Nanolithography and subnanomolecular interactions for biomimetic sensors

Materials Science and Engineering C 26 (2006) 924 – 928 www.elsevier.com/locate/msec

Nanolithography and subnanomolecular interactions for biomimetic sensors O. Hayden a, D. Podlipna a, X. Chen a, S. Krassnig a, A. Leidl b, F.L. Dickert a,* a

Institute of Analytical Chemistry and Food Chemistry, Vienna University, Waehringer Str. 38, A-1090 Vienna, Austria b EPCOS AG, SAW RD IP, Anzingerstr. 13, D-81617 Muenchen, Germany Dedicated to Prof. Dr. Heinz Hoffman on the occassion of his 70th birthday Available online 26 October 2005

Abstract Template-directed moulding of polymer thin films is performed to generate artificial receptors directly on pre-coated piezoacoustic devices and interdigital capacitors (IDCs). This generally applicable approach is useful for the fabrication of robust nano- and microstructured sensor coatings. Exemplary label-free detection results of yeast and mammalian cells are shown with these biomimetic receptors allowing even single cell detection. Grid electrodes were applied to quartz crystal microbalances (QCM) screening the dielectric properties of viable Saccharomyces cerevisiae and unusual non-gravimetric frequency shifts are observed. Autocorrelation analysis of capacitive responses to yeast cells yields characteristic time constants related to cell trapping by the sensor surface. D 2005 Elsevier B.V. All rights reserved. Keywords: Molecular imprinting; Viruses; Yeast cells; Erythrocytes; Quartz crystal microbalance; Interdigital capacitor

1. Introduction Template-directed synthesis of synthetic receptors from highly crosslinked polymers is an innovative and generally applicable technique [1]. The synthesized materials are usually termed molecularly imprinted polymers (MIPs) [2]. The recognition sites in the bulk of these materials are formed by self-organizing principles during the polymerization process. The cavities in the polymer bulk remaining after the dissolution or evaporation of the small organic molecules in the polymer are rather rigid due to the high crosslinking and thus do not collapse. The selective chemical recognition of the templating species by re-incorporation is achieved by a geometrical fit in combination with a pronounced adhesion to the cavities based on noncovalent phenomena as electrostatics, hydrogen bonding, vander-Waals and hydrophobic forces. The geometrical dimensions range from diameters of molecules to micrometers characterizing cells. The selective interactions between analyte and sensor surface, however, are in any case due to subnanomolecular interactions. These MIPs are most often applied in analytical sciences for solid phase extraction or chromatographic separation. Now, an increasing number of publications are applying * Corresponding author. Tel.: +43 1 4277 52317; fax: +43 2243 3627315. E-mail address: [email protected] (F.L. Dickert). 0928-4931/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.msec.2005.09.014

this self-organization technology as an effective strategy to create receptors for chemical sensing [3,4]. Our work is motivated by the possibility to extend the imprinting concept with small organic molecules towards the recognition of biological specimen. Bioimprinting with bulky specimen as templates, such as viruses or cells, can rationally only be performed on the polymer surface due to diffusion limitations [5]. We therefore developed a surface imprinting concept of thin films based on a stamping procedure to mould the polymer surfaces during the polymer curing [6]. This

Fig. 1. Schematic illustration of surface imprinting using biotemplates directly on a pre-coated transducer.

O. Hayden et al. / Materials Science and Engineering C 26 (2006) 924 – 928

925

Fig. 2. Schemes of transducers applied for biomimetic sensing. (a) QCMs with two screen-printed channels, a conventional and a grid electrode. (b) High frequency SAWs for gas and liquid sensing are shown. (c) IDCs for dielectric measurements.

technique is related to soft- and imprint-lithography and is generally applicable to biological templates spanning the nanoand micrometer scale [7]. 2. Experimental Materials for surface imprinting are prepared by polymerizing a monomer solution until reaching the gel point. Polymers synthesized by either radical polymerization or polyaddition can be applied for this process [8]. The results presented were obtained with polyurethane composed of bisphenol A, 4,4Vdiisocyanatodiphenylmethane (30% trifunctional) and phloroglucinol as cross-linker. The purified compounds were used as received from MERCK. The gel-like pre-polymer is redissolved in tetrahydrofurane and diluted in an appropriate concentration for the transducer coating. Thin films of these pre-polymers are formed by either spin or drop-coating of the transducer. To perform the surface imprinting, a stamp covered with a layer of purified templates is prepared and pressed onto the polymerizing thin film during the curing (Fig. 1a and b). The critical point is to remove the stamp from the cured polymer layers without peeling-off the sensor layer. This is possible since stamp and surface of the sensor layer are slightly separated by the bulky bio-templates. The aqueous solution can wet the void between thin film and stamp making possible a non-destructive lift-off. The templates remaining on the polymer surface are removed by

brief rinsing in an ultrasonic bath (Fig. 1c). To minimize a covalent embedding of templating structures reactive isocyanato groups are either blocked by brief exposure to humid air or by blocking agents with amino groups. Mass producible and highly sensitive transducers are used for the sensing experiments, quartz crystal microbalances (QCMs), surface acoustic wave resonators (SAWs) and interdigital capacitors (IDCs). Two or three gold electrodes were screen printed on a single quartz substrate to fabricate a QCM with a fundamental resonance frequency of 10 MHz. The multielectrode structure is used for differential measurements to compensate for temperature, viscosity and pressure fluctuations. Electrodes facing the aqueous solution have larger diameter than the electrodes oriented to the gas phase to minimize conductivity effects [9]. In contrast to this electrode geometry of QCMs, we fabricated grid electrodes facing the aqueous phase to increase the stray-field influence. This allows us to monitor dielectric properties of the enriched analytes (Fig. 2a). Additionally, high frequency shear wave SAWs were applied for ultra-sensitive mass resolution in the picogram range. The LiTaO3 SAWs had a fundamental resonance frequency of 428 MHz. Here, surface acoustic shear waves are used avoiding the high damping observed for Rayleigh waves [10]. Piezoacoustic measurements are performed with home-built oscillator circuits and a network analyzer. A precision LCR meter is used for impedance and capacitance measurements

Fig. 3. AFM micrographs and section analyses of molded polyurethane thin films by surface imprinting (Digital Instruments Nanoscope III atomic force microscope). (a) The TMV MIP is imprinted with a sub-monolayer of viruses. The virus imprint depth corresponds to half the TMV diameter. (b) A single S. cerevisiae imprint from a packed sensor layer is shown. Around 20% of the yeast diameter moulds the polyurethane layer. (c) The red blood cell imprint has a lower mechanical stiffness due to the missing cell wall and thus rather shallow imprint depths are obtained compared to yeast imprints.

O. Hayden et al. / Materials Science and Engineering C 26 (2006) 924 – 928

Frequency [Hz]

926

300 200

K2

100

K1

0 0

200

400

600

800

1000

1200

TMV [µg/ml] Fig. 4. 10 MHz QCM, adhesion of TMV viruses to the imprinted coating— strong (K1) and a less pronounced (K2) adhesion to the sensor layer.

with IDCs. Gold electrodes are prepared by standard optical lithography (Fig. 2c). Saccharomyces cerevisiae are arrested with hydroxyurea for synchronization. Freshly prepared synchronized yeast samples in the early G1 (cells without bud) and early S (cells with small buds) phase are used for sensing experiments. Red blood cell samples were purchased from the Austrian Red Cross. 3. Results and discussion The moulded sensor layers reflect the geometry of the biotemplates, which is best observed by atomic force microscopy (AFM). Fig. 3 summarizes AFM results of surface imprinted layers obtained with viruses and cellular entities as biotemplates. Tobacco mosaic virus (TMV) is a model plant virus and has a rod-like shape, with an average length of 300 nm and a diameter of 18 nm. A sub-monolayer of templating TMV generates a nanostructured imprint pattern on a polyurethane surface (Fig. 3a). The virus imprints are the result of individual and bundles of viruses. Imprints of individual viruses in polyurethane have depths of half the virus diameter. Imprinting with microorganism and yeasts having a mechanically stabilizing cell wall is a very straightforward process and honeycomb-like patterns are usually obtained by the surface imprinting with imprint depths of up to 1.5 Am (Fig. 3b). Now, we are extending bioimprinting to templating mammalian cells, such as erythrocytes. Mammalian cells lack stabilizing cell walls and thus are fragile when exposed to mechanical stress in comparison to yeasts. However, surface imprints can be

achieved by careful handling and red blood cell imprints are generated with depths of more than 500 nm (Fig. 3c). As stated before, the imprinting is performed directly on the coated transducer, which reduces considerably necessary steps for developing a sensor. In Fig. 4 the characteristic for the enrichment of TMV viruses on an imprinted QCM-coating can be seen. A titration curve results with a characteristic end point indicating the formation of an approximate monolayer on the sensor. Further TMV are adsorbed with a minor binding strength leading to a multilayer arrangement. Complementary information of the adhesion process of cellular analytes is obtained depending on their mass and their dielectric properties of viable cells with both QCM and IDC measurements. Fig. 5 shows mass-sensitive measurements with a conventional QCM gold electrode geometry to minimize stray-field influences, and with a grid electrode to monitor dielectric properties of the cells close to the QCM electrode. The sensor effects with QCMs having imprinted, non-imprinted sensor layer on uncoated electrodes are compared. All measurements were performed at equal conditions with yeast cells suspended in distilled water. Methylene blue staining has been performed to ensure the viability of the yeast cells in these conditions during the time of the measurement. The yeast suspensions were injected in the flow cells and the sensor effect due to sedimentation was recorded. The sedimentation of cells on a non-imprinted layer (1.5 Am thickness) as well as on bare electrodes resulted in equal gravimetric frequency shifts on normal electrodes. With an imprinted thin film (3.3 Am thickness) a strong enhancement of the sensor effect is achieved. The yeast MIP sensor is highly selective and other microorganism, such as Gram-negative Escherichia coli or Gram-positive Leuconostoc oenus, are barely detectable [4]. The sensor results are reversible by brief rinsing with a flow rate of ¨ 100 ml/min. Contrary to the gravimetric frequency decrease recorded with the conventional electrode, we observe an enormous non-gravimetric frequency [11] increase on grid electrodes due to the yeast sedimentation. Almost no sensor effect is observed on bare grid electrodes and the sensor effect on non-imprinted layers is a nearly nonspecific adsorption on conventional QCM electrodes. The extreme difference of the frequency shifts with imprinted sensor layers is also observed during passive network analyzer measure-

Fig. 5. QCM measurements with (a) an unconverntional and (b) a grid electrode facing to the liquid phase (closed squares: S. cervevisiae MIP coating; open squares: bare electrodes; open circles: non-imprinted layer). The sedimentation of yeasts from a suspension of 1 108 cells/ml in distilled water is observed after injection.

O. Hayden et al. / Materials Science and Engineering C 26 (2006) 924 – 928

927

Fig. 6. Autocorrelation plots of capacitive time responses performed with S. cerevisiae MIP coated IDCs. The IDCs had finger electrode spacings and electrode widths of 10 Am, respectively. The yeast suspensions (20 mg/ml) in distilled water were continuously flowing with a flow rate of 10 ml/min in a flow cell volume of 200 Al: (a) non-budding yeasts and (b) budding yeasts. (c) Autocorrelation cross-check of piston pump influences with distilled water.

ments. The increase of the serial resonance frequency is accompanied by a pronounced phase shift greater than + 10with grid electrodes. Phase shifts smaller than + 5- are occurring with gravimetric sensor responses. Since these measurements were performed with identical MIP coatings and conditions, the results can only originate from changes in the field strength at edges. Unusual effects are also observed for ionic solutions when normal electrodes are applied and these findings are more pronounced for grid structures [12,13]. IDC measurements were performed with microfluidic flow cells. The flow direction was perpendicular to the finger electrodes. A variety of finger electrode spacing, flow rates and flow cell volumes were used to observe the trapping of yeasts on imprinted surfaces. Here, preliminary results with synchronized yeast cell samples are shown. To distinguish random from correlated movement of yeast cells along the patterned sensor layer, autocorrelation of capacitive measurements was performed. If random processes occur, the autocorrelation function of time dependent sensor responses should be nearly zero. If non-random, then the autocorrelations will be significantly non-zero, which can be interpreted as characteristic residence times of the cells adhered to the imprinted layers. Only steady portions of the capacitive measurements with a minimum of 500 data points and a sampling rate of 0.8 data

Fig. 7. Specific and nonspecific erythrocyte adsorption on a QCM having two coated conventional electrodes (closed squares: red blood cell imprinted layer; open squares: non-imprinted polyurethane layer). The network analyzer measurement shows the adhesion of erythrocytes from an isotonic suspension (pH 7.2) of 6  108 cells/ml.

points/s have been used for autocorrelation calculations. All autocorrelation plots of the capacitance time series with the sample pumped through the flow cell with a piston pump show that the time series is not random and a characteristic time constant at the beginning can be seen. (Fig. 6). Only in the case, when the synchronized yeast cells correspond to the templating yeast cells, we can observe an additional modulation of the autocorrelation function (Fig. 6a). This modulation disappears when budding yeast cells are present (Fig. 6b) since due to the nanometer buds these cells cannot be effectively engulfed by the patterned sensor coating. To exclude any influences of the piston pump the autocorrelation was investigated in identical conditions without cells as shown in Fig. 6c. Simultaneously performed optical microscopy studies revealed no influence of sedimenting yeast cells with these flow rates. Characteristic time constants in the autocorrelation functions only occur with IDC finger spacing with a distance corresponding to the yeast cell diameter of ¨ 5 Am. No specific effects can be observed with finger electrode distances smaller than the cell diameter as well as larger than ¨ 3 times the yeast size. Thus, we assume that cells flowing close to the imprinted sensor surface roll over the imprinted pits and get trapped in imprinted pits of appropriate size. The average residence time of yeast cells due to this behaviour can be deduced from the autocorrelation function to be approximately ¨ 30 s. This time constant depends on flow rate, flow cell volume, concentration and IDC structure width. The periodic increase of the maxima in the autocorrelation function according to Fig. 6c can be interpreted as a multiple of the average trapping time.

Fig. 8. Sensor response of a 428 MHz shear wave SAW to the adhesion of a single HeLa tumor cell on the coated delay range of the device.

928

O. Hayden et al. / Materials Science and Engineering C 26 (2006) 924 – 928

The chemical recognition of extremely bulky analytes, such as mammalian cells with surface imprinted layers, is a challenging new task. As an example, the specific and nonspecific adhesion of red blood cells measured with a network analyzer is shown in Fig. 7. The erythrocytes are bulky doughnut-shaped specimen and thus a pronounced nonspecific sensor effect is expected to occur on coated QCMs. However, only minor sensor effects are observed on non-imprinted layers in comparison to the red blood cell imprinted thin film. In spite of the high flexibility of the blood cells, the results show that selectivity is achievable due to interactions on the molecular scale. QCMs operating in the lower MHz regime have detection limits of a few nanogram. Higher sensitivities in the picogram range can be achieved with high frequency devices, such as SAWs. These extreme sensitivities are demonstrated in Fig. 8 for the detection of a single HeLa tumor cell in buffered suspension. 4. Conclusion A general surface imprinting concept based on bio-template directed moulding of thin films has been described. Nano- and microstructured coatings are fabricated by surface imprinting reflecting the size and geometry and molecular interaction sites of the templates. These imprinted materials allow in combination with appropriate transducers, such as QCMs, SAWs and IDCs, the design of sensitive and selective sensors based on chemical recognition and biomimetic strategies. Even single cell detection as shown for a Hela cell can be performed with these layers when a high frequency SAW is used. The yeast

adhesion on MIP coated grid electrodes on QCMs reveal unusual non-gravimetric effects, which are influenced by the dielectric properties of viable cells. In addition, autocorrelation of capacitive IDC responses reveal characteristic time constants of yeast cell trappings by the layers. The recognition of mammalian cells, such as tumor and blood cells, is introduced emphasizing the generally applicable concept of surface imprinting. Acknowledgement This work was supported by the FWF, the Austrian Science Fund, project number P15512. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13]

K. Mosbach, Tr. Bioch. Sci. 19 (1994) 9. L. Ye, K. Haupt, Anal. Bioanal. Chem. 378 (2004) 1887. K. Haupt, K. Mosbach, Chem. Rev. 100 (2000) 2495. O. Henry, D. Cullen, S. Piletsky, Anal. Bioanal. Chem. 382 (2005) 1618. A. Aherne, C. Alexander, M.J. Payne, N. Perez, E.N. Vulfson, J. Am. Chem. Soc. 118 (1996) 8771. O. Hayden, F.L. Dickert, Adv. Mater. 12 (2001) 311. R.S. Kane, S. Takayama, E. Ostuni, D.E. Ingber, G.M. Whitesides, Biomaterials 20 (1999) 2363. B. Ebarvia, F. Sevilla, Sens. Actuators, B, Chem. 107 (2005) 782. M. Rodahl, F. Ho¨o¨k, B. Kasemo, Anal. Chem. 68 (1996) 2219. F.L. Dickert, M. Tortschanoff, W.-E. Bulst, G. Fischerauer, Anal. Chem. 71 (1999) 4559. R. Lucklum, C. Behling, P. Hauptmann, Anal. Chem. 71 (1999) 2488. A.V. Mamishev, K. Sundara-Rajan, F. Yang, Y. Du, M. Zahn, Proc. IEEE 92 (2004) 808. F.L. Dickert, K. Halikias, O. Hayden, L. Ping, R. Sikorski, Sens. Actuators, B, Chem. 76 (2001) 295.