Article pubs.acs.org/Langmuir
Molecularly Designed Layer-by-Layer (LbL) Films to Detect Catechol Using Information Visualization Methods Pedro H. B. Aoki, † Priscila Alessio, † Leonardo N. Furini, † Carlos J. L. Constantino, Tácito T. A. T. Neves,‡ Fernando V. Paulovich,‡ Maria Cristina F. de Oliveira,‡ and Osvaldo N. Oliveira, Jr.*,§
†
†
Faculdade de Ciências e Tecnologia, UNESP, Presidente Prudente, SP, 19060-900, Brazil Instituto de Ciências Matemáticas e de Computaçaõ , USP, CP 668, 13560-970 São Carlos, SP, Brazil § Instituto de Física de São Carlos, SP, USP, CP 369, 13560-970 São Carlos, Brazil ‡
S Supporting Information *
ABSTRACT: The control of molecular architectures has been exploited in layer-by-layer (LbL) films deposited on Au interdigitated electrodes, thus forming an electronic tongue (e-tongue) system that reached an unprecedented high sensitivity (down to 10−12 M) in detecting catechol. Such high sensitivity was made possible upon using units containing the enzyme tyrosinase, which interacted specifically with catechol, and by processing impedance spectroscopy data with information visualization methods. These latter methods, including the parallel coordinates technique, were also useful for identifying the major contributors to the high distinguishing ability toward catechol. Among several film architectures tested, the most efficient had a tyrosinase layer deposited atop LbL films of alternating layers of dioctadecyldimethylammonium bromide (DODAB) and 1,2-dipalmitoyl-sn-3-glycero-fosfo-rac-(1-glycerol) (DPPG), viz., (DODAB/DPPG)5/DODAB/Tyr. The latter represents a more suitable medium for immobilizing tyrosinase when compared to conventional polyelectrolytes. Furthermore, the distinction was more effective at low frequencies where double-layer effects on the film/liquid sample dominate the electrical response. Because the optimization of film architectures based on information visualization is completely generic, the approach presented here may be extended to designing architectures for other types of applications in addition to sensing and biosensing.
■
INTRODUCTION The design of new supramolecular materials has been at the forefront of materials science, especially in nanotechnology efforts. There are many methods of producing such materials, but there has been an emphasis on fabrication techniques that allow for control of molecular architectures, including the Langmuir−Blodgett (LB)1,2 and the electrostatic layer-by-layer (LbL)3,4 methods. In some cases, the final properties of the fabricated materials can mimic the highest-performance natural materials, as in the LbL films made with carbon nanotube composites5 and clay nanosheets6 that reached record strength. Efficient polymer light-emitting diodes were obtained with LbL films where molecular-scale engineering of charge-injection layers was exploited to generate graded electronic profiles.7 In fact, for organic electronics devices the LbL method has been proven suitable to control exciton diffusion8 and energy transfer.9 Other examples of controlled architectures include film coatings with antireflection, antifogging, and self-cleaning properties, produced with all-nanoparticle LbL films,10 and the hollow spheres with controlled size and shape that can be used in drug delivery.11 Indeed, the search for new biological © 2013 American Chemical Society
applications has driven developments in novel architectures, in various instances incorporating proteins12 or serving to control cell interactions.13 LbL assembly was also exploited to produce functional electrically conducting networks starting with 1D nanostructures with such a degree of control that cross points in the networks could be addressed individually.14 A common feature in the work mentioned above is the combination of distinct materials in a single film architecture, normally found when seeking synergy among the film components. The importance of controlling architecture has led to the coinage of the term nanoarchitectonics,15,16 which is defined as “a technology system to be used for arranging nanoscale structural units, i.e., the nanostructure unit as a group of atoms or molecules in a predesignated configuration”.17 Numerous materials can indeed be explored within the nanoarchitectonics paradigm, from the obvious organic Special Issue: Interfacial Nanoarchitectonics Received: November 14, 2012 Revised: January 1, 2013 Published: January 28, 2013 7542
dx.doi.org/10.1021/la304544d | Langmuir 2013, 29, 7542−7550
Langmuir
Article
Figure 1. Proposed molecular architectures for the LbL films containing Tyr, which is always immobilized on the outer layer. The inset shows the molecular structures of PAH (mer), PSS (mer), DPPG, DODAB, and catechol.
and the parallel coordinates33 that have been used in feature selection in optimization processes.29
materials to mesoporous structures and inorganic components.18 Equally wide is the spectrum of possible applications, including dielectric nanosheets for nanoelectronics,19 the removal of arsenic from water,20 and sensing.21 With such a variety of materials and purposes, the optimization of the properties of nanoarchitectonics products becomes crucial. This requires not only experimental methods to characterize the properties of the system but also techniques to analyze the large amounts of data normally generated in the characterization process. In this article, we report on supramolecular architectures obtained with the LbL method that are used to detect catechol with impedance spectroscopy measurements. Sensor arrays were fabricated with LbL films of distinct materials in order to explore the cross-sensitivity of electronic tongues (e-tongues),22 which is again an application of nanoarchitectonics. In particular, we compare the performance of two types of molecular architectures, one of which includes an enzyme capable of specific interactions with catechol. The importance of detecting catechol has already been highlighted in the literature because phenolic compounds are broadly used in wood preservatives, textiles, herbicides, and pesticides23,24 and may be water pollutants.23−25 The enzyme chosen was tyrosinase (Tyr), which is known to be efficient in detecting phenolic compounds such as catechol, chlorophenol, and phenol.24 To reach high performance, Tyr was immobilized in LbL films containing liposomes made with dioctadecyldimethylammonium bromide (DODAB) and 1,2-dipalmitoyl-sn-3glycero-fosfo-rac-(1-glycerol) (DPPG) because they are charged structures amenable to forming films based on electrostatic interactions. Furthermore, these lipids are used in simple models of biological membranes for their ability to preserve the activity of biomolecules.26,27 In addition to demonstrating a high sensitivity for the sensor array containing the biosensor, made possible with the use of information visualization methods28,29 to process the impedance spectroscopy data, we discuss how the performance can be optimized. With regard to the latter, we build upon our previous experience in applying sophisticated computational methods for sensing and biosensing.30,31 These methods include multidimensional projection techniques32 employing nonlinear cost functions that have been proven suitable for biosensing
■
MATERIALS AND METHODS
Materials. Anionic phospholipid DPPG was purchased from Genzyme Pharmaceuticals. Cationic lipid DODAB, anionic poly(sodium 4-styrenesulfonate) (PSS), and cationic poly(allylamine hydrochloride) (PAH) polyelectrolytes were acquired from SigmaAldrich. Catechol was acquired from Panreac. All chemicals were used without further purification. The molecular weights of DPPG, DODAB, PAH, PSS, and catechol are 744.95, 630.95, 15 000, 70 000, and 110.11 g/mol, respectively, and their molecular structures are shown in the inset of Figure 1. Tyrosinase (Tyr) from mushroom EC 232-653-4 with a noted activity of 5370 U/mg for the solid (catalog no. T3824-250KU) was acquired from Sigma-Aldrich. LbL Film Fabrication. The LbL films were grown using solutions and dispersions prepared with ultrapure water (18.2 MΩ·cm and pH 6.2) obtained from a Milli-Q system, model Simplicity. The 0.74 mg/ mL (1.0 mM) concentration of the DPPG dispersion was prepared without any special procedure: the powder was simply added to ultrapure water at room temperature (22 °C), and the dispersion was gently stirred. The same methodology was applied to preparing solutions of PAH and PSS, both at 1.0 mg/mL. The 0.63 mg/mL (1.0 mM) concentration of the DODAB dispersion was prepared following the methodology presented by Feitosa et al.:34 DODAB powder was added to ultrapure water, and the mixture was heated under gentle stirring to 60 °C (above the DODAB Tm of ∼45 °C) for a few minutes. Then, the DODAB dispersion was cooled to room temperature (22 °C). The PAH/DPPG LbL film was grown following the procedure reported by Aoki et al.35 The DODAB/DPPG LbL films were produced adjusting the latter procedure.35 Basically, the substrate was sequentially immersed in a DODAB dispersion (3 min), ultrapure water that was gently stirred to remove excess adsorbed DODAB (1 min), a DPPG dispersion (3 min), and ultrapure water to remove excess adsorbed DPPG (1 min). Then the first bilayer of DODAB/ DPPG was formed, and the multilayer DODAB/DPPG LbL film was grown by repeating this four-step sequence. The same methodology was applied to grow PAH/PSS LbL films with DODAB being replaced with PAH and DPPG being replaced with PSS. The enzyme Tyr was always immobilized onto the outer layer as demonstrated to be suitable for enzymatic biosensors reported by Caseli et al.36 and Crespilho et al.,37 leading to the LbL film architectures shown in Figure 1. The Tyr enzyme was dissolved in phosphate buffer (pH 7.1) in a concentration of 0.0175 mg/mL (1.3672 × 10−7 mol/L), where the LbL films were allowed to soak for 15 min for Tyr adsorption. 7543
dx.doi.org/10.1021/la304544d | Langmuir 2013, 29, 7542−7550
Langmuir
Article
that biosensors made with enzymes immobilized in LbL films are stable for 4−6 weeks.41,42 Data Analysis. Data from impedance spectroscopy from e-tongues and other types of sensors and biosensors are normally treated with statistical and computational methods, including multidimensional projection techniques,32 which are aimed at reducing the dimensionality of the data in order to derive a visual representation. A technique commonly employed when analyzing data for sensing is principal component analysis (PCA), a linear statistical dimension reduction method that seeks to identify linear combinations of the data attributes that account for most of the data covariance.43 Using this type of method is essential to exploiting the cross-sensitivity from different sensing units in arrays (such as in e-tongues and e-noses).44 We have found recently that for biosensing data nonlinear projection techniques usually provide better support for data analysis assisted by informative visual representations, probably because specific interactions between film components in the sensing units and the analytes introduce nonlinearities in the electrical response (review in Oliveira et al.29). Typical procedures inherent in the multidimensional projections are as follows. A projection technique allows the mapping of each data element into a graphical marker, which is placed on a plane in a way in which markers representing “similar” instances are placed close to each other. Conversely, dissimilar instances are placed far apart, and this requires a measure of dissimilarity in the original high-dimensional data space. In sensing and biosensing, the Euclidean distance is normally employed to convey dissimilarity. For a mathematical formulation of the projection scheme, let us assume that a data set is represented by X = {x1, x2,..., xn}, with δ(xi, xj) being the dissimilarity (distance) between two data instances i and j. The projected graphical markers corresponding to X are represented by Y = {y1, y2,..., yn}, which are determined through an injective function f = X → Y that attempts to make |δ(xi, xj) − d( f(xi), f(yj))| ≈ 0, ∀xi, xj ∈ X,45 where d(yi, yj) is the distance function on the projected plane. In this article, we present visualizations obtained with two nonlinear techniques, namely, Sammon’s mapping33 and IDMAP.46 The error functions minimized in Sammon’s mapping and IDMAP are, respectively,
Optimizing the Immobilization of Tyr. The Tyr immobilization was monitored by mass adsorption using a Stanford Research Systems Inc. model QCM 200 quartz crystal microbalance. In QCM, shifts in frequency were continuously followed with time, and the mass gain (Δm, g) was obtained using the Sauerbrey equation:38 ΔF = (−2.3 × 106F02Δm)/A, where ΔF is the frequency shift (Hz), F0 is the initial frequency (without a coating, MHz), and A is the electrode area (cm2). After Tyr immobilization, the LbL films were washed using either ultrapure water or buffer and monitored by QCM, as shown in Figure SI 1 in the Supporting Information. We noted that 100% of Tyr was removed from the (PAH/PSS)5/Tyr LbL film upon washing with water (Figure SI 1a). However, 45% (by mass) of the Tyr remained on the (PAH/PSS)5/Tyr LbL film when it was washed in buffer (Figure SI 1b). The latter could be enhanced to 60% of the remaining mass of Tyr if it was immobilized onto PAH as the last layer, forming a (PAH/ PSS)5/PAH/Tyr LbL film (Figure SI 1c). By replacing PSS with DPPG (i.e., in a (PAH/DPPG)5/PAH/Tyr LbL film), the remaining mass of Tyr reached 80%, as shown in Figure SI 2a. The latter dropped slightly to 75% when PAH was replaced with DODAB in the (DODAB/DPPG)5/DODAB/Tyr LbL film, as shown in Figure SI 2b. The effectiveness of the adsorption process for Tyr and its stability in the LbL film clearly depended on the choice of molecular architecture and the pH of the washing solutions. Therefore, they depend on the electrostatic interactions between the polyelectrolytes or DPPG and Tyr, whose isoelectric point is 4.7−5.39 At pH 6.2 (ultrapure water), Tyr was already negatively charged, and at pH 7.1 (buffer), Tyr is even more negatively charged whereas PAH is highly ionized at both pH values.40 However, secondary interactions surely play some role in allowing for negatively charged Tyr to be immobilized onto PSS or DPPG. Sensing Experiments. The electronic tongue applied here was composed of four sensing units, which are basically formed by LbL films deposited onto Au interdigitated electrodes (IDEs) characterized by impedance spectroscopy, as in the pioneering work by Riul et al.22 The IDE is composed of 100 digits separated by 10 μm from each other, forming 50 pairs. The dimensions of each digit are 0.5 mm in length, 10 μm in width, and 100 nm in height, as illustrated by a cartoon in Figure SI 3a. The following LbL films were deposited onto the electrodes: (PAH/PSS)5/PAH/Tyr, (PAH/DPPG)5/PAH/Tyr, and (DODAB/DPPG)5/DODAB/Tyr. The fourth sensing unit was a bare Au IDE, which is applied to check the effect from the LbL films on the electrical responses (real capacitance vs frequency). This array of sensing units allows one to explore the performance of biosensors with the enzyme Tyr immobilized either on a matrix of polyelectrolytes or on a matrix of lipids. In a second set of experiments, we directly compared the performance of a sensor made with a (DODAB/DPPG)5/DODAB LbL film and a biosensor made with a (DODAB/DPPG)5/DODAB/Tyr LbL film in order to verify the importance of the enzyme Tyr. Impedance spectroscopy measurements were carried out with a Solartron 1260A impedance analyzer. All sensing experiments were carried out by recording impedance curves as a function of frequency from 1 to 106 Hz with an ac voltage of 50 mV at 22 °C. Prior to the LbL deposition, several bare Au IDEs were subjected to impedance measurements in buffer to check the reproducibility of their signal in order to ensure that the main changes observed in the electrical responses from each sensing unit are due to the distinct LbL films instead of differences among the Au IDEs themselves. The real capacitance versus frequency curves for the four bare Au IDEs chosen are shown in Figure SI 3b (Supporting Information). After LbL deposition, three consecutive measurements were performed with sensing units immersed in the buffer used as reference and into catechol solutions at 10−12, 5 × 10−12, 10−11, 5 × 10−11, 10−10, 5 × 10−10, and 10−9. The measurements started with the buffer and then from the lowest (10−12 M) to the highest (10−9 M) concentration of catechol solution. The sensing units were left soaking 20 min inside the solutions before data acquisition to enable a stable reading. In this article, we did not make a systematic stability study of the sensing units, but in similar work in the literature, it has been shown
SSam =
1 ∑i < j δ(xi , xj)
∑
(d(yi , yj ) − δ(xi , xj))2 δ(xi , xj)
where δ and d denote the distance functions as explained above and
SIDMAP =
δ(xi , xj) − δmin δmax − δmin
− d(yi , yj )
where δmin and δmax are the minimum and maximum distances between the data instances. Further details about the use of multidimensional projections for sensing and biosensing are provided in the review by Oliveira et al.29
■
RESULTS AND DISCUSSION The strong dependence on the matrix for immobilizing enzymes in biosensors has been known for a long time, but it is difficult to quantify the relative importance of distinct film components. Here we compare biosensors with Tyr deposited onto LbL films of polyelectrolytes and phospholipids, in addition to seeking a high sensitivity with the sensor array containing all sensing units. The real capacitance versus frequency curves for sensing units (PAH/PSS) 5/PAH/Tyr, (PAH/DPPG)5/PAH/Tyr, (DODAB/DPPG)5/DODAB/Tyr, and bare Au IDE immersed in different concentrations of catechol in buffer (10−12, 5 × 10−12, 10−11, 5 × 10−11, 10−10, 5 × 10−10, and 10−9 mol/L) are shown in Figure 2. Taylor and Macdonald47 proposed an equivalent circuit to model an electrode coated with organic films immersed in an electrolyte solution, with the capacitance of the equivalent circuit being analyzed as a function of the frequency of the ac bias voltage. Basically, at frequencies 7544
dx.doi.org/10.1021/la304544d | Langmuir 2013, 29, 7542−7550
Langmuir
Article
Figure 3. IDMAP multidimensional projection grouping the results by different concentrations of catechol solutions considering the data measured with all sensing units.
is made as a case study to demonstrate the capability of information visualization methods in this scenario. Figure 4 shows the plot where the capacitance data from Figure 2 were treated using Sammon’s mapping33 multidimensional projection. The electrical responses from each sensing unit and for each solution are plotted individually. Now, circle colors identify the sensing unit, as shown. Because the closer the circles the more similar the electrical responses, the highest distinction ability is displayed by the biosensing unit formed with a lipid matrix, viz., the (DODAB/DPPG)5/DODAB/Tyr LbL film, where the red circles are scattered all over the diagram plot. In contrast, the bare Au IDE was not able to distinguish the solutions, as expected (blue circles). These results confirm the importance not only of the ultrathin films deposited onto the IDE but also the characteristics of the molecules forming the matrix where the biomolecule is immobilized regarding the biosensing performance. The importance of the matrix where the enzyme is immobilized was further investigated using a new set of sensing units, again using catechol as the analyte to be detected. Figure 5 shows the electrical responses (real capacitance vs frequency, Figure SI 4, Supporting Information) after being treated with the IDMAP projection for a sensing unit composed of Au IDE coated with a (DODAB/DPPG)5/DODAB LbL film and a biosensing unit with the same type of LbL film containing Tyr immobilized onto the outer layer. On the left-hand side of the plot, one notes that the distinction of the various samples was relatively poorer for sensor 2, in comparison to the more efficient distinction obtained with biosensor 2 whose data appear on the right-hand side of the plot. This was to be expected because the specific interaction with catechol involved in biosensing tends to lead to a higher sensitivity. By the same token, the variability among samples with the same concentration was lower for sensor 2 because a higher reproducibility reflects the lower sensitivity in this case. Note that in this article we used the term sensitivity interchangeably with distinguishing ability rather than using a calibration curve as is normal in the field of sensing. Among the various visualization methods useful in assessing the influence of molecular architecture on the film properties is the so-called parallel coordinates,54 which is a technique aimed at helping users to identify patterns visually and to understand relationships between the different dimensions (also called attributes or features) in multidimensional data. Parallel
Figure 2. Real capacitance vs frequency curves for sensing units (PAH/PSS) 5 /PAH/Tyr, (PAH/DPPG) 5 /PAH/Tyr, (DODAB/ DPPG)5/DODAB/Tyr, and bare Au IDE immersed in different concentrations of catechol in buffer (10−12, 5 × 10−12, 10−11, 5 × 10−11, 10−10, 5 × 10−10, and 10−9 mol/L). The experiments were always performed starting from the lower concentration.
between 1 and 102 Hz capacitance effects are governed by the formation of an electrical double layer at the film/solution interface; between 102 and 105 Hz, the electrical response is dominated by the properties of the film coating the electrodes; above 105 Hz, the geometric capacitance of the electrodes is the most important factor.48 The dielectric constant of the solution also plays a role in frequencies higher than 105 Hz.49 The capacitance curves can be shifted to higher or lower frequencies depending on both the ion concentration in solution,35,50 which clearly is the case here (buffer medium), and the electrical characteristics of the films coating the electrode.51 A visual inspection of Figure 2 does not allow one to infer whether it is possible to distinguish the samples with different catechol concentrations. In fact, when measurements are performed in duplicate or triplicate with nominally identical sensing units or distinct days, the dispersion in the data is already within a few percent, which would apparently hamper any distinction among samples with small analyte concentrations. This is the reason that one has to resort to statistical or computational methods for analyzing the data. Indeed, the distinction ability is demonstrated by treating the capacitance data in Figure 2 using the IDMAP multidimensional projection technique, whose results are given in Figure 3. It is clear that the electronic tongue can easily distinguish catechol solutions even down to 10−12 mol/L. To our knowledge, this is the lowest concentration ever detected for catechol. In a search of the literature, we found reports of electrochemical sensors with catechol detection in the range from 10−3 M down to 10−10 M25,52,53 and an earlier report from our group using optical absorbance and impedance spectroscopy, with detection down to 10−10 M.25 The difference between the latter work and this paper is in the optimization of the processing of the impedance spectroscopy data, where the whole capacitance versus frequency plots were analyzed. We further analyze the data obtained in order to identify the contribution from each sensing unit for the distinction ability, which is important to optimizing the sensor array. The analysis 7545
dx.doi.org/10.1021/la304544d | Langmuir 2013, 29, 7542−7550
Langmuir
Article
Figure 4. Sammon’s mapping layout of the impedance measurements obtained from the distinct sensing units.
Figure 5. IDMAP multidimensional projection considering sensing units (DODAB/DPPG)5/DODAB/Tyr and (DODAB/DPPG)5/DODAB.
is indicated by the better spatial separation of the lines and also by the little boxes placed above the axes in the plot, which visually indicate a quantitative measurement for the distinguishability known as the silhouette value S, which is calculated with eq 1
coordinates have already been employed to analyze the performance of biosensors,28 where feature selection was required. Here we confirm the expectation from the physicochemical analysis of impedance spectroscopy results obtained with organic films immersed in a liquid because a higher distinction capability is observed at low frequencies as indicated in the parallel coordinates plot of Figure 6. In this plot, all frequency curves are mapped to the 2D space by drawing m equally spaced axes, with each one representing one of the m frequencies. Each capacitance curve will be shown as a polygonal line that intersects each axis at the point corresponding to its measured value at that particular frequency. Axes are arranged from lower to higher frequencies. The data in Figure 6 are the same as in Figure 5, only the curves are colored on the basis of the compound concentration and the type of sensor. The better distinction at low frequencies
S=
1 n
n
∑ i=1
(bi − ai) max(ai , bi)
(1)
where ai is the average of the computed distances between the frequency curve of the ith sample and the curves of all of the samples with the same concentration as I, bi is the minimum distance between the frequency curve of the ith sample and the curves of all of the other samples with different concentrations from i, and n is the total number of samples. 7546
dx.doi.org/10.1021/la304544d | Langmuir 2013, 29, 7542−7550
Langmuir
Article
Figure 6. Parallel coordinate plots for the data obtained with sensor 2 and biosensor 2 data.
Figure 7. IDMAP multidimensional projection of measurements obtained in data sets 1 and 2.
S takes values in the range of [−1, 1], with higher values indicating a better discrimination capability. A fully filled blue box indicates the maximum value S = 1, a fully filled red box indicates the minimum value S = −1, and an empty white box indicates S = 0. Again, one may note visually that biosensor 2 has a better distinction capability than sensor 2, as one should expect from the plot in Figure 5. It is also clear from Figure 6 that the best distinction is achieved with measurements at low frequencies. We recall that the region with frequencies below 100 Hz corresponds to the contribution to the signal from the double layer.47 For taste sensors and electronic tongues whose principle of detection is
impedance spectroscopy, the distinction capability is normally low at high frequencies, with the exception of samples where strong analyte adsorption occurs on the film surface, as is the case for coffee samples.55 In the intermediate frequency range, the sensitivity can be high because changes in the liquid sample may affect the film properties. Obviously, however, the stronger impact should occur for the double-layer contribution, which varies with changes in both the film vicinity and the liquid sample. The key role played by the matrix where the enzyme is immobilized could be further analyzed with Figure 7, which shows the capacitance data recorded for all sensing units 7547
dx.doi.org/10.1021/la304544d | Langmuir 2013, 29, 7542−7550
Langmuir
Article
the combination of a film component capable of the molecular recognition of catechol with a suitable matrix in the LbL film. Significantly, the visualization methodsmultidimensional projections and parallel coordinatescan also be used to optimize sensor performance. In this article, they were used to confirm that a biosensor (i.e., a unit containing Tyr) was more efficient than a sensor without the enzyme. Furthermore, with the parallel coordinates technique we could quantitatively determine the distinction ability of sensors at different frequencies in the impedance spectroscopy measurements. For the data analyzed, we found that the best distinction was possible at lower frequencies, where the electrical response is dominated by double-layer effects. Therefore, one envisages a further enhancement in sensitivity upon selecting the best frequencies for distinction, which may allow the detection of concentrations even lower than 10−12 M in future studies. The high sensitivity achieved does not mean that the sensor arrays presented here are selective for catechol, and we plan to address this issue in further experiments with similar compounds, such as phenol, resorcinol, hydroquinone, and salicylic acid. Finally, we emphasize that all methods used here for LbL films containing an enzyme and for detecting catechol using impedance spectroscopy measurements are absolutely generic. The optimization procedure using information visualization methods can be applied to any other type of film architecture to detect other analytes and to use other principles of detection. The approach we presented can actually be adapted to any nanoarchitectonics endeavor in which molecular architectures are to be developed for any classification task, which includes but is not limited to sensing and biosensing.
previously discussed in this work (experiments with sets 1 and 2), projected with IDMAP. The performance of the following sensing units could be compared: bare Au IDE, (DODAB/ DPPG)5/DODAB, (PAH/PSS)5/PAH/Tyr, (PAH/DPPG)5/ PAH/Tyr, and (DODAB/DPPG)5/DODAB/Tyr (duplicate). Biosensor (DODAB/DPPG)5/DODAB/Tyr presented the best performance, consistent with the discussion above. Note that the circles depicting measurements from samples at the same concentrations do not overlap because the experiments were carried out on different days using distinct, fresh solutions of catechol. Therefore, considering the very low solution concentrations and the high sensitivity of the system, we must discuss the reproducibility in terms of the trend (pattern) in the sensing unit response rather than absolute values. Another important feature is the similar performance of the sensor with (DODAB/DPPG)5/DODAB (red) and biosensor (PAH/ PSS)5/PAH/Tyr (light blue). In spite of the presence of Tyr in the latter, its stability was known to be poor (cf. experiments in Materials and Methods for the optimization of Tyr immobilization); therefore, the sensing performance was not significantly better than that of the corresponding LbL film without Tyr. The intermediate performance of biosensor (PAH/DPPG)5/PAH/Tyr, which contains a mixture of polyelectrolyte and lipid, is consistent with our discussion. Although the higher performance of the units containing Tyr can be readily attributed to the specific interaction with the analytes, the reasons that the matrix is also relevant are more difficult to establish. The most important factor is the preserved activity of the enzyme, and this is very dependent on the matrix. For instance, immobilization in LbL films has been extensively exploited because of the mild conditions needed for film fabrication and retention of entrapped water that helps preserve activity in biomolecules (review in Siqueira et al.56). Indeed, in many cases the use of LbL films has been proven to be superior to other immobilization strategies.56 In comparing the units containing Tyr but immobilized in distinct matrices, we noted a higher performance for the film with a lipid matrix. The interactions between Tyr and the other film components include electrostatic interactions, H bonding, and hydrophobic forces, which may occur for the polyelectrolytes as well as for the lipids. The advantage of lipid matrices may have to do with a more biologically friendly environment, but the precise reason for this has not been determined. Another immobilization procedure is the Langmuir−Blodgett (LB) technique, where enzymes are incorporated into lipid matrices that are also suitable for activity preservation.57−61 With regard to the molecular architecture, performance optimization may be reached with either enzymes in all of the bilayers in LbL films62 or only on the top layer.36 In the latter case, the relevant feature was the favorable electron transport to the electrode in electrochemical measurements. Because with impedance spectroscopy the interaction with the topmost layer is the most relevant, we chose to adsorb Tyr only on the last layer, which proved to be adequate for sensing catechol.
■
ASSOCIATED CONTENT
S Supporting Information *
QCM data, interdigitated electrode geometry, and real capacitance versus frequency curves. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Notes
The authors declare no competing financial interest.
■ ■
ACKNOWLEDGMENTS This work was supported by FAPESP, CNPq, CAPES, and rede nBioNet (Brazil). REFERENCES
(1) Langmuir, I. The Constitution and Fundamental Properties of Solids and Liquids. II. Liquids. J. Am. Chem. Soc. 1917, 39, 1848−1906. (2) Blodgett, K. B. Deposition of Successive Monomolecular Layers. J. Am. Chem. Soc. 1935, 57, 1007−1022. (3) Decher, G. Fuzzy Nanoassemblies: Toward Layered Polymeric Multicomposites. Science 1997, 277, 1232−1237. (4) Decher, G.; Hong, J. D.; Schmitt, J. Buildup of Ultrathin Multilayer Films by a Self-Assembly Process 0.3. Consecutively Alternating Adsorption of Anionic snd Cationic Polyelectrolytes on Charged Surfaces. Thin Solid Films 1992, 210, 831−835. (5) Shim, B. S.; Zhu, J.; Jan, E.; Critchley, K.; Ho, S.; Podsiadlo, P.; Sun, K.; Kotov, N. A. Multiparameter Structural Optimization of Single-Walled Carbon Nanotube Composites: Toward Record Strength, Stiffness, and Toughness. ACS Nano 2009, 3, 1711−1722. (6) Kotov, N. Clay Nanosheets Comprise New Material with Strength of Steel. Mater. Perform. 2008, 47, 20−21.
■
CONCLUSIONS The choice of suitable film architectures for sensing units has led to an e-tongue-type sensor array with unprecedented sensitivity toward catechol. Concentrations down to 10−12 M could be detected upon using a layer of tyrosinase (Tyr) on top of LbL films made with lipids and treating the capacitance data with information visualization methods. With regard to contributions akin to nanoarchitectonics, one may mention 7548
dx.doi.org/10.1021/la304544d | Langmuir 2013, 29, 7542−7550
Langmuir
Article
(7) Ho, P. K. H.; Kim, J. S.; Burroughes, J. H.; Becker, H.; Li, S. F. Y.; Brown, T. M.; Cacialli, F.; Friend, R. H. Molecular-Scale Interface Engineering for Polymer Light-Emitting Diodes. Nature 2000, 404, 481−484. (8) Favarim, H. R.; Spadacio, D.; Faceto, A. D.; Zucolotto, V.; Oliveira, O. N., Jr; Guimaraes, F. E. G. Energy-Modulated Heterostructures Made with Conjugated Polymers for Directional Energy Transfer and Carrier Confinement. Adv. Funct. Mater. 2007, 17, 2862−2868. (9) Postacchini, B. B.; Zucolotto, V.; Dias, F. B.; Monkman, A.; Oliveira, O. N., Jr. Energy Transfer in Nanostructured Films Containing Poly(p-phenylene vinylene) and Acceptor Species. J. Phys. Chem. C 2009, 113, 10303−10306. (10) Lee, D.; Rubner, M. F.; Cohen, R. E. All-nanoparticle thin-film coatings. Nano Lett. 2006, 6, 2305−2312. (11) Caruso, F.; Caruso, R. A.; Mohwald, H. Nanoengineering of Inorganic and Hybrid Hollow Spheres by Colloidal Templating. Science 1998, 282, 1111−1114. (12) Lvov, Y.; Ariga, K.; Ichinose, I.; Kunitake, T. Assembly of Multicomponent Protein Films by Means of Electrostatic Layer-byLayer Adsorption. J. Am. Chem. Soc. 1995, 117, 6117−6123. (13) Berg, M. C.; Yang, S. Y.; Hammond, P. T.; Rubner, M. F. Controlling Mammalian Cell Interactions on Patterned Polyelectrolyte Multilayer Surfaces. Langmuir 2004, 20, 1362−1368. (14) Huang, Y.; Duan, X. F.; Wei, Q. Q.; Lieber, C. M. Directed Assembly of One-Dimensional Nanostructures into Functional Networks. Science 2001, 291, 630−633. (15) Hecht, S. Welding, Organizing, and Planting Organic Molecules on Substrate Surfaces - Promising Approaches Towards Nanoarchitectonics from the Bottom Up. Angew. Chem., Int. Ed. 2003, 42, 24−26. (16) Ariga, K.; Lee, M. V.; Mori, T.; Yu, X.-Y.; Hill, J. P. TwoDimensional Nanoarchitectonics Based on Self-Assembly. Adv. Colloid Interface Sci. 2010, 154, 20−29. (17) Ariga, K.; Li, M.; Richards, G. J.; Hill, J. P. Nanoarchitectonics: A Conceptual Paradigm for Design and Synthesis of DimensionControlled Functional Nanomaterials. J. Nanosci. Nanotechnol. 2011, 11, 1−13. (18) Ariga, K.; Ji, Q.; McShane, M. J.; Lvov, Y. M.; Vinu, A.; Hill, J. P. Inorganic Nanoarchitectonics for Biological Applications. Chem. Mater. 2012, 24, 728−737. (19) Osada, M.; Sasaki, T. Two-Dimensional Dielectric Nanosheets: Novel Nanoelectronics from Nanocrystal Building Blocks. Adv. Mater. 2012, 24, 210−228. (20) Rajbhandari, R.; Shrestha, L. K.; Pradhananga, R. R. Nanoporous Activated Carbon Derived from Lapsi (Choerospondias Axillaris) Seed Stone for the Removal of Arsenic from Water. J. Nanosci. Nanotechnol. 2012, 12, 7002−7009. (21) Ariga, K.; Ishihara, S.; Abe, H.; Li, M.; Hill, J. P. Materials Nanoarchitectonics for Environmental Remediation and Sensing. J. Mater. Chem. 2012, 22, 2369−2377. (22) Riul, A., Jr; dos Santos, D. S., Jr; Wohnrath, K.; Di Tommazo, R.; Carvalho, A.; Fonseca, F. J.; Oliveira, O. N., Jr; Taylor, D. M.; Mattoso, L. H. C. Artificial Taste Sensor: Efficient Combination of Sensors Made from Langmuir-Blodgett Films of Conducting Polymers and a Ruthenium Complex and Self-Assembled Films of an Azobenzene-Containing Polymer. Langmuir 2002, 18, 239−245. (23) Yang, L. J.; Xiong, H. Y.; Zhang, X. H.; Wang, S. F. A Novel Tyrosinase Biosensor Based on Chitosan-Carbon-Coated Nickel Nanocomposite Film. Bioelectrochemistry 2012, 84, 44−48. (24) Karim, F.; Fakhruddin, A. N. M. Recent Advances in the Development of Biosensor for Phenol: A Review. Rev. Environ. Sci. Bio/Technol. 2012, 11, 261−274. (25) Zucolotto, V.; Pinto, A. P. A.; Tumolo, T.; Moraes, M. L.; Baptista, M. S.; Riul, A.; Araujo, A. P. U.; Oliveira, O. N., Jr Catechol Biosensing Using a Nanostructured Layer-by-Layer Film Containing Cl-Catechol 1,2-Dioxygenase. Biosens. Bioelectron. 2006, 21, 1320− 1326.
(26) Moraes, M. L.; Baptista, M. S.; Itri, R.; Zucolotto, V.; Oliveira, O. N., Jr. Immobilization of Liposomes in Nanostructured Layer-byLayer Films Containing Dendrimers. Mater. Sci. Eng., C 2008, 28, 467−471. (27) Yoshimoto, M.; Sakamoto, H.; Shirakami, H. Covalent Conjugation of Tetrameric Bovine Liver Catalase to Liposome Membranes for Stabilization of the Enzyme Tertiary and Quaternary Structures. Colloids Surf., B 2009, 69, 281−287. (28) Paulovich, F. V.; Moraes, M. L.; Maki, R. M.; Ferreira, M.; Oliveira, O. N., Jr; de Oliveira, M. C. F. Information Visualization Techniques for Sensing and Biosensing. Analyst 2011, 136, 1344− 1350. (29) Oliveira, O. N., Jr; Pavinatto, F. J.; Constantino, C. J. L.; Paulovich, F. V.; de Oliveira, M. C. F. Information Visualization to Enhance Sensitivity and Selectivity in Biosensing. Biointerphases 2012, 7, 53−. (30) Moraes, M. L.; Maki, R. M.; Paulovich, F. V.; Rodrigues Filho, U. P.; de Oliveira, M. C. F.; Riul, A., Jr; de Souza, N. C.; Ferreira, M.; Gomes, H. L.; Oliveira, O. N., Jr. Strategies to Optimize Biosensors Based on Impedance Spectroscopy to Detect Phytic Acid Using Layerby-Layer Films. Anal. Chem. 2010, 82, 3239−3246. (31) Perinoto, A. C.; Maki, R. M.; Colhone, M. C.; Santos, F. R.; Migliaccio, V.; Daghastanli, K. R.; Stabeli, R. G.; Ciancaglini, P.; Paulovich, F. V.; de Oliveira, M. C. F.; Oliveira, O. N., Jr; Zucolotto, V. Biosensors for Efficient Diagnosis of Leishmaniasis: Innovations in Bioanalytics for a Neglected Disease. Anal. Chem. 2010, 82, 9763− 9768. (32) de Oliveira, M. C. F.; Levkowitz, H. From Visual Data Exploration to Visual Data Mining: A Survey. IEEE Trans. Visualization Comput. Graph. 2003, 9, 378−394. (33) Sammon, J. W., Jr. A Nonlinear Mapping for Data Structure Analysis. IEEE Trans. Comput. 1969, C18, 401−409. (34) Feitosa, E.; Karlsson, G.; Edwards, K. Unilamellar Vesicles Obtained by Simply Mixing Dioctadecyldimethylammonium Chloride and Bromide with Water. Chem. Phys. Lipids 2006, 140, 66−74. (35) Aoki, P. H. B.; Volpati, D.; Riul, A., Jr; Caetano, W.; Constantino, C. J. L. Layer-by-Layer Technique as a New Approach to Produce Nanostructured Films Containing Phospholipids as Transducers in Sensing Applications. Langmuir 2009, 25, 2331−2338. (36) Caseli, L.; dos Santos, D. S.; Foschini, M.; Goncalves, D.; Oliveira, O. N., Jr. The Effect of the Layer Structure on the Activity of Immobilized Enzymes in Ultrathin Films. J. Colloid Interface Sci. 2006, 303, 326−331. (37) Crespilho, F. N.; Ghica, M. E.; Florescu, M.; Nart, F. C.; Oliveira, O. N., Jr.; Brett, C. M. A. A Strategy for Enzyme Immobilization on Layer-by-Layer Dendrimer-Gold Nanoparticle Electrocatalytic Membrane Incorporating Redox Mediator. Electrochem. Commun. 2006, 8, 1665−1670. (38) Sauerbrey, G. Verwendung Von Schwingquarzen zur Wagung Dunner Schichten und zur Mikrowagung. Z. Phys. 1959, 155, 206− 222. (39) Robb, D. A.; Gutteridge, S. Polypeptide Composition of 2 Fungal Tyrosinases. Phytochemistry 1981, 20, 1481−1485. (40) Shiratori, S. S.; Rubner, M. F. pH-Dependent Thickness Behavior of Sequentially Adsorbed Layers of Weak Polyelectrolytes. Macromolecules 2000, 33, 4213−4219. (41) Moraes, M. L.; Petri, L.; Oliveira, V.; Olivati, C. A.; de Oliveira, M. C. F.; Paulovich, F. V.; Oliveira, O. N., Jr; Ferreira, M. Detection of Glucose and Triglycerides Using Information Visualization Methods to Process Impedance Spectroscopy Data. Sens. Actuators, B 2012, 166, 231−238. (42) Moraes, M. L.; Rodrigues, U. P.; Oliveira, O. N., Jr; Ferreira, M. Immobilization of Uricase in Layer-by-Layer Films Used in Amperometric Biosensors for Uric Acid. J. Solid State Electrochem. 2007, 11, 1489−1495. (43) Jolliffe, I. T. Principal Component Analysis; Springer-Verlag: New York, 2002. (44) Reed, J. L.; Kaas, J. H. Statistical Analysis of Large-Scale Neuronal Recording Data. Neural Networks 2010, 23, 673−684. 7549
dx.doi.org/10.1021/la304544d | Langmuir 2013, 29, 7542−7550
Langmuir
Article
(45) Tejada, E.; Minghim, R.; Nonato, L. G. On Improved Projection Techniques to Support Visual Exploration of Multi-Dimensional Data Sets. Inf. Visualization 2003, 2, 218−231. (46) Minghim, R.; Paulovich, F. V.; Lopes, A. A. Content-Based Text Mapping Using Multi-Dimensional Projections for Exploration of Document Collections. In Proceedings of the Conference on Visualization and Data Analysis 2006 (EI10), San Jose, CA, Jan 15−19, 2006; Erbacher, R. F., Roberts, J. C., Grohn, M. T., Borner, K., Eds.; IS&T/ SPIE; Vol. 6060. (47) Taylor, D. M.; Macdonald, A. G. ac Admittance of the MetalInsulator-Electrolyte Interface. J. Phys. D: Appl. Phys. 1987, 20, 1277− 1283. (48) Riul, A.; Soto, A. M. G.; Mello, S. V.; Bone, S.; Taylor, D. M.; Mattoso, L. H. C. An Electronic Tongue Using Polypyrrole and Polyaniline. Synth. Met. 2003, 132, 109−116. (49) Zaretsky, M. C.; Mouayad, L.; Melcher, J. R. Continuum Properties from Interdigital Electrode Dielectrometry. IEEE Trans. Electric. Insul. 1988, 23, 897−917. (50) Volpati, D.; Aoki, P. H. B.; Dantas, C. A. R.; Paulovich, F. V.; de Oliveira, M. C. F.; Oliveira, O. N., Jr; Riul, A., Jr; Aroca, R. F.; Constantino, C. J. L. Toward the Optimization of an e-Tongue System Using Information Visualization: A Case Study with Perylene Tetracarboxylic Derivative Films in the Sensing Units. Langmuir 2012, 28, 1029−1040. (51) Aoki, P. H. B.; Alessio, P.; Riul, A., Jr.; De Saja Saez, J. A.; Constantino, C. J. L. Coupling Surface-Enhanced Resonance Raman Scattering and Electronic Tongue as Characterization Tools to Investigate Biological Membrane Mimetic Systems. Anal. Chem. 2010, 82, 3537−3546. (52) Ensafi, A. A.; Karimi-Maleh, H.; Mallakpour, S.; Rezaei, B. Highly Sensitive Voltammetric Sensor Based on Catechol-DerivativeMultiwall Carbon Nanotubes for the Catalytic Determination of Captopril in Patient Human Urine Samples. Colloids Surf., B 2011, 87, 480−488. (53) Mobin, S. M.; Sanghavi, B. J.; Srivastava, A. K.; Mathur, P.; Lahiri, G. K. Biomimetic Sensor for Certain Phenols Employing a Copper(II) Complex. Anal. Chem. 2010, 82, 5983−5992. (54) Inselberg, A.; Dimsdale, B. Parallel Coordinates: A Tool for Visualizing Multi-Dimensional Geometry, Visualization, 1990. Visualization ’90; In Proceedings of the First IEEE Conference on Visualization, Oct 23−26, 1990; IEEE Computer Society Press: Washington, DC; pp 361−378. (55) Ferreira, E. J.; Pereira, R. C. T.; Delbem, A. C. B.; Oliveira, O. N., Jr; Mattoso, L. H. C. Random Subspace Method for Analysing Coffee with Electronic Tongue. Electron. Lett. 2007, 43, 1138−1140. (56) Siqueira, J. R.; Caseli, L.; Crespilho, F. N.; Zucolotto, V.; Oliveira, O. N., Jr Immobilization of biomolecules on nanostructured films for biosensing. Biosens. Bioelectron. 2010, 25, 1254−1263. (57) Apetrei, C.; Alessio, P.; Constantino, C. J. L.; de Saja, J. A.; Rodriguez-Mendez, M. L.; Pavinatto, F. J.; Fernandes, E. G. R.; Zucolotto, V.; Oliveira, O. N., Jr. Biomimetic Biosensor Based on Lipidic Layers Containing Tyrosinase and Lutetium Bisphthalocyanine for the Detection of Antioxidants. Biosens. Bioelectron. 2011, 26, 2513− 2519. (58) Zanon, N. C. M.; Oliveira, O. N., Jr; Caseli, L. Immbolization of Uricase Enzyme in Langmuir and Langmuir-Blodgett Films of Fatty Acids: Possible Use as a Uric Acid Sensor. J. Colloid Interface Sci. 2012, 373, 69−74. (59) Girard-Egrot, A. P.; Godoy, S.; Blum, L. J. Enzyme Association with Lipidic Langmuir-Blodgett Films: Interests and Applications in Nanobioscience. Adv. Colloid Interface Sci. 2005, 116, 205−225. (60) Caseli, L.; Crespilho, F. N.; Nobre, T. M.; Zaniquelli, M. E. D.; Zucolotto, V.; Oliveira, O. N., Jr. Using Phospholipid Langmuir and Langmuir-Blodgett Films As Matrix for Urease Immobilization. J. Colloid Interface Sci. 2008, 319, 100−108. (61) Caseli, L.; Moraes, M. L.; Zucolotto, V.; Ferreira, M.; Nobre, T. M.; Zaniquelli, M. E. D.; Rodrigues, U. P.; Oliveira, O. N., Jr Fabrication of Phytic Acid Sensor Based on Mixed Phytase-Lipid Langmuir-Blodgett Films. Langmuir 2006, 22, 8501−8508.
(62) Ferreira, M.; Fiorito, P. A.; Oliveira, O. N., Jr; de Torresi, S. I. C. Enzyme-Mediated Amperometric Biosensors Prepared with the Layerby-Layer (LbL) Adsorption Technique. Biosens. Bioelectron. 2004, 19, 1611−1615.
7550
dx.doi.org/10.1021/la304544d | Langmuir 2013, 29, 7542−7550
