Biosensors and Bioelectronics 196 (2022) 113737 Contents lists available at ScienceDirect Biosensors and Bioelectronics journal homepage: www.elsevier.com/locate/bios A plug, print & play inkjet printing and impedance-based biosensing technology operating through a smartphone for clinical diagnostics Giulio Rosati a, Massimo Urban a, b, Lei Zhao a, Qiuyue Yang a, Cecilia de Carvalho Castro e Silva a, c, Stefano Bonaldo d, Claudio Parolo a, Emily P. Nguyen a, Gabriel Ortega e, Paolo Fornasiero b, Alessandro Paccagnella d, Arben Merkoçi a, f, * a Catalan Institute of Nanoscience and Nanotechnology (ICN2), Edifici ICN2, Campus UAB, 08193, Bellaterra, Barcelona, Spain b University of Trieste, Department of Chemical and Pharmaceutical Sciences, Via Giorgieri 1, Italy c MackGraphe-Mackenzie Institute for Research in Graphene and Nanotechnologies, Mackenzie Presbyterian University, Consolação Street 930, 01302-907, São Paulo, Brazil d Department of Information Engineering, University of Padova, Via G. Gradenigo 6b, 35131, Padova, Italy e Department of Chemistry and Biochemistry, And Center for Bioengineering. University of California Santa Barbara. Santa Barbara CA, 93117, USA f Catalan Institution for Research and Advanced Studies (ICREA), Passeig de Lluís Companys, 23, 08010, Barcelona, Spain A R T I C L E I N F O A B S T R A C T Keywords: Simplicity is one of the key feature for the spread of any successful technological product. Here, a method for Nanofunctional inks rapid and low-cost fabrication of electrochemical biosensors is presented. This “plug, print & play” method in­ Inkjet-printed sensors volves inkjet-printing even in an office-like environment, without the need of highly specialized expertise or Aptasensors equipment, guaranteeing an ultra-fast idea to (scaled) prototype production time. The printed biosensors can be Flexible electronics connected to a smartphone through its audio input for their impedance readout, demonstrating the validity of the Impedimetric biosensors Smartphone readout system for point-of-care biosensing. Proper electrodes layout guarantees high sensitivity and is validated by finite element simulations. The introduction of a passivation method (wax printing) allowed to complete the devices fabrication process, increasing their sensitivity. Indeed, the wax allowed reducing the interference related to the parasitic currents flowing through the permeable coating of the employed substrates, which was used for the chemical sintering, thus avoiding the common thermal treatment after printing. As a case study, we used the devices to develop an electrochemical aptamer-based sensor for the rapid detection of neutrophil gelatinase- associated lipocalin (NGAL) in urine – a clinically important marker of acute kidney injury. The aptasensor platform is capable of detecting clinically relevant concentrations of NGAL with a simple and rapid smartphone readout. The developed technology may be extended in the future to continuous monitoring, taking advantage of its flexibility to integrate it in tubes, or to other diagnostic applications where cost/efficiency and rapidity of the research, development and implementation of point of care devices is a must. 1. Introduction the purity of our atmosphere, water and soil (Jaiswal and Anand, 2021), are on the rise. However, as technology becomes increasingly complex, Biosensors first entered the market several decades ago and today - at simple and customizable fabrication of ultra-low-cost sensors together the interface between research and the market - there is a plethora of with user friendly testing in any environment is highly sought after. new generation biosensors with the potential to profoundly influence Currently, the production of biosensors is generally expensive, as spe­ medical decision-making. Applications of point-of-care (POC) di­ cifically designed facilities (e.g., clean rooms) are required. Moreover, agnostics (C. Parolo et al., 2020), such as wearables for non-invasive the need for trained personnel and specialized materials, as well as the health and lifestyle monitoring (Christodouleas et al., 2018), and the time taken for production can further drive up costs. Thus, to enable Internet of Things (IoT) network of environmental sensors, measuring widespread integration and accessibility to advanced POC technologies, * Corresponding author. Nanobioelectronics & Biosensors Group, Institut Català de Nanociància i Nanotecnologia (ICN2), CSIC and The Barcelona Institute of Science and Technology (BIST), Campus UAB, 08193, Bellaterra, Barcelona, Spain. E-mail address: arben.merkoci@icn2.cat (A. Merkoçi). https://doi.org/10.1016/j.bios.2021.113737 Received 14 September 2021; Accepted 25 October 2021 Available online 29 October 2021 0956-5663/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). G. Rosati et al. Biosensors and Bioelectronics 196 (2022) 113737 the fabrication process should aim to reduce costs to manufactures and, prove its validity for biosensing purposes, we functionalized the fabri­ ultimately, everyday users. cated sensors with aptamers specific for NGAL (Azzalini et al., 2016; Inkjet printing and smartphone sensing are promising candidates to Wang et al., 2017), an early biomarker correlated to important syn­ fulfil these requirements, with the first allowing mass fabrication at low dromes such as acute kidney injury (AKI) and acute to chronic liver cost, and the second enabling easy POC and even at-home applications failure (ACLF, Kara et al., 2010; Mishra et al., 2004). Our inkjet sensors (Da Silva et al., 2014; Hernández-Neuta et al., 2019; Manoto et al., 2020; showed excellent performances in the clinical range (3–30 nM), with a Medina-Sánchez et al., 2014; Quesada-González and Merkoçi, 2017). sensitivity of 31.49 Ohm/nM, corresponding to a LOD of 3 nM in arti­ Inkjet printing represents as a highly cost-effective, user-friendly and ficial urine. Furthermore, our biosensor proved to be resilient to the versatile method for the fabrication of biosensors. In comparison to interference of Bovine Serum Albumin, which we used to emulate the other printing fabrication methods, such as screen printing, the internal presence of albumins which are one of the most typical class of protein multichannel ink reservoirs can be adapted to print with many and found in human urine (Cheng et al., 2012). Therefore, effective various types of ink (i.e. conductive, dielectric etc.) and is well suited for impedance-based smartphone biosensors can not only be used at the various flexible substrates, such as paper and plastics (i.e. PET). PoC, but also produced by anyone with just office equipment (an inkjet Furthermore, inkjet printing also allows for mass production of bio­ printer, a wax printer, and a laminator). sensors with the capability to automate the printing process (Rosati In the following, we present our journey in the development and et al., 2019a). Despite these excellent properties, inkjet devices fabri­ characterization of our Plug, Print & Play technology (Fig. 1). Starting cation is still bound to a laboratory-based paradigm, preventing its full from the materials selection (inkjet printer model and ink/substrate), we potential as a ubiquitous production method to be exploited. This is then characterized the printed lines morphology and conductivity, mainly due to the treatments needed by the devices after printing. defining the parameters we needed for the electrodes design. We vali­ Thermal or optical procedures are in fact usually required to achieve the dated the design with FEM simulations to assess the uniformity of the desired electrical and mechanical properties of the devices, such as electrical field along the electrodes and guarantee satisfactory perfor­ thermal sintering on nanoparticles-based inks or photonic curing of mances in affinity sensing. Once theoretically validated, we produced insulating and polymeric inks (Cummins and Desmulliez, 2012; the devices and we optimized their fabrication to avoid parasitic effects Wünscher et al., 2014). during the impedance measurements due to the semi-permeable sub­ Inkjet manufacturing is particularly convenient for the fabrication of strate allowing room temperature sintering. The sensitivity of the affinity-based electrochemical label-free biosensors (Sui and Zorman, impedance measurements to the coverage of the electrodes surface by 2020; Tortorich et al., 2018). These are composed by electrodes modi­ biomolecules have been experimentally assessed and modelled by data fied with biological receptors that specifically bind the target molecule fitting to an equivalent electrical circuit (EEC). Finally, we modified the (affinity) if present in the liquid sample. The binding is detected with devices with the NGAL specific aptamers and we performed the electrochemical techniques, such as impedance, that can be performed smartphone-based quantification. directly in the sample solution, without the addition of any further The proposed detection mechanism, simply based on the non-faradic compound or label (label-free). Aptamers, which are short DNA se­ (label-free and redox-free) impedimetric detection of the NGAL binding quences iteratively selected for their specific affinity to the target to the aptamers immobilized on the electrode surface, is usually molecule of interest, are very proficient as bioreceptors due to their high considered prone to unspecific binding and interference in real samples. stability, low-cost, and customizability, which makes them an excellent Despite that, the combination of this detection method with well- choice for this and other kinds of biosensors (Arroyo-Currás et al., 2017, characterized additional methods such as anti-fouling strategies, nano­ 2020; Idili et al., 2019; Claudio Parolo et al., 2020). Electrochemical shearing, and differential measurements, can guarantee very good per­ label-free and redox-free strategies, are typically considered prone to formances for clinical testing, allowing also for the compensation of interference. Impedance aptamer-based affinity systems are no excep­ inter and intra-patient samples variations (McGrath et al., 1995; Sabaté tion. Despite this, the level of interference mainly depends on two fac­ del Río et al., 2019; Vaidyanathan et al., 2014). All the previous tech­ tors: the amount of interfering molecules present in the sample of nologies can be easily exploited adding up additional electrodes in the interest (the so-called matrix effect), and the implementation of design, i.e. asymmetric interdigitated electrodes for the nanoshearing anti-fouling strategies and materials. Based on the expected level of and a second couple of the electrodes functionalized with similar but interference in the sample of interest, different antifouling strategies non-specific aptamers for differential measurements to remove the in­ may be considered. Between them, surface modification with terferences due to the variability of the real samples. Self-Assembling Monolayers (SAM) is probably the most diffused one. Several molecules can serve to this scope (Lin and Li, 2020). Despite 2. Material and methods that, also microfluidic and electrochemical protocols proved to be effective, e.g. continuous flow conditions or nanoshearing (Vaidyana­ Device fabrication and testing: The devices were fabricated following than et al., 2014). three procedures in order to compare the effect of two coating passiv­ In response to the need for a manufacturing technology that allows ation strategies on electrochemical impedance measurements for for convenient, cheap and easily transferable manufacturing of elec­ affinity-based detection of biomolecules. Devices without any coating trochemical biosensors, in this work, we propose a reliable Plug, Print & passivation were fabricated and tested as negative controls, as sche­ Play technology – plug: office-like equipment (no specific skills/labo­ matized in Fig. 2. All the device electrodes and contacts were inkjet- ratory needed), print: ease of design and of production by simply printed with Mitsubishi Paper Mills Silver Nano™ AgNP ink (NBSIJ- printing an image, & play: connection of the device to a smartphone and MU01) on Mitsubishi Paper Mills special substrate (NB-TP-3GU100) by the simple drop-casting of the sample for immediate testing. We use this an EPSON XP15000 printer. The surface was dried immediately after technology for the fabrication of aptamer-based biosensors for both printing thanks to the substrates coating absorbing the ink solvent. We office-like environments and POC testing. Hitherto, this goal has proved waited 72 h keeping the devices at RT before performing the following challenging from a technical/technological point of view due to the need step to ensure proper drying of the solvent. The wax-passivated devices for controlled operations/conditions and skilled personnel. In order to were fabricated by printing a layer of wax (black colour, Xerox Phaser simplify these constraints, we introduced an innovative sinter-free 8850) in the area over and around the electrodes, with a 1 cm wide method to produce electrochemical impedance-based devices. These, square pattern per each device. After the wax printing, we kept the combined with a simple and cost-effective (~$5) circuit connected to device in the oven at 95◦ for 20 s to melt the wax and let it slightly the audio and microphone channels of a smartphone, and a commercial penetrate the coating structure. We verified that this step could also be app, allow for fast impedance measurements at the point of care. To performed with a simple hairdryer without losing reproducibility since 2 G. Rosati et al. Biosensors and Bioelectronics 196 (2022) 113737 Fig. 1. The figure shows the steps in the process of the inkjet devices simulation, fabrication and opti­ mization, and validation. The interdigitated elec­ trodes (IDE) layout was selected thanks to the simulations (1) and inkjet-printed with silver nanoparticles (AgNP) ink on coated substrate for room temperature sintering (2). Two coating passivation strategies have been tested (3) and the devices were then laminated with patterned plastic sheets for contacts insulation (4). Bare, wax and laser etched devices were tested with mercapto­ hexanol (MCH) Self-Assembled Monolayer immo­ bilization to select the most effective passivation method (5). Wax passivated devices were func­ tionalized with NGAL Aptamers (6) and used to detect NGAL in Phosphate Buffered Saline (PBS) (7) and artificial urine (8). Smartphone measurements were performed on the same devices for the cali­ bration of NGAL in PBS. Fig. 2. The figure shows examples of the FEM simulation results for the layout described in the text and sketches the steps of the fabrication pro­ cesses which has been tested, with the used equip­ ment and SEM pictures of the resulting devices. a, cross-section sketch and contour plots of the elec­ tric field along the 50 μm wide electrodes at the base (continuous), middle (dashed), and end of the fingers (point-dashed). b, same representations of (a) for the 400 μm wide electrodes. c, Schematic of the fabrication steps of the bare, wax-passivated and laser etched devices. d, Corresponding SEM images of the gap between the printed electrodes and of the electrode/substrate cross-sections for the bare, wax and laser devices (from top to bottom, the images all have the same scale). its goal was just to obtain a strong attachment between the wax and the come in contact with the solution. As performed in a previous study coating. After this, we performed the deposition of a second wax layer to (Rosati et al., 2019b) the lamination sheet can also be patterned by a ensure a complete coverage of the area and we kept the devices in the low-cost digital blade cutter (e.g. CraftROBO CC330), which can be oven for 5 min (this step can also be performed with an hairdryer). safely used in any office-like environment (Fig. S5). The laser-etched devices were fabricated engraving the inkjet- The fabricated devices were tested in 10 mM PBS (Merck-Sigma printed devices with a Rayjet 50™ CO2 laser (power: 17%, speed: Aldrich) by drop-casting 50 μl of the solution on the active window and 100%, number of passages: 3) using the same PDF file employed for the connecting them to a Metrohm AutoLab PGSTAT12 (FRA2) potentiostat wax printing. The laser vaporized the coating in the engraved area but by a custom adapter using a FPC Molex connector to ensure proper not under the silver electrodes, which protected the coating reflecting electrical contact. The Electrochemical Impedance Spectroscopy (EIS) the beam. measurements were performed in triplicate (if not otherwise specified) All the devices (inkjet-printed, wax-passivated, and laser-etched) between 0.1 Hz and 100 kHz, 5 pt/dec, with 10 mVpp AC and 0 V DC. were then laminated with a laser-patterned 100 μm thick lamination The open circuit characterizations were performed by repeating the EIS sheet (Fellowes A4 80 μm gloss) at the temperature recommended by the measurements with the described method three times: before the solu­ laminator manual for sheets of this thickness (temperature selector in tion dropcasting (Open Circuit dry – OC dry), after the solution drop- orange position), exposing the device contacts and the electrodes in a 2 casting (PBS), and immediately after the solution removal with nitro­ mm wide and 4 mm high window, through which the electrodes could gen (Open Circuit wet – OC wet). 3 G. Rosati et al. Biosensors and Bioelectronics 196 (2022) 113737 MCH functionalization, testing and EIS data fitting: All the fabricated drop volume of about 5 pl, the availability of empty cartridges on the device types were modified by incubating them in 50 μl mercaptohex­ market, the relatively low cost of the printhead, and the recent date of anol (Merck-Sigma Aldrich) dilutions in 10 mM PBS for 1 h at RT. The entrance in the market to ensure the longest availability of the printer on dilutions ranged from 10 nM to 10 mM, one per decade. After the in­ the market. Table S1 lists the settings used for the printing with Epson cubation period, the devices were washed with milliQ and dried with XP15000. nitrogen. Then, the devices were tested by EIS with the same settings Regarding the inks and the substrates used for the printing, the described in the previous paragraph. On the basis of the impedance data choice of conductive inks is fundamental to define the electronic and we got from the PBS characterization and considering both the nature of electrochemical properties of the printed electrodes. Following the conduction through the solution and through the wet coating, we principle of an easy and user-friendly fabrication platform, we checked defined an electrical equivalent model (Fig. 4b). The EIS data collected the commercially available inks based on different metal nanoparticles, from the MCH-modified devices were fitted with the defined model with i.e. mainly gold and silver. However, only the widespread silver- ZView2™ (Scribner’s and associates). nanoparticles-based inks (AgNPs) were found to meet the re­ Aptamer immobilization: The NGAL aptamers were designed and quirements of a sustainable price-over-quantity ratio, sufficient ease of selected by the Plaxco group and were composed of a thiol-modified management, and long-term stability at room temperature (RT). The end. They were reduced in Tris(2-carboxyethyl)phosphine hydrochlo­ main producers of inkjet AgNP-based inks are Mitsubishi Paper Mills and ride (TCEP) 10 mM for 1 h in the dark and then diluted in 10 mM PBS Novacentrix, which also produce the semi-absorbing plastic substrates, with 1 mM MgCl2 to 1000 nM. From this concentration, we prepared typically used with these inks for the high-quality printing of electronics. two other dilutions, respectively of 100 and 10 nM. The wax-passivated Semi-absorbing plastic substrates guarantee the chemical sintering of devices, selected for the following tests thanks to the MCH tests and data the AgNPs at RT, where a reagent in the absorbing coating of the sub­ fitting, were incubated for 2 h drop-casting 50 μl of the three aptamer strates remove the outer protective shell of the nanoparticles and dras­ dilutions. After the incubation, the devices were washed with milliQ and tically improve their conductivity. Inks were tested in combination with dried with nitrogen. Before testing the devices, we waited for 3 h, several substrates and their printing performance was compared in keeping them at RT to ensure a complete drying of any liquid residue terms of minimum conductive width of the electrode lines (or fingers) eventually infiltrated under the wax layer. The functionalized devices and their resistance with respect to the printing direction. The study, were tested by EIS in 10 mM PBS (in triplicate) with bare wax-passivated based on a previous work (Rosati et al., 2019b), allowed us to select the electrodes tested as negative control (Fig. 5a). Mitsubishi ink and substrate for the following steps. NGAL detection: The aptamer-functionalized devices were separately tested by EIS in 5 NGAL dilutions ranging from 0.3 to 3000 nM and in 3.2. Printed lines morphology and conductivity PBS. Each dilution was tested on three devices for each aptamer con­ centration. The 10 nM and 1000 nM aptamer concentrations did not In order to evaluate the printing results of the Epson printer, we show a clear response to the NGAL dilutions, while the 100 nM aptamer performed an analytical comparison with a standard Dimatix 2800 showed a good response at frequencies in the range between 1 and 100 printer. The lines printed with the Mitsubishi AgNP ink on the respective Hz. Therefore, 100 nM aptamer concentration and 100 Hz were selected substrate were characterized by optical microscopy and interferometric for the device calibration. measurements (Fig. S1). Despite the Dimatix’s lines appear more uni­ The calibration was performed by connecting 6 devices to the form, the Epson reached similar resolution and conductivities (Figs. S1 AutoLab multiplexer module (MUX) and setting up a customized NOVA and S2), and even a much faster resistance stabilization time. The function for the continuous measurement of impedance at 100 Hz. We interferometric analysis clearly evidenced the single jetted drops in 3D, dropcasted 50 μl of 10 mM PBS on all the devices and we started the with an average drop diameter of about 20–30 μm and a thickness of impedance measurements. After 10 min we replaced 25 μl of the drop on between 125 and 200 nm (Fig. S1d). Furthermore, the thickness of a the devices with the first NGAL dilution on three of the six devices, and printed line significantly varied through its width. In fact, the central with PBS on the other three. We repeated the procedure for each NGAL regions of lines had a thickness of ~800 nm, while their lateral borders dilution, with a 10 min interval between each substitution. had a thickness of ~150 nm. This difference is related to the different The staircase calibration data was analyzed by subtracting the data number of overlapped layers of jetted drops, as the printer overlaps 4 of the PBS sensors from the NGAL ones to compensate the eventual drift different layers at a distance of more than 150 μm from the line border. related to the long measurements (almost 2 h). The added NGAL con­ This 3D profile of the lines implies that lines with widths lower than centrations are reported in Fig. 6a. However, the effective NGAL con­ 300 μm have a significantly different thickness and thus conductance centration on the sensor was generally slightly higher than half the with respect to larger ones. In turn, this explains why the resistance of 2 added concentration due to the half drop replacement protocol. The cm lines of different widths did not follow Ohm’s first law for lines under kinetics related to four NGAL concentrations between 0.15 and 25.10 the 300 μm (Fig. S2). In these cases, the terrace-like 3D structure of the nM were fitted to a mono-exponential equation. jetted drops along the line section strongly influences the lines’ re­ The smartphone NGAL measurements were performed with the same sistances. Moreover, optical characterizations of the lines (Fig. S1) evi­ staircase calibration protocol but taking full 50 Hz–20 kHz impedance denced several defects on the overlapping of drops along the line, spectra from three Android 8.1.0 smartphones (Huawei Mate 20 lite) creating bottlenecks between two consecutive drops, which inhibit the connected to three identical reader circuits. Every EIS measurement was electric conduction. It is important to underline that the precision of the taken after 10 min of incubation with the respective NGAL dilution. The overlapping of the drops strongly depends on the printing direction. As calibration curve was obtained plotting the impedance module at 150 studied in one of our previous works (Rosati et al., 2019a), the highest Hz (Fig. 6c). The data was fitted using the Hill equation. Finally, the quality is obtained when the lines are printed along the same direction of same measurements were repeated in artificial urine (Surine Negative the printhead movement, called "inline"-print, while the worst quality is Urine Control, Sigma). perpendicular to it, called "outline"-print (Fig. S1a). All the electrodes fabricated in this study were printed "inline", in 3. Results and discussion order to obtain the highest quality in terms of precision and reproduc­ ibility. The measured resistivity for lines of width greater than 300 μm 3.1. Materials selection and consumer printer setup resulted in 0.685 μΩm, corresponding to 2.32% of the bulk silver con­ ductivity. Only the contacts were printed “outline”, as their extended The printer selected for the new developed platform was the Epson area with large width neutralizes the weight of defects on the line XP 15000. The main advantages of this printer model are the minimum resistance. 4 G. Rosati et al. Biosensors and Bioelectronics 196 (2022) 113737 Another important element to consider is the difference between the impedance-based electrochemical detection performance. The focus of layout width and the real width of the lines (Fig. S2a). The real printed this process is the removal or passivation of the absorbing coating on the lines were generally 1.34 times wider than the nominal layout value substrate. Indeed, the coating guarantees a rapid and precise printing with an offset of 85 μm, as obtained from the linear regression. However, with an RT chemical sintering and also absorbs any liquid solution in this law was not respected only for nominal dimensions lower than 100 contact with it. The solution penetrates and slowly diffuses around the μm and a minimum printable width of 50 μm. Based on these analysis, electrodes causing instabilities and undesired short-circuiting effects. we defined equivalent profiles for 400 μm nominal width lines and for Even if the “short-circuit” resistance has values in the order of 100 the thinnest printable lines. kOhm, this prevents the devices from optimally performing affinity- based detection of the analyte. 3.3. Electrodes design and simulations As illustrated in Fig. 2, we compared two strategies for the coating passivation/removal with the bare devices (with the coating intact and For our device, we chose a simple 2 interdigitated electrodes design. unmodified). In the first strategy, two layers of wax ink were deposited As a first step, our devices are characterized by two parallel electrodes over the electrodes around the solution interface area and heated up to with a nominal width of 30 μm (real width = 50 μm), a gap of 200 μm penetrate in the coating. While, in the second strategy, the corre­ and a total length of 2 mm, since this was the minimum printable width sponding area was etched by a desktop CO2 laser. for a conductive line. However, we found that, even though thinner The wax devices appears clearly black and the fact that their SEM electrodes are typically associated with higher sensitivity, this scaled images show a similar surface to the bare devices indicates that the wax design may result in a not-uniform electric field along the electrodes. effectively penetrated in the coating, even if in the cross-section image This effect is due to the balance between the lines conductance and the no difference between the coating and wax can be distinguished. This is conduction through the liquid sample over the electrodes, as described due to the similar density and elemental composition of the two mate­ in a previous work [23]. The electric field uniformity is important in the rials. SEM images acquired at higher magnification and lower acceler­ development of electrodes in order to guarantee an equal sensitiveness ation voltage allowed us to slightly distinguish the two (Fig. S4), further of all regions of the electrode surface. Therefore, we designed and proving the permeation of the coating by wax. printed two other electrodes with nominal widths up to 400 μm (real Interestingly, the laser was reflected by the AgNP electrodes and as a width = 600 μm) and the same 200 um gap between the electrodes result, the substrate coating beneath them remained intact (bottom SEM (Fig. S3a). The real gap distance has been kept constant in the two pictures of Fig. 2). layouts introducing an offset proportional to the theoretical width to All the devices (bare, wax and laser) were then encapsulated in a compensate the spreading of the printed lines on the substrate. laser-patterned lamination sheet in order to define the sensor active To further verify the optimization of the electrode design, simula­ window, where the two parallel electrodes were exposed to the testing tions using COMSOL Multiphysics™ were conducted to simulate their solution. The lamination sheet can be also patterned by office-like electrical behaviour and, in particular, observe the uniformity of the cheaper equipment such as digital cutter. There is no need for a laser electric field along the electrodes. Using the morphological and elec­ cutter for this as demonstrated in a previous work (Rosati et al., 2019a). trical characterizations described in paragraph 2.2, we modelled the We carried out 3 studies to verify the most effective coating passiv­ electrode-solution interface. The mesh of the electrodes had the previ­ ation strategy and compare the device performance: 1) we simply ously defined profile and conductivity on a dielectric substrate. The measured the impedance spectra in 10 mM PBS; 2) we compared the electrodes were surrounded by a 3D rectangle, which simulates the so­ impedance of the dry electrodes with the that of the electrodes dried lution drop over the electrodes. The base dimension of the rectangle right after a measurement in PBS; and 3) we studied the sensor equalled that of the device active window with a width of 2 mm and a impedimetric response to the functionalization with 6-Mercapto-1-hex­ height of 4 mm (Fig. S3b). We set the conductivity of the 3D rectangle to anol (MCH) dilutions ranging from 10 nM to 10 mM. 16 mS/cm, equal to the conductivity of 10 mM PBS, used for the EIS EIS measurements are used to characterize the printed electrodes. measurements. Such measurements are minimally perturbing the electrochemical cell, Simulation of the electric potential along the meshed electrodes thus it is possible to avoid the use of a reference electrode. In this way, allowed us to check for its uniformity when an electrical stimulation of the device can be composed of just 2 identical electrodes. Moreover, we 10 mV DC was applied between the two electrodes ends. This study was did not use any redox mediator on our sensors, allowing direct mea­ fundamental to assess if variations of impedance due to passivation of surement of the sample solutions, without any pre-treatment. In this the electrodes by the binding of the analyte to the functionalized surface configuration, the binding of biomolecules on the electrodes surface is was equally effective along the electrode length. A uniform field would usually detected as an increase of the impedance module at the fre­ mean a smaller impedance signal variation due to the same analyte quencies resulting in the phase peak, i.e. where the impedance spectra is concentration in solution. dominated by the double-layer capacitance of the electrochemical cell. The potential isosurfaces, shown in Fig. S3d, reveals a non-uniform However, electrodes printed on porous substrates could be affected by electrical field along the electrodes for the 50 μm layout. This is even the parasitic conductance of the solution infiltrating under the elec­ more evident in Figs. 2a and S3e, which underlines the non-overlapped trodes and limiting the maximum increase of impedance achievable contour plots in different planes along the electrodes. The width and gap with their top surface coverage with the biomolecules of interest. From between the electrodes is paramount to define the field uniformity. For this, the rationale of the substrate coating passivation strategies is the example, a nominal width of 400 μm with a gap of 200 μm with mesh reduction of these parasitic components. shown in Fig. S3f resulted in a uniform electrical field as can be seen The wax-passivated devices EIS response (Fig. 3a) clearly shows a from both the isosurfaces and contour plots in Fig S3g, 3b and S3h. The higher impedance with respect to the bare electrodes at low frequencies. trade-off between the field uniformity and the reduction of the surface The laser-treated devices surprisingly show lower impedance values in area of the electrodes is fundamental to ensure the highest signal vari­ the same region. In the high-frequency range, where the impedance ation related to the analyte binding on the electrodes (Rosati et al., response is mainly related to the resistance of the solution, the curves are 2014). relatively similar. This indicates that the modification of the devices did not significantly change the electrode surface and interaction with the 3.4. Fabrication of the biosensors and initial electrochemical validation solution. However, the spectra of all the devices show a phase peak around 20 Hz, which is not coherent with the R–C behaviour expected in After choosing the validated 400 μm device layout, we studied and PBS not containing any redox probe. optimized the devices fabrication process in order to obtain the best The EIS tests on the dry electrodes before and after the PBS 5 G. Rosati et al. Biosensors and Bioelectronics 196 (2022) 113737 Fig. 3. The figure represents the EIS responses of the fabricated devices (bare, wax-passivated and laser-engraved) in order to define the most effective coating passivation method. a, Bode diagrams of the devices’ impedance spectra measured in PBS. b, Comparison of the 100 mHz impedance modules of the devices in PBS, immediately before the PBS in­ jection, and immediately after it. c, Schematics illustrating the theorized conduction between the bare, wax-passivated and laser-treated electrodes. Fig. 4. Impedance spectra and modelling of the bare, wax-passivated and laser-etched devices modified with a MCH self-assembled monolayer. a, Impedance modules at 10 Hz recorded for bare, wax-passivated and laser-etched devices. b, Elec­ trical equivalent circuit and schematic of the de­ vices model of conduction. c,d EEC parameters values with respect to the MCH concentration for the bare, wax-passivated and laser-etched devices. The values have been obtained by fitting the average impedance spectra in the three cases, so no standard deviations are reported. The percentage errors obtained for the two elements were smaller than 5% in all the cases. 6 G. Rosati et al. Biosensors and Bioelectronics 196 (2022) 113737 Fig. 5. NGAL aptamer concentration optimization and detection testing. a, Impedance module and phase spectra obtained after the functionalization with 10, 100 and 1000 nM NGAL aptamer (N = 3). The 10 nM case probably resulted with the highest impedance module values because the aptamers may be lying horizontally with a so low packing density and no spacer. b, Average impedance module values with respect to NGAL concentration at 0.1, 1, 10, 100, and 1000 Hz for the aptasensors functionalized with 100 nM NGAL aptamer. c, Sin­ gle frequency plots of the data in b. for the fre­ quencies 1, 10 and 100 Hz (N = 3) compared to measurements in PBS for a better visualization of the impedance values dependence on the NGAL concentration. The error bars represent the stan­ dard deviations in all the plots and in some case are relatively big because of a possible lack of repro­ ducibility between the three different devices used for each concentration. Fig. 6. NGAL calibration on the impedance apta­ sensors. a, Staircase NGAL calibration in PBS at 100 Hz by potentiostat with the injected NGAL concen­ trations in the PBS drop over the aptasensor b, Ki­ netics associated to the effective NGAL concentrations in the drop in the first seconds after the injection and respective fitting with single exponential equation. c, Smartphone-measured NGAL calibration curve at 100 Hz with Langmuir fitting (|Z|max = 373.56, K = 13.92, n = 1.30, Adj. R2 = 0.99477) and 95% confidence and prediction bounds. d, NGAL detection in artificial urine, impedance module variation at 1 Hz with respect to pure artificial urine with linear fitting of the values in the range 3–300 nM (Δ|Z| = 31.49[NGAL]+ 3845, R2 = 0.9998). In this case, due to the change of the buffer properties, the response at 1 Hz resulted more significative with respect to the one at 100 Hz. The top inset shows a magnification of the 0–30 nM range and the bottom inset shows the result of the interference study performed with BSA. measurements (Fig. 3b) gave further tips to understand the real nature of These results evidence what is happening in the three types of tested the events taking place at the electrode interfaces and inside the coating. devices. Fig. 3c illustrates the hypothesized behaviour of the electric As expected, the bare devices showed a difference of 4 orders of field between the electrodes of the three device types. The bare devices magnitude between the 0.1 Hz impedance modules of the dry electrodes coating between the two electrodes immersed in solution is immediately before and after PBS testing. The wax-passivated devices significatively penetrated by the solution, which diffuses under the bottom face of increased the impedance absolute values and reduced the difference electrodes. The conduction through the coating is thus possible and, for between the two conditions to 3 orders of magnitude. While, the laser- low frequencies values, the EIS curve follows the trend of a constant treated devices showed the highest impedance module values and phase element with a behaviour closer to that of a resistance with a completely closed the gap between the dry measurements before and negative phase lower than 30◦ . This reduces the impedance values in after PBS testing, saturating the potentiostat at this frequency (10 that region, dominating the double layer capacitance related one and GOhm). forming a peak in the phase, like in a redox-based system. This 7 G. Rosati et al. Biosensors and Bioelectronics 196 (2022) 113737 behaviour is confirmed by the impedance module values at 0.1 Hz on the functionalization was performed simply by drop-casting the activated dry electrodes after the PBS testing, which are just 1 order of magnitude thiolated aptamers on the electrodes for 2 h. Three different aptamer greater than the ones recorded in PBS. For the laser-treated devices, the concentrations were tested, ranging from 10 nM to 1 μM. impedance of the dried electrodes after PBS testing is so high because the Fig. 5a shows a large variation in EIS responses between the bare coating has been completely etched between the electrodes, thus electrodes and the 10 nM functionalized ones. While the variations be­ nothing connects them when dried. However, the coating is still present tween different concentrations of the aptamers (100 nM vs. 1 μM) are under the electrodes and this does not solve the problem during the PBS significant, they are smaller. As shown in Fig. 5b, we interpreted this measurements, which in fact appears similar or even worse than the bare result as the effect of the aptamers lying in a horizontal configuration, device ones. In wax-passivated devices, the conduction through the due to the lack of the repulsion between the nucleotides negative coating is drastically reduced making it more hydrophobic even if not charges in such a dispersed solution. This configuration produced a completely impermeable. The measurements on the dry electrodes greater coverage of the surface with a consequent decrease of the double clearly show a net improvement in the gap between the impedance layer capacitance and an increase of the related impedance module. modules at 0.1 Hz of the PBS test and of the dried electrodes, reaching Interestingly, in the case of 100 nM and 1 μM aptamers dilutions, the almost 4 orders of magnitude. Even the impedance of the dry electrodes greatest variation between the respective spectra was observed around improved, reaching the values obtained with the laser treatment. This is 100 Hz, instead at lower frequencies as observed for the other dilutions. clearly reflected in the PBS tests, with the phase reaching − 55◦ even if For all the aptamer-modified devices, we performed EIS measure­ still affected by a reduced coating conduction. ments in NGAL dilutions in PBS. The tested NGAL concentrations range was between 0.3 nM and 3000 nM, covering the range of clinical interest 3.5. Impedance sensitivity to surface coverage and electrical modelling (3–30 nM). The highest impedance variations were obtained with 100 nM aptamers (Fig. 5c), showing a maximum variation of more than 1.5 The devices functionalization with mercaptohexanol (MCH) lipidic kOhm at 100 Hz and a range of detection between 0.3 and 30 nM. For self-assembled monolayer allowed defining their sensitivity to surface higher concentrations, i.e., 300 and 3000 nM, the impedance variations coverage and a comparison of how it is influenced by the coating saturated probably because of the full occupation of the available passivation method. The average impedance response of the function­ aptamers on the electrodes surface. This may have also caused in some alized devices showed its highest variations at 10 Hz (Fig. 4a). As ex­ case a partial disruption of the aptamers self-assembled monolayer due pected, the impedance module increases with the MCH concentration, to electrostatic forces, increasing the variability of the results as can be until reaching a plateau in correspondence with the highest concentra­ seen from the higher errorbars. tions. Comparing device impedance, the bare devices showed the lowest values, with a slow increase starting at 1 μM MCH and never really 3.7. Smartphone-based NGAL quantification reaching a saturation value. The laser treated devices showed a similar behaviour but the impedance started from and reached higher values, To prove the feasibility of continuous monitoring and real-time with a sigmoid shaped curve. Conversely, the wax passivated devices detection (Curtis et al., 2019), we performed the calibration on single showed the greatest impedance variation. Furthermore, this variation electrodes injecting increasing NGAL concentrations over time and was achieved at a MCH concentration between 104 and 105 nM, at least monitoring the impedance variation at 100 Hz. As previously proved by one order of magnitude smaller with respect to the other two device Parolo et al. (Claudio Parolo et al., 2020), these aptamers allow the types (between 105 and 106 nM). This indicates their highest sensitivity monitoring of NGAL for hours if placed in a catheter. This is possible to surface coverage, thus offering the best detection performance for thanks to the aptamer affinity to the protein, which is strong enough to surface binding events, i.e. for affinity-based biosensing. Based on the capture it but also weak enough to release it in presence of a moderate collected data, an electric equivalent model was developed for bare, flow of the surrounding solution. Fig. 6a shows the average variations laser-treated and wax-passivated devices. The main part of the model with the respective standard deviations over time, corresponding to the (Fig. 4b) consisted of a simple series of the resistance of the solution (Rs) NGAL injections performed every 10 min. The maximum variation and the electrode double layer constant phase element (Cdl). No charge reached was with the 25 nM injection with a corresponding value of transfer resistance was added since the measurements were performed almost 300 Ohms. The calibration on single electrodes also allowed without any additional redox probe to allow direct sample testing, lowering the standard deviations, minimizing the effect of the devices without any sample pre-treatment or solution addition. We added to this reproducibility. simple model a constant phase parasitic element (Ccoating), which In the case of electrochemical biosensors, to guarantee POC use of accounted for the conduction through the wet coating with phase values the developed system, it is essential to push both towards the scalability in the range 30-40◦ . The experimental MCH data were fitted to the of the proposed biosensor fabrication method and to the availability of model and the variation of its key parameters, as the double-layer portable and widespread instrumentation to perform the measurements capacitance and the coating parasitic capacitance are shown in Fig. 4c and analysis. In our work, the first goal was accomplished with the and d. simple “plug & print” approach for device production, which guarantees As the defined model in Fig. 3b indicates, the higher the coating reliability and scalable low-cost production in office-like environments impedance the more effective the coating passivation. With the absolute as well as a high level of customization (the total fixed and variables value of impedance inversely proportional to the capacitance, we would fabrication costs are presented in detail in Table S2, in the supporting expect a better passivated device to have lower Ccoating values. In these information). To pursue the second goal, the smartphone with its terms, the fitting results in Fig. 3c prove again that wax offered the best widespread use and computational power is the best strategy. To this passivation performance, while the Cdl clearly indicates how the wax- end, we used the audio and microphone channels of an Android passivated devices increased their sensitivity to the surface modifica­ smartphone and a commercially available app (Google Play link) tion by MCH (significant variation at 10 times lower MCH concentra­ together with a simple custom-made $5 electronic circuit to connect our tions) with respect to the other two types. devices to the smartphone and perform EIS measurements in the 50 Hz - 20 kHz frequency range. 3.6. Modification of biosensors with NGAL aptamers We tested the devices with the same testing protocol previously used with the potentiostat (increasing NGAL injection concentrations every As the wax-passivated devices showed the most promising results in 10 min), but measuring the full spectra through the smartphone after 10 the previous tests, we decided to functionalize them with aptamers min of incubation. The impedance variations at 150 Hz (which resulted specific for NGAL for real-time detection in urine. The device slightly less noisy rather than the 100 Hz values) are reported in Fig. 6c 8 G. Rosati et al. Biosensors and Bioelectronics 196 (2022) 113737 and are coherent with the ones observed in Fig. 6a with the potentiostat. Declaration of competing interest We also tested both setups for variations obtained by repeated injection of the buffer only on the functionalized devices, obtaining negligible The authors declare that they have no known competing financial variations over the whole duration of the test. interests or personal relationships that could have appeared to influence Finally, the NGAL detection sensor was tested with NGAL dilutions in the work reported in this paper. artificial urine to emulate the environment of a real application. The results, represented in Fig. 6d, show a linear variation of the impedance Acknowledgements in the 1–300 nM range. Furthermore, we tested our sensors with 30 nM BSA, 30 nM NGAL and 30 nM BSA+30 nM NGAL, obtaining the results The development of the device fabrication platforms has been resumed in the boxplot in inset of Fig. 6d. These results prove the partially funded through the MICROB-PREDICT project (https://micr biosensor reliability for the NGAL POC monitoring and its resilience to ob-predict.eu/). unspecific binding. The MICROB-PREDICT project has received funding from the Euro­ On the basis of the data obtained by the staircase calibration, after pean Union’s Horizon 2020 research and innovation programme under the analysis of the impedance variations referred to the blank mea­ grant agreement No 825694. This reflects only the author’s view, and surements, we determined the performance of our biosensor by means of the European Commission is not responsible for any use that may be its sensitivity, linear range, and limit of detection. For the latter we made of the information it contains. needed to introduce an alternative definition. In general, the LOD is We thank Dr Alex Chamorro and Prof. Kevin Plaxco for the NGAL calculated as the analyte concentration that causes a signal variation aptamers sequence used for the case study of our biosensors. E. P. N. greater to 3 times the standard deviation of the blank measurements. acknowledges funding through the EU’s Horizon 2020 research and However, in our case the calibration curves are built as the variations of innovation programme under the Marie Skłodowska-Curie grant the impedance signal at a certain frequency with respect to the blank agreement No. 754510. C.C.C.S. acknowledges funding through CAPES measurement. Therefore, the blank measurements result all at zero and a – PRINT (Programa Institucional de Internacionalização; grant standard deviation value cannot be obtained. We decided to calculate #88887.310281/2018-00 and 88887.467442/2019-00) and Mackpes­ the LOD as the first tested concentration causing an impedance variation quisa-UPM. greater than 3 times its standard deviation. This method should over­ This project has received funding from the European Union’s Hori­ estimate the actual LOD of the sensor but we think it is the more suitable zon 2020 research and innovation programme under grant agreement for our type of calibration curves. No 101008701 (Project acronym: EMERGE) Also our sensitivity definition may appear unusual, particularly for its measurement unit. Despite that, in this case, we used the common Appendix A. Supplementary data definition, i.e., the slope of the linear part of the calibration curve, which in our case is measured in Ohms over NGAL concentration (Ohm/ Supplementary data to this article can be found online at https://doi. nanomolar). org/10.1016/j.bios.2021.113737. Considered this, our biosensor showed a sensitivity of 31.49 Ohms/ nM, a linear range from 3 to 300 nM and a LOD of 3 nM in artificial References urine, considering the impedance variations at 1 Hz. In PBS and at 100 Hz, the biosensor showed a sensitivity of 7.92 Ohms/nM, a linear range Arroyo-Currás, N., Dauphin-Ducharme, P., Scida, K., Chávez, J.L., 2020. Anal. Methods 12, 1288–1310. https://doi.org/10.1039/d0ay00026d. from 3 to 30 nM and a LOD of 10 nM. Despite it may looks counterin­ Arroyo-Currás, N., Somerson, J., Vieira, P.A., Ploense, K.L., Kippin, T.E., Plaxco, K.W., tuitive to obtain a better performance in artificial urine rather than in 2017. Proc. Natl. Acad. Sci. U. S. A. 114, 645–650. https://doi.org/10.1073/ PBS, we recall that it strongly depends on the chosen frequency for the pnas.1613458114. Azzalini, L., Spagnoli, V., Ly, H.Q., 2016. Can. J. Cardiol. 32, 247–255. https://doi.org/ impedance variations analysis. The frequency optimization is thus a very 10.1016/J.CJCA.2015.05.013. important step, which needs to be performed in the final matrix to obtain Cheng, L., Lopez-Beltran, A., Bostwick, D.G., 2012. https://doi.org/10.1002/978111827 the best performance results. 5436.ch32. Christodouleas, D.C., Kaur, B., Chorti, P., 2018. ACS Cent. Sci. 4, 1600–1616. https://doi. org/10.1021/acscentsci.8b00625. 4. Conclusions Cummins, G., Desmulliez, M.P.Y., 2012. Circuit World 38, pp. 193–213. https://doi.org/ 10.1108/03056121211280413. Curtis, S.D., Ploense, K.L., Kurnik, M., Ortega, G., Parolo, C., Kippin, T.E., Plaxco, K.W., In conclusion, in this work, we designed a simple IDE layout and we Arroyo-Currás, N., 2019. Anal. Chem. 91, 12321–12328. https://doi.org/10.1021/ simulated it by FEM, selecting optimal geometrical parameters on the acs.analchem.9b02553. base of the electronic characteristics of the AgNP lines. Then, we printed Da Silva, E.T.S.G., Miserere, S., Kubota, L.T., Merkoçi, A., 2014. Anal. Chem. 86, the IDE with a consumer printer and commercially available ink and 10531–10534. https://doi.org/10.1021/ac503029q. Hernández-Neuta, I., Neumann, F., Brightmeyer, J., Ba Tis, T., Madaboosi, N., Wei, Q., substrate without any sintering step. These devices were then charac­ Ozcan, A., Nilsson, M., 2019. J. Intern. Med. 285, 19–39. https://doi.org/10.1111/ terized with EIS measurements for use as affinity label-free and redox- joim.12820. free biosensors. We introduce and studied an effective method to Idili, A., Parolo, C., Ortega, G., Plaxco, K.W., 2019. ACS Sens. 4, 3227–3233. https://doi. org/10.1021/acssensors.9b01703. reduce the parasitic conduction through the semi-permeable substrate Jaiswal, K., Anand, V., 2021. In: Rizvanov, A.A., Singh, B.K., Ganasala, P. (Eds.). Springer coating allowing the room temperature sintering by using a consumer Singapore, Singapore, pp. 459–471. wax printer and we isolated the contacts of the devices by plastic Kara, P., de la Escosura-Muñiz, A., Maltez-da Costa, M., Guix, M., Ozsoz, M., Merkoçi, A., 2010. Biosens. Bioelectron. 26, 1715–1718. https://doi.org/10.1016/j. lamination. Therefore, we introduced a complete biosensors fabrication bios.2010.07.090. platform using only office low-cost equipment. We used MCH SAM as a Lin, P.H., Li, B.R., 2020. Analyst 145, 1110–1120. https://doi.org/10.1039/c9an02017a. model to validate the sensitivity of the EIS devices to surface coverage by Manoto, S.L., Mabena, C., Malabi, R., Ombinda-Lemboumba, S., El-Hussein, A., Kasem, M., Lugongolo, M., Mthunzi-Kufa, P., 2020. Proc.SPIE. https://doi.org/ biomolecules. Finally, we functionalized the validated devices with 10.1117/12.2546444. NGAL aptamers as a proof-of-concept of their validity for biosensing McGrath, M.J., Iwuoha, E.I., Diamond, D., Smyth, M.R., 1995. Biosens. Bioelectron. 10, purposes and we detected NGAL in PBS and artificial urine using a low- 937–943. https://doi.org/10.1016/0956-5663(95)99231-9. Medina-Sánchez, M., Martínez-Domingo, C., Ramon, E., Merkoçi, A., 2014. Adv. Funct. cost smartphone-based system in the clinical range for AKI continuous Mater. 24, 6291–6302. https://doi.org/10.1002/adfm.201401180. monitoring at the point of care. Mishra, J., Mori, K., Ma, Q., Kelly, C., Barasch, J., Devarajan, P., 2004. Am. J. Nephrol. 24, 307–315. https://doi.org/10.1159/000078452. 9 G. Rosati et al. Biosensors and Bioelectronics 196 (2022) 113737 Parolo, Claudio, Idili, A., Ortega, G., Csordas, A., Hsu, A., Arroyo-Currás, N., Yang, Q., Sabaté del Río, J., Henry, O.Y.F., Jolly, P., Ingber, D.E., 2019. Nat. Nanotechnol. 14, Ferguson, B.S., Wang, J., Plaxco, K.W., 2020. ACS Sens. 5, 1877–1881. https://doi. 1143–1149. https://doi.org/10.1038/s41565-019-0566-z. org/10.1021/acssensors.0c01085. Sui, Y., Zorman, C.A., 2020. J. Electrochem. Soc. 167, 037571 https://doi.org/10.1149/ Parolo, C., Sena-Torralba, A., Bergua, J.F., Calucho, E., Fuentes-Chust, C., Hu, L., 1945-7111/ab721f. Rivas, L., Álvarez-Diduk, R., Nguyen, E.P., Cinti, S., Quesada-González, D., Tortorich, R.P., Shamkhalichenar, H., Choi, J.W., 2018. Appl. Sci. 8 https://doi.org/ Merkoçi, A., 2020. Nat. Protoc. https://doi.org/10.1038/s41596-020-0357-x. 10.3390/app8020288. Quesada-González, D., Merkoçi, A., 2017. Biosens. Bioelectron. 92, 549–562. https://doi. Vaidyanathan, R., Shiddiky, M.J.A., Rauf, S., Dray, E., Tay, Z., Trau, M., 2014. Anal. org/10.1016/j.bios.2016.10.062. Chem. 86, 2042–2049. https://doi.org/10.1021/ac4032516. Rosati, G., Daprà, J., Cherré, S., Rozlosnik, N., 2014. Electroanalysis 26, 1400–1408. Wang, J., Yu, J., Yang, Q., McDermott, J., Scott, A., Vukovich, M., Lagrois, R., Gong, Q., https://doi.org/10.1002/elan.201400062. Greenleaf, W., Eisenstein, M., Ferguson, B.S., Soh, H.T., 2017. Angew. Chem. Int. Ed. Rosati, G., Ravarotto, M., Sanavia, M., Scaramuzza, M., De Toni, A., Paccagnella, A., 56, 744–747. https://doi.org/10.1002/anie.201608880. 2019a. Sens. Bio-Sensing Res. 26, 100308. https://doi.org/10.1016/j. Wünscher, S., Abbel, R., Perelaer, J., Schubert, U.S., 2014. J. Mater. Chem. C 2, sbsr.2019.100308. 10232–10261. https://doi.org/10.1039/C4TC01820F. Rosati, G., Ravarotto, M., Scaramuzza, M., De Toni, A., Paccagnella, A., 2019b. Sensor. Actuator. B Chem. 280, 280–289. https://doi.org/10.1016/j.snb.2018.09.084. 10