, Mack DP & Hay- ward GL (1999) Interactions of HIV-1 TAR RNA with Tat-derived peptides discriminated by on-line acoustic wave detector. Anal Chem 71, 1167–1175. 17 Percival CJ, Stanley S, Braithwaite A, Newton MI & McHale G (2002) Molecular imprinted polymer coated QCM for the detection of nandrolone. Analyst 127, 1024–1026. 18 Hook F, Ray A, Norden B & Kasemo B (2001) Characterization of PNA and DNA immobilization and subsequent hybridization with DNA using acoustic- shear-wave attenuation measurements. Langmuir 17, 8305–8312. 19 Zhou XD, Liu LJ, Hu M, Wang LL & Hu JM (2002) Detection of hepatitis B virus by piezoelectric biosensor. J Pharmaceut Biomed Anal 27 , 341–345. 20 Fung YS & Wong YY (2001) Self-assembled monolay- ers as the coating in a quartz piezoelectric crystal Detection of low molecular mass analytes Y. Uludag ˘ et al. 5478 FEBS Journal 274 (2007) 5471–5480 ª 2007 Akubio Ltd. Journal compilation ª 2007 FEBS immunosensor to detect salmonella in aqueous solution. Anal Chem 73, 5302–5309. 21 Richert L, Lavalle P, Vautier D, Senger B, Stoltz JF, Schaaf P, Voegel JC & Picart C (2002) Cell interactions with polyelectrolyte multilayer films. Biomacromolecules 3, 1170–1178. 22 Liu Y, Yu X, Zhao R, Shangguan D-H, Bo Z & Liu G (2003) Real time kinetic analysis of the interaction between immunoglobulin G and histidine using quartz crystal microbalance biosensor in solution. Biosens Bioelectron 18, 1419–1427. 23 Halamek J, Hepel M & Skladal P (2001) Investigation of highly sensitive piezoelectric immunosensors for 2,4- dichlorophenoxyacetic acid. Biosens Bioelectron 16, 253–260. 24 Nakamura C, Song S-H, Chang S-M, Sugimoto N & Miyake J (2002) Quartz crystal microbalance sensor tar- geting low molecular weight compounds using oligopep- tide binder and peptide-immobilized latex beads. Anal Chim Acta 469, 183–188. 25 Karousos NG, Aouabdi S, Way AS & Reddy SM (2002) Quartz crystal microbalance determination of organophosphorus and carbamate pesticides. Anal Chim Acta 469, 189–196. 26 Carmon KS, Baltus RE & Luck LA (2004) A piezoelec- tric quartz crystal biosensor: the use of two single cyste- ine mutants of the periplasmic Escherichia coli glucose ⁄ galactose receptor as target proteins for the detection of glucose. Biochemistry 43, 14249–14256. 27 Wang X, Ellis JS, Lyle EL, Sundaram P & Thompson M (2006) Conformational chemistry of surface-attached calmodulin detected by acoustic shear wave propaga- tion. Mol Biosystems 2, 184–192. 28 Ban T, Morigaki K, Yagi H, Kawasaki T, Kobayashi A, Yuba S, Naiki H & Goto Y (2006) Real-time and single fibril observation of the formation of amyloid beta spherulitic structures. J Biol Chem 281, 33677–33683. 29 Saraswathi R, Hillman AR & Martin SJ (1999) Mechanical resonance effects in electroactive polycar- bazole films. J Electroanalyt Chem 460, 267–272. 30 Hook F, Kasemo B, Nylander T, Fant C, Sott K & Elwing H (2001) Variations in coupled water, viscoelastic properties, and film thickness of a Mefp-1 protein film during adsorption and cross-linking: a quartz crystal microbalance with dissipation monitoring, ellipsometry, and surface plasmon resonance study. Anal Chem 73, 5796–5804. 31 Bu ¨ ttgenbach S, Rabea J, Zimmermann B & Hauptmann P (2001) High-frequency thickness-shear mode resonators for sensor application in liquids, Proceedings of the SPIE Photonics East Conference, Volume 207. Boston, MA. 32 Brown MJ, Hillman AR, Martin SJ, Cernosek RW & Bandey HL (2000) Manipulation of electroactive polymer film viscoelasticity: the roles of applied poten- tial and frequency. J Mater Chem 10, 115–126. 33 Wulff G (1986) Molecular recognition in polymers pre- pared by imprinting with templates. ACS Symposium Series, pp. 186–230. American Chemical Society, Washington, DC. 34 Haupt K & Mosbach K (1998) Plastic antibodies: devel- opments and applications. Trends Biotechnol 16, 468–475. 35 Piletsky SA, Turner NW & Laitenberger P (2006) Molecularly imprinted polymers in clinical diagnostics ) future potential and existing problems. Med Eng Phys 28, 971–977. 36 Shiomi T, Matsui M, Mizukami F & Sakaguchi K (2005) A method for the molecular imprinting of hemo- globin on silica surfaces using silanes. Biomaterials 26, 5564–5571. 37 Piletsky SA, Matuschewski H, Schedler U, Wilpert A, Piletsky EV, Thiele TA & Ulbricht M (2000) Surface functionalization of porous polypropylene membranes with molecularly imprinted polymers by photograft copolymerization in water. Macromolecules 33, 3092–3098. 38 Wu L & Li Y (2004) Metal ion-mediated molecular- imprinting polymer for indirect recognition of formate, acetate and propionate. Anal Chim Acta 517, 145–151. 39 Yan M & Ramstrom O (2005) Design and analysis of molecular imprinting, in Molecularly imprinted materi- als: science and technology, pp. 363–394. Marcel Dekker, New York, NY. 40 Kugimiya A & Takeuchi T (2001) Surface plasmon res- onance sensor using molecularly imprinted polymer for detection of sialic acid. Biosens Bioelectron 16, 1059–1062. 41 Wu AH & Syu MJ (2006) Synthesis of bilirubin imprinted polymer thin film for the continuous detection of bilirubin in an MIP ⁄ QCM ⁄ FIA system. Biosens Bio- electron 21, 2345–2353. 42 Lin CY, Tai DF & Wu TZ (2003) Discrimination of peptides by using a molecularly imprinted piezoelectric biosensor. Chem Eur J 9, 5107–5110. 43 Ling TR, Yau ZS, Tasi YC, Chou TC & Liu CC (2005) Size-selective recognition of catecholamines by molecu- lar imprinting on silica–alumina gel. Biosens Bioelectron 21, 901–907. 44 Liu F, Liu X, Ng SC & Chan HSO (2006) Enantioselec- tive molecular imprinting polymer coated QCM for the recognition of 1-tryptophan. Sens Actuators B Chem 113, 234–240. 45 Piacham T, Josell A, Arwin H, Prachayasittikul V & Ye L (2005) Molecularly imprinted polymer thin films on quartz crystal microbalance using a surface bound photo-radical initiator. Anal Chim Acta 536, 191–196. 46 Cao L, Zhou XC & Li SFY (2001) Enantioselective sensor based on microgravimetric quartz crystal Y. Uludag ˘ et al. Detection of low molecular mass analytes FEBS Journal 274 (2007) 5471–5480 ª 2007 Akubio Ltd. Journal compilation ª 2007 FEBS 5479 microbalance with molecularly imprinted polymer film. Analyst 126, 184–188. 47 Liu K, Wei WZ, Zeng JX, Liu XY & Gao YP (2006) Application of a novel electrosynthesized polydop- amine-imprinted film to the capacitive sensing of nico- tine. Anal Bioanalyt Chem 385, 724–729. 48 Percival CJ, Stanley S, Galle M, Braithwaite A, Newton MI, McHale G & Hayes W (2001) Molecular-imprinted, polymer-coated quartz crystal microbalances for the detection of terpenes. Anal Chem 73, 4225–4228. 49 Haupt K, Noworyta K & Kutner W (1999) Imprinted polymer-based enantioselective acoustic sensor using a quartz crystal microbalance. Anal Commun 36, 391–393. 50 Sallacan N, Zayats M, Bourenko T, Kharitonov AB & Willner I (2002) Imprinting of nucleotide and monosac- charide recognition sites in acrylamidephenylboronic acid–acrylamide copolymer membranes associated with electronic transducers. Anal Chem 74, 702–712. 51 Feng L, Liu Y, Tan Y & Hu J (2004) Biosensor for the determination of sorbitol based on molecularly imprinted electrosynthesized polymers. Biosens Bioelec- tron 19, 1513–1519. 52 Liao HP, Zhang ZH, Li H, Nie LH & Yao SZ (2004) Preparation of the molecularly imprinted polymers- based capacitive sensor specific for tegafur and its characterization by electrochemical impedance and piezoelectric quartz crystal microbalance. Electrochim Acta 49, 4101–4107. 53 Chianella I, Piletsky SA, Tothill IE, Chen B & Turner AP (2003) MIP-based solid phase extraction cartridges combined with MIP-based sensors for the detection of microcystin-LR. Biosens Bioelectron 18, 119–127. 54 Tan Y, Peng H, Liang C & Yao S (2001) New assay system for phenacetin using biomimic bulk acoustic wave sensor with a molecularly imprinted polymer coat- ing. Sens Actuators B Chem 73, 179–184. 55 Tan Y, Yin J, Liang C, Peng H, Nie L & Yao S (2001) A study of a new TSM bio-mimetic sensor using a molecularly imprinted polymer coating and its applica- tion for the determination of nicotine in human serum and urine. Bioelectrochemistry 53, 141–148. 56 Ersoz A, Denizli A, Ozcan A & Say R (2005) Molecu- larly imprinted ligand-exchange recognition assay of glucose by quartz crystal microbalance. Biosens Bioelec- tron 20, 2197–2202. 57 Pogorelova SP, Bourenko T, Kharitonov AB & Willner I (2002) Selective sensing of triazine herbicides in imprinted membranes using ion-sensitive field-effect transistors and microgravimetric quartz crystal micro- balance measurements. Analyst 127, 1484–1491. 58 Tan YG, Zhou ZL, Wang P, Nie LH & Yao SZ (2001) A study of a bio-mimetic recognition material for the BAW sensor by molecular imprinting and its applica- tion for the determination of paracetamol in the human serum and urine. Talanta 55, 337–347. 59 Yan S, Fang Y & Gao Z (2007) Quartz crystal micro- balance for the determination of daminozide using molecularly imprinted polymers as recognition element. Biosens Bioelectron 22, 1087–1091. 60 Hirayama K, Sakai Y, Kameoka K, Noda K & Naganawa R (2002) Preparation of a sensor device with specific recognition sites for acetaldehyde by molecular imprinting technique. Sens Actuators B Chem 86, 20–25. 61 Tsuruoka M, Yano K, Ikebukuro K, McNiven S, Takeuchi T & Karube I (1996) Rapid detection of com- plementary- and mismatched DNA sequences using flu- orescence polarization. Anal Lett 29, 1741–1749. 62 Dickert FL, Lieberzeit P, Miarecka SG, Mann KJ, Hay- den O & Palfinger C (2004) Synthetic receptors for chemical sensors – subnano- and micrometre patterning by imprinting techniques. Biosens Bioelectron 20, 1040–1044. Detection of low molecular mass analytes Y. Uludag ˘ et al. 5480 FEBS Journal 274 (2007) 5471–5480 ª 2007 Akubio Ltd. Journal compilation ª 2007 FEBS . MINIREVIEW Piezoelectric sensors based on molecular imprinted polymers for detection of low molecular mass analytes Yildiz Uludag ˘ 1,2 ,. combination of molecularly imprinted polymers as recognition elements with piezoelectric biosensors for quantita- tive detection of small molecules. Analytes