1. Introduction
Food safety plays a crucial role in detecting the presence of harmful substances, such as pesticide residues, which can severely harm consumers’ health. Regulatory control of random samples, as commonly implemented in laboratories, yields less than 1% positive detections [
1]. Thus, there is a demand for a more comprehensive complementary screening in order to protect the population from uncounted threats. In contrast to instrumental methods, which are expensive and time-consuming, simpler and more affordable screening tools are necessary. Here, we investigate such a possibility with a hybrid lab-on-a-chip multiple injector disposable device and processing methodology. The platform is able to perform the complete protocol necessary for a robust detection with smartphone cameras, within predicted semiquantitative categories and with the ability to reject failed measurements.
Carbamate (CM) and organophosphate (OP) insecticides in foods are globally used to boost agricultural productivity, and thus constitute a relevant target for in situ screening of samples. Although CMs and OPs do not usually release toxic metabolites and are degradable under ambient conditions, they feature neurotoxic effects originating from acetylcholinesterase (AChE) inhibition, a vital enzyme in the neural system of many organisms. AChE inhibition results in acetylcholine (ACh) accumulation, which can lead to serious health problems, including respiratory and myocardial malfunctions [
2]. One such CM, carbofuran, is a broad-spectrum systemic insecticide, nematicide, and acaricide commonly used worldwide [
3]. Although carbofuran has been reported as a highly harmful compound [
4] and is banned in the European Union (EU), there are still carbofuran violation notifications in the EU Rapid Alert System for Food And Feed (RASFF,
https://goo.gl/h1Hc41, last accessed on 13 September 2019), mostly from imported foodstuffs. Maximum residue limits (MRLs) have been set in various food matrices, for instance 0.001 mg/kg in apple, which is ten times less than the EU default MRL (0.010 mg kg
−1), indicating harmful carbofuran potential.
Liquid and gas chromatography (LC and GC) coupled to tandem mass spectrometry (MS/MS) are the reference methods used for pesticide residues detection in foodstuffs [
5,
6]. Despite being highly accurate, robust, and with low limits of detection (LODs) at the sub-ppb level, chromatographic methods are also expensive, time-consuming, and require highly skilled operators. Therefore, such reference methods do not fit into the affordable complementary screening concept [
7,
8]. Alternative methods have been investigated, including immunoassays against specific CMs [
9], electrochemical detection [
10,
11], surface plasmon resonance (SPR), and fluorescence aided by nanomaterials (NMs) [
12]. Although these alternatives attain competitive LODs, they also suffer from high cost, complicated procedures, shelf-life limitations, and can require complex instrumentation [
2,
9,
10,
11,
12].
On the contrary, paper-based assays fulfil many of the practical requirements for in situ detection. The technology is not only portable, easy to use, cost-effective, disposable, and available worldwide, but can also be functionalized with enzymes for affordable detection [
13]. On the other hand, the integration of multiple-step protocols, which are well developed for classical lab-on-a-chip (LOC) methods, are not simple with paper devices. The challenge in this case was to enable the capabilities of active injectors in a disposable device integrating volume-metered reagents and references, as well as the paper detection stage.
AChE in vitro can hydrolyze certain substrates to colored products, and in the presence of an inhibitor its activity is reduced (producing less color). Although such sensors have been used for the detection of CMs and OPs, they were purely qualitative [
14], relied on inaccurate visual inspection of the results [
15,
16], and lacked sample handling integration.
In this study, semiquantitative, smartphone-based carbofuran screening was achieved using a hybrid 3D-printed paper–lab-on-a-chip injector device. Active injection in disposable devices is unusual due to the costs and complexity of micro injectors capable of delivering well-defined volumes [
17]. The present work reports on a disposable device featuring four integrated injectors able to deliver 15 μL volumes on demand at freely configurable protocol stages. This capability is achieved by an asymmetric flow resistance provided by the functionalized AChE paper assay itself, acting as a permeable barrier in the LOC micro channels. The devices were measured in a 3D-printed holder under controlled illumination provided by a smartphone camera flash, while the phone rear side camera video recorded the dynamic color response of sample and reference carbofuran concentrations. Recording the dynamic response enabled the actual analyte effect to be disentangled from the wicking, incubation, and fluidic effects in coloration. The observed responses can be explained by a double exponential model, which was used to predict semiquantitative ranges for robust identification and rejection of suboptimal detections, while providing a generic procedure able to accommodate the binary detection of carbofuran in apple extracts.
2. Materials and Methods
2.1. Design and Fabrication
2.1.1. Unibody-LOC (ULOC) Device
The ULOC devices were designed using Autodesk Inventor Fusion for Mac, and the generated .stl files were printed with a stereolithography (SLA) 3D printer (Formlabs, Form 1+) controlled by the PreForm 1.9 software, using Clear Type O4 resin and printing at 100 μm layer resolution.
ULOC devices were designed to self-support their features, thus avoiding the extra cost and time of creating supporting structures. For the current total volume of 2.28 mL, the cost of each prototype was 0.3€. Additionally, ULOC-implemented channels or chamber architectures open on at least one side, thus allowing easy removal of uncured resin by sonication in ethanol. The surfaces of the devices were refined using sandpaper, which allows the open channels to be sealed with adhesive tape. This step is no longer necessary with the next generation of SLA printers.
In the case of the current device, Whatman paper functionalized for the colorimetric detection was fixed on the underside of the ULOC using a plastic spring element, which secured the contact of the paper with the end of the microfluidic channel, thus creating four porous injector barriers in each device. The top tape was trimmed to expose the cells to the ambient illumination in the detection zone, preventing any limitation to the flow and avoiding interferences to the optical path.
2.1.2. Device Holder
The ULOC devices were measured within a holder that shielded ambient illumination and set a repeatable positioning with respect to the phone illumination (rear side flash) and camera. The holder was designed using Autodesk Inventor Fusion for Mac and printed with a fused deposition modeling (FDM) 3D printer (M3D) using a PLA filament. The design consists of 2 complementary parts used to attach the phone and guide the ULOC into position, and its final version was painted with black mat paint.
2.2. Chemicals and Reagents
Phosphate buffer saline (PBS) tablets, AChE from Electrophorus electricus, acetylthiocholine iodide (AThI, purity >99%), 5,5′-dithio bis-2-nitrobenzoic acid (DTNB, purity >99), bovine serum albumin (BSA), and Whatman® cellulose chromatography paper were from Sigma-Aldrich (Prague, Czech Republic). Anhydrous magnesium sulphate (MgSO4) was obtained from Fluka (Buchs, Germany), and sodium chloride (NaCl) from Penta (Chrudim, Czech Republic). Sorbent primary–secondary amine (PSA, Bondesil, 40 μm) was purchased from Agilent Technologies (USA). Deionized water was purified using a Milli-Q system (Millipore; Bedford, MA, USA). The 96-microwell plates were supplied by Gama Group (Ceske Budejovice, Czech Republic). AChE stock solution in PBS (2 U μL−1) was stored in the fridge at 4 °C. Acetonitrile were supplied as high purity solvents from Merck (Darmstadt, Germany). Carbofuran was analytical standard grade and purchased by Sigma Aldrich (Taufkirchen, Germany). Carbofuran stock solution (3.0 mg mL−1) was prepared in acetonitrile and kept in a freezer at −20 °C, protected from ambient light. Carbofuran working solutions were prepared daily in PBS.
2.3. Enzyme Immobilization
AChE was physically adsorbed on Whatman paper strips with 1.4 × 1 cm dimensions. The immobilization procedure was simple, as 10 μL of AChE working solution in PBS (0.5 U μL
−1) were pipetted into a microwell of a 96-microwell plate and a strip was dipped in the well using tweezers, resulting in a 5 U AChE/strip. Afterwards, the strips were dried at room temperature for 2 h and were ready for use. Measurements at this stage of proof of concept were conducted at room temperature, and the shelf-life of the paper assay was evaluated within a 56-day period (
Figure S5), showing a stable behavior.
2.4. Paper-Based AChE Assay
Immobilized AChE activity was measured both in the presence and absence of an inhibitor, carbofuran insecticide, using the Ellman’s assay, which is composed of two steps. Firstly, the paper strip was incubated with a sample for 8 min and then a solution of the enzyme substrate and the chromogenic reagent (75 mM AThI: 7.5 mM DTNB, 9:1 (v/v), respectively) was added, and the yellow color development was monitored for 3 mins using the 3D-printed smartphone reader.
2.5. Smartphone Readout
A Huawei P8 lite smartphone was used for the detection coupled to a 3D-printed holder. To achieve standardized and repeatable video acquisition, the free OpenCamera app was used, which permits the adjustment of exposure, focus, and illumination (see
Supplementary Materials, Table S1). Videos of the whole enzymatic reaction were recorded at 30 fps (frames per second) and processed by Python software.
2.6. Python Software and Graphical User Interface (GUI)
Acquired videos were imported in Python, using a program with a tkinter user interface (
Figure S6) that displayed an initial video frame to allow the interactive selection of coordinates for the regions of interest (ROIs), corresponding to the sample and reference cells. Besides the ROIs coordinates, the initial cut-off time (
to) to eliminate the wicking stage and the total considered response time could be specified. Once such selection was complete, data evaluation consisted of the recording of the blue channel intensity for each frame averaged for all pixels of each ROI. Such response for the reference and sample cells were generated and saved in .csv format for further analysis.
A complementary Python program was developed to collect multiple .csv outputs and operate with such data to create response signals by subtracting the sample dynamic response from the reference response. Such signal was resampled to 5000 data points and curve fitted to a 6-degree polynomial function, to avoid overfitting when the signals were projected in the principal component (PC) space for categorization. The correct categories were produced by the model in the same PC space, which enabled semiquantitative validated ranges to be created that could be used for automatic validation or rejection of individual measurements.
2.7. Food Samples
Apple bio-samples were purchased by the local market. A representative portion was homogenized with a PM-100 Retsch Planetary Ball Mill (Haan, Germany) and the samples were kept at –20 °C. Prior to paper assay, the samples were analyzed with the liquid chromatography tandem mass spectrometry (LC-MS/MS) method (see
Supplementary Materials) to verify that they did not contain detectable levels of CMs and OPs.
2.8. Food Samples Extraction
Quick, easy, cheap, efficient, rugged, and safe (QuEChERS) extraction is the golden standard for pesticide residue extraction, and thus it was applied to extract the tested samples as described in [
18], with slight modifications. To begin with, 10 g of representative sample was weighted into a 50 mL solution. When needed, samples were spiked with target analytes from a freshly prepared solution (20 µg mL
−1) in acetonitrile to achieve the final concentration and left for at least 30 min at room temperature. Then, 10 mL of acetonitrile was added and the tubes were shaken vigorously for 2 min. To achieve phase separation, 4 g MgSO
4 and 1 g sodium NaCl were added to each tube, shaken vigorously for 1 min, and finally centrifuged for 5 min at 10,000 rpm. Following the partitioning step, 3 mL of the acetonitrile extract was transferred to a 15 mL centrifuge tube containing 75 mg of PSA sorbent and 450 mg of MgSO
4. The tube was shaken for 1 min and centrifuged for 2 min at 6000 rpm. Then, 500 µL of the cleaned-up extract was transferred to a vial and evaporated until nearly dry under a gentle nitrogen steam. Finally, samples were reconstituted to 500 μL of PBS and subjected to the enzyme paper assay.
4. Conclusions
The current concept provides an autonomous platform to exploit the benefits of low-cost paper assays by complementing its weaknesses with tailormade disposable microfluidics, thus delivering on demand reagent and sample injection for a multiple-step protocol (0.30 €/device at the development stage). It must be noted that the reported cost of 0.30 €/device is a highly affordable development cost, which is made possible by modern 3D printing technology. This cost enables multiple design iterations a day with any number of modifications. In this particular case, more than 14 different optimization cycles were conducted during a 10 day period, with the ability to test each physical outcome. The natural next step is to migrate the design to scale manufacturing (e.g., injection molding) to make the final cost compatible with the paper fluidics.
Importantly, the collateral use of the paper assay as a porous barrier enabled the simplification of the LOC architecture, and allowed the integration of a reference channel at the same cost. The combination of the autonomous device with a proper light- and exposure-controlled acquisition extended the possibilities of the colorimetric assay to adopt a robust semiquantitative configuration, which enabled the differentiation of valid response identification from spurious measurements. The proposed concept provides a screening procedure for the binary identification of carbofuran in apple matrix and could support screening of different CMs and OPs at different concentration levels, since they share a similar inhibitory effect against AChE.