Information Visualization and Feature Selection ... - ACS Publications

Subscriber access provided by CORNELL UNIVERSITY LIBRARY

Article

Information Visualization and Feature Selection Methods Applied to Detect Gliadin in Gluten-Containing Foodstuff with a Microfluidic Electronic Tongue Cristiane Margarete Daikuzono, Flavio Makoto Shimizu, Alexandra Manzoli, Antonio Riul, Maria Helena de Oliveira Piazzetta, Angelo Luiz Gobbi, Daniel S Correa, Fernando V. Paulovich, and Osvaldo Novais Oliveira, Jr. ACS Appl. Mater. Interfaces, Just Accepted Manuscript • Publication Date (Web): 08 May 2017 Downloaded from http://pubs.acs.org on May 13, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

ACS Applied Materials & Interfaces is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 24

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

Information Visualization and Feature Selection Methods Applied to Detect Gliadin in Gluten-Containing Foodstuff with a Microfluidic Electronic Tongue Cristiane M. Daikuzono,∗,†,‡ Flavio M. Shimizu,∗,† Alexandra Manzoli,¶ Antonio Riul Jr,§ Maria H. O. Piazzetta,k Angelo L. Gobbi,§ Daniel S. Correa,¶ Fernando V. Paulovich,⊥ and Osvaldo N. Oliveira Jr∗,† †São Carlos Institute of Physics, University of São Paulo, P.O Box 369, 13560-970 São Carlos, SP, Brazil ‡São Carlos School of Engineering, University of São Paulo, 13560-000, São Carlos, SP, Brazil ¶Nanotechnology National Laboratory for Agriculture (LNNA), Embrapa Instrumenta¸caõ, 13560-970, São Carlos, SP, Brazil §DFA, IFGW, Universidade Estadual de Campinas/Unicamp, 13083-859 Campinas, SP, Brazil kLNNano, Centro Nacional de Pesquisa em Energia e Materiais/CNPEM, 13083-970 Campinas, SP, Brazil ⊥Institute of Mathematical Sciences and Computing, University of São Paulo, 13566-590 São Carlos, SP, Brazil E-mail: [email protected]; [email protected]; [email protected]

1

ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Abstract The fast growth of celiac disease diagnosis has sparked the production of glutenfree food and the search for reliable methods to detect gluten in foodstuff. In this paper, we report on a microfluidic electronic tongue (e-tongue) capable of detecting trace amounts of gliadin, a protein of gluten, down to 0.005 mg.kg−1 in ethanol solutions, and distinguishing between gluten-free and gluten-containing foodstuff. In some cases it is even possible to determine whether gluten-free foodstuff has been contaminated with gliadin. That was made possible with an e-tongue comprising four sensing units, three of which made of layer-by-layer (LbL) films of semiconducting polymers deposited onto gold interdigitated electrodes placed inside microchannels. Impedance spectroscopy was employed as the principle of detection, and the electrical capacitance data collected with the e-tongue were treated with information visualization techniques with feature selection for optimizing performance. The sensing units are disposable to avoid cross-contamination as gliadin adsorbs irreversibly onto the LbL films according to polarization-modulated infrared reflection absorption spectroscopy (PM-IRRAS) analysis. Small amounts of material are required to produce the nanostructured films, however, and the e-tongue methodology is promising for low cost, reliable detection of gliadin and other gluten constituents in foodstuff.

Keywords electronic tongue, microfluidics, gliadin, celiac disease, impedance spectroscopy, feature selection, information visualization

1

Introduction

Celiac disease (CD), also known as gluten-sensitive enteropathy, is an autoimmune disorder whose diagnosis has increased consistently in recent years, reaching 1% of population worldwide nowadays. 1 It affects mainly people intolerant to gluten, 2,3 which is a protein containing 2


Page 2 of 24

Page 3 of 24

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


glutamines (insoluble fraction) and prolines (soluble fraction). 4 The only treatment available for CD is a gluten-free diet, which is obviously challenging because gluten is present in cereals such as wheat, rye, barley, oat, malt and their derivatives. Needless to say, the food industry has to comply with regulation for properly labeling food contents. Still, consumers should have to take precautions in their own hands, since inadequately labeled food or inadvertently gluten contaminated food remains a threat. Providing consumers with such means will only be possible if cheap, reliable methods become available to detect whether a sample of food or drink contains gluten. Detection of gluten has been mostly done indirectly by determining the presence of gliadin, a protein fraction present in gluten. 5,6 Techniques such as polymerase chain reaction (PCR), 7,8 mass spectroscopy (MS), 9 high performance liquid chromatography (HPLC) 10–12 and enzyme linked immunosorbent assay (ELISA) 6 are effective for the task. 9,13,14 However, these are expensive, time consuming techniques that require trained laboratory personnel. Viable alternatives for easy-to-use, low cost diagnostic devices may arise from electroanalytical sensors, based on immobilization of biomolecules such as lipids, 5,15 antibody 16–19 and DNA 20,21 to detect gliadin in foodstuff down to 0.5 mg.kg−1 . The drawback lies in the need of a conventional three electrode setup, electrolyte solution and sometimes a marker for signal amplification in these experiments. In this paper, we present a simpler, fast methodology with no need of biomolecule immobilization for specific recognition of species, by using a microfluidic electronic tongue (e-tongue) 22,23 based on the global selectivity concept. E-tongues have been exploited in biomedical, 24,25 pharmaceutical, 26 monitoring of chemical reactions 27,28 and chemical/bioanalysis. 28–34 In their microfluidic version, e-tongues are advantageous for requiring small volumes for sampling and discharge, and for allowing integration with other methodologies. High sensitivity can be reached owing to the ultrathin nature of the sensing units, 35 and here gliadin was easily detected in ethanol solutions of foodstuffs down to 0.005 mg.kg−1 using electrical impedance measurements. The analysis of impedance data was performed with the aid of multidimensional information visualization methods 36,37 where feature selection was

3



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 24

used, similarly to machine learning applications. 38 The mechanisms behind gliadin sensing were also investigated using atomic force microscopy (AFM) and polarization-modulated infrared reflection absorption spectroscopy (PM-IRRAS).

2 2.1

Experimental Electrode Modification

Poly(allyllamine hydrochloride) (PAH), poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS), polypyrrole (PPy), copper (II) tetrasulphonatedphthalocyanine (CuTsPc) and wheat gliadin were purchased from Sigma-Aldrich. All solutions were prepared using ultrapure water from a Millipore Direct-Q5 system. PAH/CuTsPc films were obtained from 1

aqueous solutions of PAH and CuTsPc at concentrations of 0.5 mg.mL− at pH = 8. 39 PAH/PEDOT:PSS films were made from aqueous solutions of PAH and PEDOT:PSS at 1

concentrations of 3.0 and 0.1 mg.mL− , respectively, at pH = 3.5. 40 PAH/PPy films were 1

obtained from aqueous solutions of PAH and PPy at concentrations of 0.5 mg.mL− and 1

1

1.1 mg.mL− , respectively, both prepared with a 0.5 mol.L− NaCl solution 41 without pH correction. PDMS microchannels and gold interdigitated electrodes (IDEs) were fabricated at the LNNano (CNPEM) onto glass slides. The microchannels were 490 µm wide, 50 µm high and 12.5 mm long, while the IDEs had 30 pairs of fingers, 40 µm wide, 3 mm long with an inter-electrode gap of 40 µm. The LbL films were deposited within the PDMS mi1

crochannel under a flow rate of 103 µL.h− . The solutions were injected alternately into the microchannel using a Hamilton microsyringe with the aid of an infusion syringe pump (New Era pump systems). Figure 1 depicts the procedure to fabricate the thin film architecture on IDEs inside the microchannel using the dynamic LbL technique. 31

4


Page 5 of 24

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 1: Films are deposited via dynamic LbL method 29 by injection of polyelectrolyte solutions into the microchannel.

2.2

Sample preparation

Food samples containing gluten (toast, salt biscuit, beer and mix of cereals Neston) and gluten free (powder milk Molico, powder milk Ninho, sake and corn flour Cremogema) were purchased at a local supermarket in São Carlos, Brazil. Gliadin standard solutions and food samples were prepared in ethanol 70% (v/v) and mixed in a vortex for about 10 min. The supernatant was carefully removed and centrifuged in a Hanil, Continent R model centrifuge, for 20 min at 5000 rpm. The gliadin solutions and the supernatant of food samples were analyzed using a UV-vis spectrophotometer (UV-160, Shimadzu) to verify the presence of gliadin at 280 nm.

2.3

Film characterization

The PDMS microchannel was removed from the device with the aid of a scalpel to verify the presence of the deposited LbL films, and a surface analysis of the IDEs was carried out using a scanning electron microscopy (SEM), JEOL model 6510. The coverage of the electrode surface was studied with atomic force microscopy (AFM), model Dimension V (Veeco), using 1

silicon tips in a microcantilever with spring constant of 5.6 nm− and resonance frequency

5



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

of 180 kHz. The AFM images were generated in the intermittent contact mode (tappingT M ) with scan frequency of 0.5 Hz. In situ polarization-modulated infrared reflection absorption spectroscopy (PM-IRRAS) was also used to confirm both the LbL film adsorption on the IDEs and gliadin adsorption on the LbL film. A KSV spectrometer (model PMI 550) was employed to obtain the spectra for each film using a HgCdTe detector, PCI-3TE -10.6 model, silicon carbide lamp and a photoelastic modulator of ZnSe crystal.

2.4

Detection

The electronic tongue comprised chips coated with 5-bilayer LbL films of PAH/CuTsPc, PAH/PEDOT:PSS, PAH/PPy LbL films, in addition to an uncoated electrode chip. Solutions with different concentrations of gliadin (0.005 mg.kg−1 , 5 mg.kg−1 , 10 mg.kg−1 , 20 mg.kg−1 , 30 mg.kg−1 , 50 mg.kg−1 , 100 mg.kg−1 , 200 mg.kg−1 , 227 mg.kg−1 , 570 mg.kg−1 , 770 mg.kg−1 , 950 mg.kg−1 and 1310 mg.kg−1 ) were dissolved in ethanol 70% (v/v), following the procedure described by Peres et al., 5 and analyzed using impedance spectroscopy in the frequency range from 1 Hz to 1 MHz, using a voltage amplitude of 20 mV, in a Solartron SI 1260. Multidimensional projection techniques implemented in PEx-Sensors software 36 were used to analyze the impedance data to distinguish between different solutions and for analyzing the UV-vis. data. In these techniques, a capacitance or UV-Vis. spectrum is reduced to a single data point in the 2D projection, and one tries to minimize the loss of information as the number of dimensions is reduced. The overall aim is to place data points representing similar spectra close to each other. A full description of these techniques and of PEx-Sensors can be found in. 36

3

Results and discussion

The darker regions in the SEM images in Figure 2 confirm the LbL film deposition along the microchannel and on the gold IDEs for PAH/CuTsPc, PAH/PEDOT:PSS and PAH/PPy

6


Page 6 of 24

Page 7 of 24

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 2: SEM images of 5-bilayer LbL films deposited on IDEs: (a) PAH/CuTsPc, (b) PAH/PPy and (c) PAH/PEDOT:PSS.

films. The films appear to be homogeneous, but the AFM images in Figures S1 through S3 in Supplementary Material show larger accumulation of material on the film edges, which was expected because the flow speed is much higher at the center region inside the microchannel.

3.1

Information Visualization

The microfluidic e-tongue was capable of distinguishing low concentrations of gliadin in ethanol solutions, starting at 0.005 mg.kg−1 up to 1310 mg.kg−1 . Because of the large amount of data collected with four sensing units (three LbL modified electrodes and one bare electrode) for many samples, information visualization methods and optimization procedures were employed to enhance data analysis. Using the PEx-Sensors suite of tools 37,43,44 we tested various multidimensional projection techniques such as Principal Component Analysis (PCA), 37 Sammon’s mapping 45 and IDMAP (Interactive Document Map) 46 to analyze the capacitance data, taking Euclidean distances as the metric for dissimilarity between samples or sensing units. Due to the high-dimensionality of the produced data, the distinguishing ability of a system may decrease if all the data frequencies or features are taken into account. To address this problem, we use a simple feature-selection procedure, which consisted in manually selecting the frequencies of the impedance measurements and the sensing units that mostly contributed to the distinguishing ability of the e-tongue system. Figure S4 shows parallel coordinate 47,48 plots used to guide the feature-selection process. The distinguishing 7



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ability was quantified in terms of a Silhouette coefficient S 49 defined as the average of the distances between each data instance and all other points belonging to the same cluster, and the minimum distance between each data instance and other instances belonging to other clusters. 36 The coefficient S varies between -1 to 1, with values near 1 meaning that the curves at the particular frequency are distinct from each other. When S ≈ 0, the data do not assist in distinguishing the samples while S ≈ -1 means that using data at these frequencies may actually be deleterious for distinction. 50 In our feature-selection process we select the frequencies with positive values of silhouette coefficient. Figure 3 shows the IDMAP plot for gliadin in ethanol solutions at concentrations up to 200 mg.kg−1 , above which the data instances are so far located that distinction is trivial, thus not relevant for the analysis. Note the large distance, i.e. dissimilarity, between control solution (prepared with 70% of ethanol and 0 gliadin) and gliadin solutions, even at 0.005 mg.kg−1 . A proper analysis of the data could only be done by removing the data point of pure ethanol, as shown in the zoomed graph. The distinguishing ability with 13 selected frequencies between 15 and 250 Hz is represented by a Silhouette coefficient of 0.74. In subsidiary tests we observed that the sensing performance decreases if data from only one or two modified electrodes are used, as illustrated in IDMAP plots in Figure S5, S = 0.37 for a single PAH/PEDOT:PSS modified electrode and S = 0.64 for the two electrodes PAH/PPy and PAH/PEDOT:PSS. One may infer from the frequency range identified for optimization in the feature selection process, below 250 Hz, that the sensing mechanisms are associated with double-layer effects. 51,52

8


Page 8 of 24

Page 9 of 24

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 3: IDMAP plot for distinguishing gliadin solutions at various concentrations, using frequency selection for the normalized capacitance data for the 4 electrodes.

The analysis of impedance (capacitance) data with 3 LbL modified electrodes and a bare electrode indicates that distinguishing gluten-free from gluten-containing foodstuff is relatively straightforward, as shown in Figure 4. The best performance with S = 0.97 was obtained with frequency selection, as explained in Figure S6 in the supplementary material.

9



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Figure 4: IDMAP plot for gluten-free (green colored region) and gluten-containing foods (red colored region) with frequency selection for the normalized capacitance data of 4 electrodes.

A very stringent test for the e-tongue in real applications is to verify whether it can distinguish between a gliadin-free food sample and a similar sample deliberately contaminated with gliadin. After an optimization procedure, we found that the IDMAP plot for capacitance data with two of the electrodes (PAH/PPy and PAH/PEDOT:PSS) yielded the best distinguishing ability, with S = 0.86. When all the electrodes and all frequencies were used, i.e. without feature selection, S = 0.81. Nevertheless, Figure 5 indicates a very feeble distinction between samples A (Cremogema), AC (Contaminated Cremogema), B (Ninho) and BC (Contaminated Ninho), in contrast to C (Molico) and D (Sake) where the contaminated samples are placed far apart.

10


Page 10 of 24

Page 11 of 24

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 5: IDMAP plot for gluten-free, gluten-containing and spiked gliadin-free foods with frequency selection for the normalized capacitance data obtained with PAH/PPy and PAH/PEDOT:PSS electrodes.

It is worth noting that the distinction of samples in the IDMAP plots in Figures 3 through 5 did not involve any machine learning strategy. That is to say, the original capacitance data were merely plotted with a dimension reduction technique (IDMAP), with no prior labels for the samples. The distinction ability of gliadin in Figure 3 may be compared to results in the literature for diversified sensors. Table 1 shows that the performance of the microfluidic e-tongue introduced here is competitive with most sensors reported in the literature. In fact, only the sensor containing DNA and aptamers had a lower limit of detection. We should nevertheless emphasize that the value quoted for our work is not truly a limit of detection because it was not determined from an analytical curve. This is why we put an asterisk (*) in the table. This value was inferred from the analysis of Figure 3, in which it is clear that detecting gliadin down to 0.005 mg.kg−1 is indeed possible. In principle, data analysis with methods such as IDMAP could lead to analytical curves but only after applying regression analysis methods, which was out of the scope of the present paper.

11



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 24

Table 1: Summary of data from the literature on gliadin detection. Technique Fluorescence Chromatography Differential Pulse Voltammetry Differential Pulse Voltammetry Chronoamperometry ELISA kit test Potentiometric electronic tongue Electronic tongue

3.2

Methodology Polyclonal Immunomagnetic beads AuNp conjugated Polyclonal Immunosensor Polyclonal Immunosensor Pencil Graphite Electrode DNA sequence, Aptamer Gli4 Poly- and monoclonal antibodies Polymeric Membranes Layer-by-layer films

Limit of detection (mg.kg−1 ) 0.6 0.2 0.008 7.11 0.0005 - 0.005 0.24 - 6 1-2 0.005

References [57] [12] [18] [19] [8,20,21] [58] [5] This work*

UV-Vis Spectroscopy

To ensure with an independent method that gliadin was indeed present in the samples analyzed by impedance spectroscopy, a control experiment using UV-Vis absorption spectroscopy was carried out, since gliadin in ethanol solutions displays an absorption band at 280 nm. 5 Figure S7 shows that this band is distinguishable (for the set of samples analyzed) above 227 mg.kg−1 of gliadin, which is approximately the threshold value for food to be considered to contain gluten. Three types of food samples were also analyzed: i) gluten-free foods: sake (alcoholic drink), Cremogena (brand of corn flour), Ninho and Molico (brands of powder milk), ii) gluten-containing foods: toast, salt biscuit, Neston and beer, iii) glutenfree food contaminated with 200 mg.kg−1 of gliadin. Since a large number of spectra were obtained and due to the difficulty in analyzing the information acquired, statistical analysis was carried out using the software PEx-Sensors. 36,53 The IDMAP (Interactive Document Map) technique was used with normalized UV-Vis absorbance data in the wavelength range of 240 - 350 nm. The IDMAP plot in Figure 6 shows that the samples containing gliadin are grouped together, being distinguishable from the gliadin-free samples and from gliadin-free samples contaminated with gliadin at a concentration of 200 mg.kg−1 .

3.3

PM-IRRAS Spectroscopy

As in other types of sensors, we observed that the changes in the impedance data upon exposure to the analyte (gliadin) in different types of samples are due to the adsorption of analyte molecules onto the film surface. This hypothesis was tested here using the PM12


Page 13 of 24

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 6: IDMAP plot of UV-Vis spectral data for foodstuff. Three groups of samples emerge from the analysis: containing gliadin (E = cookie, F = beer, G = Neston, H = toast, blue colored); gliadin-free food (A = Cremogena, B = Ninho, C = Molico, D = sake, green colored) and gliadin-free samples contaminated with 200 mg.kg−1 of gliadin (red colored).

IRRAS technique. The PAH/CuTsPc, PAH/PEDOT:PSS and PAH/PPy films were immersed in a gliadin solution (200 mg.kg−1 ) for 20 min and then dried, analyzed and washed (10 min with ethanol 70%). The washing test in this case is important to check whether the device is reusable, in case gliadin adsorption is reversible. Figures S8, S9 and S10 in Supplementary Information show the PM-IRRAS spectra for 5-bilayer LbL films of PAH/CuTsPc, PAH/PEDOT:PSS and PAH/PPy, while the band assignments are given in Tables S1, S2 and S3. Figure 7 shows that the bands (listed in Table 2) assigned to gliadin adsorbed onto the CuTsPc LbL film remain even after washing the sensing unit, thus indicating an irreversible gliadin adsorption. Similar results with irreversible adsorption were observed for the other electrodes, as illustrated in Figures S11 and S12. The irreversible incorporation of gliadin in the LbL films implied the disposal of the sensing units after measurements to avoid cross-contamination. This limitation in terms of cost can be mitigated by exploiting paper-based devices where the sensing units will be very cheap.

13



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(a)

Page 14 of 24

(b)

Figure 7: (A) PM-IRRAS spectrum of a 5-bilayer CuTsPc LbL film, following adsorption of gliadin and (B) after washing. Table 2: Vibration band assignment of gliadin onto the film after washing. 54–56 Region (cm−1 ) 1547 1600 -1609 1606 -1611 1617 1623 1627 1632 1636 1643 1647 1655 1668 1671 1692 -1693

4

Functional group/vibration Amide I NH2+ NH2+ β-sheet β-sheet Random coil α helix β-turn β-sheet

Conclusions

We successfully fabricated a microfluidic e-tongue device capable of determining the presence of gliadin in ethanol solutions containing foodstuff. The LbL film deposition onto the microchannels of the e-tongue was confirmed by PM-IRRAS measurements, which were particularly useful to demonstrate that the sensing units are not affected by ethanol solutions, being therefore robust for gliadin detection in real situations. To ensure practical applications of the e-tongue system stringent tests were carried out, where samples of gliadin in concentrations down to 0.005 mg.kg−1 and gluten-free foodstuff deliberately contaminated were successfully discriminated by the microfluidic system. The high discrimination power 14


Page 15 of 24

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


performance attained could only be achieved by processing the capacitance data with multidimensional projection techniques and feature selection procedures. The non-linear IDMAP technique provided excellent results, with clear distinction between gluten-free from glutencontaining foodstuff, which is the primary target for developing portable, cost-effective devices for gluten detection in real situations. In summary, the approach used here with a microfluidic e-tongue is promising to generate a low-cost, easy-to-use device sufficiently robust to be handled by regular users, with no need of trained personnel.

Acknowledgement This work was supported by FAPESP (2013/14262-7), CNPq, CAPES, MCTI-SisNano (Brazil). F.M.S. thanks FAPESP Grant 2012/15543-7. A.R.Jr. thanks FAPESP (2014/036917) for financial support. D.S.C. thanks FAPESP (2014/16789-5) and Embrapa - Rede Agronano for support.

Supporting Information Available Complementary characterization of LbL films and feature selection method. This material is available free of charge via the Internet at http://pubs.acs.org/.

References (1) Kelly, C. P.; Bai, J. C.; Liu E.; Leffler, D. A. Advances in Diagnosis and Management of Celiac Disease Gastroenterology 2015, 148, 1175-1186. (2) Green, P. H. R.; Jabri, B. Celiac Disease Annu. Rev. Med. 2006, 57, 207-221. (3) Jenkins, H. R. Coeliac Disease Current Paediatrics 1997, 7, 203-206.

15



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(4) Rostom, A.; Murray, J. A.; Kagnoff, M. F. American Gastroenterological Association (AGA) Institute Technical Review on the Diagnosis and Management of Celiac Disease Gastroenterology 2006, 131, 1981-2002. (5) Peres, A. M.; Dias, L. G.; Veloso, A. C. A.; Meirinho, S. G.; Morais, J. S.; Machado, A. A. S. C. An Electronic Tongue for Gliadins Semi-Quantitative Detection in Foodstuffs Talanta 2011, 83, 857-864. (6) Diaz-Amigo, C.; Popping, B. Accuracy of ELISA Detection Methods for Gluten and Reference Materials: A Realistic Assessment J. Agric. Food Chem. 2013, 61, 5681-5688. (7) Pinto, A.; Polo, P. N.; Henry, O.; Redondo, M. C. B.; Svobodova, M.; O'Sullivan, C. K. Label-Free Detection of Gliadin Food Allergen Mediated by Real-Time Apta-PCR Anal. Bioanal. Chem. 2014, 406, 515-524. (8) Mart´ın-Fernández, B.; Miranda-Ordieres, A. J.; Lobo-Castan, M. J.; Frutos-Cabanillas, ´ G.; de-los-Santos-Alvarez, N.; López-Ruiz B. Strongly Structured DNA Sequences as Targets for Genosensing: Sensing Phase Design and Coupling to PCR Amplification for a Highly Specific 33-mer Gliadin DNA Fragment Biosens. Bioelectron. 2014, 60, 244-251. (9) Hernando, A.; Valdes, I.; Mndez, E. New Strategy for the Determination of Gliadins in Maize- or Rice-Based Foods Matrix-Assisted Laser Desorption/ionization Time-of-Flight Mass Spectrometry: Fractionation of Gliadins from Maize or Rice Prolamins by Acidic Treatment J. Mass Spectrom. 2003, 38, 862-871. (10) Nicolas, Y.; Martinant, J.-P.; Denery-Papini, S.; Popineau, Y. Analysis of Wheat Storage Proteins by Exhaustive Sequential Extraction Followed by RP-HPLC and Nitrogen Determination J. Sci. Food Agric. 1998, 77, 96-102. (11) Wieser, H.; Antes, S.; Seilmeier, W. Quantitative Determination of Gluten Protein Types in Wheat Flour by Reversed-Phase High-Performance Liquid Chromatography Cereal Chem. 1998, 75, 644-650. 16


Page 16 of 24

Page 17 of 24

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(12) Molina-Delgado, M. A.; Aguilar-Caballos, M. P.; Gómez-Hens A. Usefulness of Gold Nanoparticles as Labels for the Determination of Gliadins by Immunoaffinity Chromatography with Light Scattering Detection Talanta 2011, 85, 2391-2396. (13) Ferranti, P.; Mamone, G.; Picariello, G.; Addeo, F. Mass Spectrometry Analysis of Gliadins in Celiac Disease J. Mass Spectrom. 2007, 42, 1531-1548. (14) Valdes, I.; Garcia, E.; Llorente, M.; Mendez, E. Innovative Approach to Low-Level Gluten Determination in Foods Using a Novel Sandwich Enzyme-Linked Immunosorbent Assay Protocol Eur. J. Gastroenterol. Hepatol. 2003, 15, 465-474. (15) Veloso, A. C. A.; Dias, L. G.; Rodrigues, L. R.; Peres. A. M. Gliadins in Foods and the Electronic Tongue in Electronic Noses and Tongues in Food Science Elsevier Science Publishing Co Inc:San Diego, 2016. (16) Nassef, H. M.; Bermudo Redondo, M. C.; Ciclitira, P. J.; Ellis, H. J.; Fragoso, A.; OSullivan, C. K. Electrochemical Immunosensor for Detection of Celiac Disease Toxic Gliadin in Foodstuff Anal. Chem. 2008, 80, 9265-9271. (17) Chiriacò, M. S.; de Feo, F.; Primiceri, E.; Monteduro, A. G.; de Benedetto, G. E.; Pennetta, A.; Rinaldi, R.; Maruccio, G. Portable Gliadin-Immunochip for Contamination Control on the Food Production Chain Talanta 2015, 142, 57-63. (18) Manfredi, A.; Giannetto, M.; Mattarozzi, M.; Costantini, M.; Mucchino, C.; Careri M. Competitive Immunosensor Based on Gliadin Immobilization on Disposable CarbonNanogold Screen-Printed Electrodes for Rapid Determination of Celiotoxic Prolamins Anal. Bioanal. Chem. 2016, 408, 7289-7298. (19) Eksin, E.; Congur, G.; Erdem A. Electrochemical Assay for Determination of Gluten in Flour Samples Food Chem. 2015, 184, 183-187.

17



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

´ (20) Amaya-González, S.; de-los-Santos-Alvarez, N.; Miranda-Ordieres, A. J.; LoboCasta˜ nón M. J. Aptamer Binding to Celiac Disease-Triggering Hydrophobic Proteins: A Sensitive Gluten Detection Approach Anal. Chem. 2014, 86, 2733-2739. ´ (21) Amaya-González, S.; de-los-Santos-Alvarez, N.; Miranda-Ordieres, A. J.; LoboCasta˜ nón M. J. Sensitive Gluten Determination in Gluten-Free Foods by an Electrochemical Aptamer-Based Assay Anal. Bioanal. Chem. 2015, 407, 6021-6029. (22) Daikuzono, C. M.; Dantas, C. A. R.; Volpati, D.; Constantino, C. J. L.; Piazzetta, M. H. O.; Gobbi, A. L.; Taylor, D. M.; Oliveira, O. N.; Riul, A. Microfluidic Electronic Tongue Sens. Actuators B Chem. 2015, 207, 1129-1135. (23) Nery, E. W.; Kubota, L. T. Integrated, Paper-Based Potentiometric Electronic Tongue for the Analysis of Beer and Wine Anal. Chim. Acta 2016, 918, 60-68. (24) Hajba, L.; Guttman, A. Circulating Tumor-Cell Detection and Capture Using Microfluidic Devices Trends Anal. Chem. 2014, 59, 9-16. (25) Zilberman, Y.; Sonkusale, S. R. Microfluidic Optoelectronic Sensor for Salivary Diagnostics of Stomach Cancer Biosens. Bioelectron. 2015, 67, 465-471. (26) Dittrich, P. S.; Manz, A. Lab-on-a-Chip: Microfluidics in Drug Discovery Nat. Rev. Drug Discov. 2006, 5, 210-218. (27) Bouchard, L.-S.; Burt, S. R.; Anwar, M. S.; Kovtunov, K. V.; Koptyug, I. V.; Pines, A. NMR Imaging of Catalytic Hydrogenation in Microreactors with the Use of ParaHydrogen Science 2008, 319, 442-445. (28) Brivio, M.; Verboom, W.; Reinhoudt, D. N. Miniaturized Continuous Flow Reaction Vessels: Influence on Chemical Reactions Lab Chip 2006, 6, 329-344. (29) Shin, J.; Park, H.; Dang, V. B.; Kim, C.-W.; Kim, S.-J. Elastomeric Microfluidic Valve with Low, Constant Opening Threshold Pressure RSC Adv. 2015, 5, 23239-23245. 18


Page 18 of 24

Page 19 of 24

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(30) Haller, A.; Spittler, A.; Brandhoff, L.; Zirath, H.; Puchberger-Enengl, D.; Keplinger, F.; Vellekoop, M. Microfluidic Vortex Enhancement for on-Chip Sample Preparation Micromachines 2015, 6, 239-251. (31) Kim, G.; Moon, J.-H.; Moh, C.-Y.; Lim, J. A Microfluidic Nano-Biosensor for the Detection of Pathogenic Salmonella Biosens. Bioelectron. 2015, 67, 243-247. (32) Xu, L.; Srinivasakannan, C.; Peng, J.; Yan, M.; Zhang, D.; Zhang, L. Microfluidic Reactor Synthesis and Photocatalytic Behavior of Cu@Cu2O Nanocomposite Appl. Surf. Sci. 2015, 331, 449-454. (33) Milani, A.; Statham, P. J.; Mowlem, M. C.; Connelly, D. P. Development and Application of a Microfluidic in-Situ Analyzer for Dissolved Fe and Mn in Natural Waters Talanta 2015, 136, 15-22. (34) Hossein-Babaei, F.; Nemati, K. A Concept of Microfluidic Electronic Tongue Microfluid. Nanofluidics 2012, 13, 331-344. (35) Ariga, K.; Yamauchi, Y.; Rydzek, G.; Ji, Q.; Yonamine, Y.; Wu, K. C.-W.; Hill, J. P. Layer-by-Layer Nanoarchitectonics: Invention, Innovation, and Evolution Chem. Lett. 2014, 43, 36-68. (36) Paulovich, F. V.; Moraes, M. L.; Maki, R. M.; Ferreira, M.; Oliveira Jr., O. N.; de Oliveira, M. C. F. Information Visualization Techniques for Sensing and Biosensing Analyst 2011, 136, 1344-1350. (37) de Oliveira, M. C. F.; Oliveira Jr, O. N.; Pavinatto, F. J.; Constantino, C. J. L.; Paulovich, F. V. Information Visualization to Enhance Sensitivity and Selectivity in Biosensing Biointerphases 2012, 7, 1-15. (38) Rodrigues Jr, J. F.; Paulovich, F. V.; de Oliveira, M. C. F.; Oliveira Jr, O. N. On the

19



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 20 of 24

Convergence of Nanotechnology and Big Data Analysis for Computer-Aided Diagnosis Nanomedicine 2016, 11, 959-982. (39) Zucolotto, V.; Ferreira, M.; Cordeiro, M. R.; Constantino, C. J. L.; Balogh, D. T.; Zanatta, A. R.; Moreira, W. C.; Oliveira, O. N. Unusual Interactions Binding Iron Tetrasulfonated Phthalocyanine and Poly(allylamine Hydrochloride) in Layer-by-Layer Films J. Phys. Chem. B 2003, 107, 3733-3737. (40) Smith, R. R.;

Smith, A. P.;

Stricker, J. T.;

Taylor, B. E.;

Durstock,

M. F. Layer-by-Layer Assembly of poly(3,4-Ethylenedioxythiophene):

Poly(3,4-

Ethylenedioxythiophene): Poly(styrenesulfonate) Macromol. 2006, 39, 6071-6074. (41) Zheng, S.; Tao, C.; He, Q.; Zhu, H.; Li, J. Self-Assembly and Characterization of Polypyrrole and Polyallylamine Multilayer Films and Hollow Shells Chem. Mater. 2004, 16, 3677-3681. (42) Kim, H.-J.; Lee, K.; Kumar, S.; Kim, J. Dynamic Sequential Layer-by-Layer Deposition Method for Fast and Region-Selective Multilayer Thin Film Fabrication Langmuir 2005, 21, 8532-8538. (43) Oliveira Jr, O. N.; Neves, T. T. A. T.; Paulovich, F. V.; de Oliveira, M. C. F. Where Chemical Sensors May Assist in Clinical Diagnosis Exploring Big Data Chem. Lett. 2014, 43, 1672-1679. (44) Paulovich, F. V.; Maki, R. M.; de Oliveira, M. C. F.; Colhone, M. C.; Santos, F. R.; Migliaccio, V.; Ciancaglini, P.; Perez, K. R.; Stabeli, R. G.; Perinoto, A. C.; Oliveira Jr, O.N.; Zucolotto, V.Using Multidimensional Projection Techniques for Reaching a High Distinguishing Ability in Biosensing Anal. Bioanal. Chem. 2011, 400, 1153-1159. (45) Sammon, J. W. A Nonlinear Mapping for Data Structure Analysis IEEE Trans. Comput. 1969, 18, 401-409.

20


Page 21 of 24

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(46) Minghim, R.; Paulovich, F. V.; Lopes, A. A. Content-Based Text Mapping Using MultiDimensional Projections for Exploration of Document Collections SPIE 2006. (47) Inselberg, A.; Dimsdale, B. A Tool for Visualizing Multi-Dimensional Geometry Proc. 1st conference on Visualization ’90 1990, 361-375. (48) Heinrich, J; Weiskopf, D. State of the Art of Parallel Coordinates Eurographics 2013, 95-116. (49) Tan, P.-N.; Steinbach, M.; Kumar, V. Introduction to Data Mining Addison-Wesley Longman Publishing Co.: Boston, 2005. (50) Soares, A. C.; Soares, J. C.; Shimizu, F. M.; Melendez, M. E.; Carvalho, A. L.; Oliveira, O. N. Controlled Film Architectures to Detect a Biomarker for Pancreatic Cancer Using Impedance Spectroscopy ACS Appl. Mater. Interfaces 2015, 7, 25930-25937. (51) Soares, J. C.; Shimizu, F. M.; Soares, A. C.; Caseli, L.; Ferreira, J.; Oliveira, O. N. Supramolecular Control in Nanostructured Film Architectures for Detecting Breast Cancer ACS Appl. Mater. Interfaces 2015, 7, 11833-11841. (52) Taylor, D. M.; MacDonald, A. G. AC Admittance of the Metal/Insulator/Electrolyte Interface J. Phys. Appl. Phys. 1987, 20, 1277-1283. (53) Moraes, M. L.; Petri, L.; Oliveira, V.; Olivati, C. A.; de Oliveira, M. C. F.; Paulovich, F. V.; Oliveira, O. N.; Ferreira, M. Detection of Glucose and Triglycerides Using Information Visualization Methods to Process Impedance Spectroscopy Data Sens. Actuators B Chem. 2012, 166-167, 231-238. (54) Popineau, Y.; Bonenfant, S.; Cornec, M.; Pezolet, M. A Study by Infrared Spectroscopy of the Conformations of Gluten Proteins Differing in Their Gliadin and Glutenin Compositions J. Cereal Sci. 1994, 20, 15-22.

21



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(55) Kretschmer, C. B. Spectroscopy and Optical Rotatory Dispersion of Zein, Wheat Gluten and Gliadin J. Phys. Chem. 1957, 61, 1627-1631. (56) Mangavel, C.; Barbot, J.; Popineau, Y.; Gugen, J. Evolution of Wheat Gliadins Conformation During Film Formation: A Fourier Transform Infrared Study J. Agric. Food Chem. 2001, 49, 867-872. (57) Chu, P.-T.; Wen, H.-W. Sensitive Detection and Quantification of Gliadin Contamination in Gluten-Free Food with Immunomagnetic Beads Based Liposomal Fluorescence Immunoassay Anal. Chim. Acta., 2013, 787, 246-253. (58) Panda, R.; Zoerb, H. F.; Cho, C. Y.; Jackson, L. S.; Garber, E. A. E. Detection and Quantification of Gluten during the Brewing and Fermentation of Beer Using AntibodyBased Technologies J. Food Prot. 2015, 78, 1167-1177.

22


Page 22 of 24

Page 23 of 24

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Graphical TOC Entry

23



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 ACS Paragon Plus Environment

Page 24 of 24

Information Visualization and Feature Selection ... - ACS Publications

Recommend Documents