*3.2. WLRS Assay*

### 3.2.1. Assay Optimization

The transfer of the carbendazim assay to the WLRS platform was based on results obtained in ELISA-optimization experiments, e.g., in terms of the composition of the assay buffer and the duration of the preincubation step. However, there were some parameters that also needed optimization. Due to the competitive nature of the assay, the two most crucial parameters to optimize were the concentration of benzimidazole conjugate used for coating and the concentration of anti-carbendazim antibody, as their combination determines the highest signal (zero standard signal) and the assay sensitivity. Hence, different concentrations of the benzimidazole conjugate (100–1000 μg/mL) in combination with different concentrations of the anti-carbendazim antibody (0.5–5.0 μg/mL) were tested. As shown in Figure S4, for all antibody concentrations tested, zero standard signal values increased as the concentration of benzimidazole-conjugate used for coating inreased and maximum signal plateau values were obtained for concentrations ≥500 μg/mL. The increase in the signal observed as the concentration of both benzimidazole-conjugate and anti-carbendazim antibody increases, depicts the increase in the biomolecular adlayer thickness due to the antigen–antibody reaction taking place onto the chip surface. Regarding the antibody concentration, it was found that the signal increased almost linearly up to an antibody concentration of 2 μg/mL, whereas higher antibody concentrations provided only marginal signal (10%) increase. In addition, as it can be concluded from Figure S5, the assay sensitivity deteriorated by approximately 50% when the anti-carbendazim antibody concentration increased from 2 to 4 μg/mL. Thus, a 2 μg/mL anti-carbendazim antibody concentration was adopted in the final assay protocol.

Another parameter to optimize was the duration of the whole assay, in order to achieve a fast and reliable assay. As shown in Figure 3, the primary immunoreaction reached a signal plateau after 30 min, the secondary immunoreaction in 20 min, and the reaction with streptavidin in 3 min. The real-time sensor response showed that the reaction with biotinylated secondary antibody and the subsequent reaction with streptavidin resulted in considerable signal enhancement. In particular, the reaction with the secondary antibody increased by 2.5-fold the signal obtained from the primary immunoreaction, whereas the reaction with streptavidin further increased the signal by approximately 3-times, leading to an overall 7.5-fold increase in the signal of the primary immunoreaction. The signal increase observed in the two reactions that follow the primary immunoreaction is ascribed to the fact that more than one biotinylated secondary antibody molecules bind per immunosorbed primary antibody, since the secondary antibody is a polyclonal one and has been raised against the whole anti-rabbit IgG molecule and it is expected, therefore, to bind to several epitopes of the primary antibody molecule. Similarly, the secondary antibody has several biotin-moieties and thus, more than one streptavidin molecule could bind per biotinylated secondary antibody molecule, providing considerable signal enhancement. Due to this accumulation of multiple molecules per immunosorbed primary antibody molecule, the effective biomolecular adlayer thickness, and consequently the WLRS sensor signal, is considerably increased.

**Figure 3.** Real time response obtained from a biochip functionalized with 500 μg/mL of benzimidazole conjugate upon running over the chip: assay buffer (from start to point A); a 1:1 (*v/v*) mixture of zero standard with a 2 μg/mL rabbit anti-carbendazim antibody solution (A–B); a 1:200 diluted solution of a biotinylated secondary antibody (B–C); a 10 μg/mL streptavidin solution (C–D).

The signal increase achieved by the implementation of the two signal enhancement steps allowed the reduction of the whole assay time. More specifically, the duration of the primary immunoreaction was set at 18 min, the reaction with the secondary antibody at 7 min, and that with streptavidin at 3 min, resulting in a total assay time of 28 min. This reduction in assay time reduced also the maximum signal obtained by 50%, however, the signal received with the shorter assay protocol was adequate for the performance of the assay. This was confirmed by the good discrimination of the real-time sensor responses obtained for carbendazim standards with concentrations ranging from 0 to 20 μg/mL as shown in Figure 4a. In addition, Figure 4b depicts the relevant calibration plot.

**Figure 4.** (**a**) Real-time responses obtained from biochips functionalized with 500 μg/mL benzimidazole conjugate upon running sequentially: assay buffer; a 1:1 *v/v* mixture of carbendazim standards (0–20,000 ng/mL) prepared in assay buffer with a 2 μg/mL anti-carbendazim antibody solution in the same buffer; a 1:200 diluted solution of biotinylated secondary antibody in assay buffer; a 10 μg/mL streptavidin solution. (**b**) Typical calibration plot for carbendazim. Each point represents the mean value of 4 runs ± SD.

The LoD of the proposed immunosensor was evaluated as described in Section 3.1.1 for the respective ELISA assay and was found to be 20 ng/mL. The assay dynamic range was 50 ng/mL–20 μg/mL.

#### 3.2.2. Optimization of Sample Preparation Procedure

The developed immunosensor was applied to carbendazim detection in fruit juices. The acidic pH and the pulp in fruit juices are reportedly impacting the performance of immunoassays by affecting the antibody–antigen binding, resulting in the so-called "matrix-effect" [23,24]. Thus, it is necessary to perform a sample preparation procedure that minimises interferences before analysing fruit juice samples. In this respect, filtration through a 0.45 μm PTFE membrane syringe filter was performed to remove the pulp, and then the pH of the filtrate was adjusted to 7.4 ± 0.2 with addition of 1 M NaOH solution (without significantly changing the sample volume). In addition, to investigate the possible matrix effect of fruit juice samples to assay performance, the signal obtained by undiluted and 2- to 20-fold diluted samples was compared to that of zero standard signal in buffer. As shown in Figure 5a for an orange juice sample, the undiluted sample, which has been used after filtration and pH neutralization, provided the same signal with the assay buffer indicating that there was no need for additional sample dilution with assay buffer in order to alleviate the matrix effect. This is supported by the fact that the undiluted sample provided statistically the same zero standard values with orange juice diluted 2–20 times with assay buffer. Similar results were obtained using juices from other fruits, e.g., lemon, and from juices made from a combination of different fruits. Thus, all juices involved in the study were analyzed without dilution after filtration and pH adjustment. Further verification that there was not any matrix-effect in the assay performance was provided by the fact that the carbendazim calibration plots obtained in assay buffer and a suitably treated, commercial orange juice were superimposed, as presented in Figure 5b.

**Figure 5.** (**a**) Zero standard signal corresponding to assay buffer and filtered and pH-adjusted juice, undiluted and diluted 2× to 20× with assay buffer. Each bar represents the mean value of 3 independent measurements ± SD. (**b**) Typical calibration plots obtained with carbendazim standards prepared in assay buffer (black squares) and treated orange juice (red circles). Each point represents the mean value of 4 runs ± SD.

#### 3.2.3. Accuracy and Precision of the Developed ELISA and Sensor Assays

The accuracy of the measurements performed with the developed immunosensor was evaluated through recovery experiments using three commercially available juices prepared from different fruits. To this end, three fruit juices previously analyzed with the carbendazim ELISA and found not to contain any detectable carbendazim were fortified with concentrations of the pesticide that corresponded to three concentration levels, i.e., 100, 500, and 1000 ng/mL. The fortified samples were analyzed in triplicates both with the developed biosensor and the ELISA assay, prior to and after the addition of carbendazim. The results obtained from the analysis of the spiked samples are shown in Table 1, while no carbendazim could be detected in any of the samples prior to the addition of the pesticide. The percent recovery was calculated as the percent ratio of the carbendazim concentration determined in the spiked samples to that expected based on the amount spiked. The recovery values determined with the two methods are presented in Table 1. As shown, the recovery values obtained with the ELISA and the biosensor assays ranged from 90 to 110% and 89 to 110%, respectively, demonstrating the high accuracy of the assays developed. Furthermore, there was a very good agreemen<sup>t</sup> of the values determined with the immunosensor to those determined for the same samples with the ELISA. A paired *t*-test confirmed that there was not statistically significant difference between the results of the two methods (*p* < 0.05).


**Table 1.** Recovery values of carbendazim in commercial fruit juices spiked with the indicated concentrations of the pesticide.

3.2.4. Analysis of Commercially Available Fruit Juices

A survey on carbendazim residues in commercial packages of fruit juices was performed. Ten different juices of four different brands made from orange, lemon, or combination of different fruits were purchased from local supermarkets and analyzed by the WLRS sensor. The results are presented in Table 2. None of the products tested contained detectable amounts of carbendazim. This finding was confirmed by analysing the same samples with the carbendazim ELISA.

**Table 2.** Carbendazim residues in commercially available fruit juices. In the parentheses the trade name and storage conditions recommended by the manufacturers are provided.


#### 3.2.5. Regeneration of Biochips

The stability of the developed immunosensor response to sequential assay/regeneration cycles was also determined as a means to exploit the use of a single biochip for analysis of several samples thus reducing the analysis cost. A low-pH buffer is often used in affinitybased biosensor assays for surface regeneration purposes, to quantitatively remove the antibody from the coating analyte, without affecting the coating analyte structure. The regeneration was achieved by running a 0.1 M glycine-HCl buffer, pH 2.5, for 3 min after completion of the assay. Figure 6 shows the zero standard responses obtained from a single biochip in 15 assay/regeneration cycles performed over a period of three days. As shown, for up to 12 assay/regeneration cycles, all values consistently fell within the mean value ± 2SD range, which demonstrates the potential reuse of biosensor.

**Figure 6.** Signal responses obtained from a single biochip for 15 regeneration/assay cycles. Dashed lines represent the mean value ± 2SD limits.

#### *3.3. Comparison with Other Biosensors*

In the last decade carbendazim has received particular attention from the research community and many efforts from research groups all over the world have been carried out towards the development of biosensors for carbendazim determination in various food commodities. The majority of the reported carbendazim biosensors are electrochemical [25–31]. Only few optical sensors for carbendazim detection have been reported in the literature. These include a sensor based on surface-enhanced Raman scattering technology, which involves cyclodextrin inclusion complexes on gold nanorods as recognition element [32], a sensor based on luminescence resonance energy transfer from aptamer-labeled upconversion nanoparticles to manganese dioxide nanosheets that act as an acceptor [33], and a surface plasmon resonance-based immunosensor [34]. The immunosensor proposed herein is the first optical sensor based on white light reflectance spectroscopy, dedicated to the determination of carbendazim in foodstuff.

With respect to actual analysis time of 28 min, our sensor is considered among the fastest sensors reported for carbendazim detection in foodstuff generally; even if the separate, 60-min preincubation step is taken into account, the proposed sensor can be still considered as an analytical tool capable of determining carbendazim very quickly. In terms of analytical sensitivity, the LoD of the developed method (20 ng/mL) is well below the current European Union regulatory limit of 200 ng/mL for carbendazim in fruit juices. It is noteworthy that in our case, as in the aforementioned SPR sensor [34], a signal enhancement step was introduced after the primary immunoreaction in order to generate a measurable response while keeping the analysis time as short as possible. Overall, the developed sensor could be regarded as a very fast and sensitive real-time biosensing platform for the detection of carbendazim, which could be considered as label-free, since it is not based on any typical label, such as a fluorophore or an enzyme, for the development of an optical signal. The sensor is accompanied by a rather simple sample preparation protocol that can be easily reproduced at the point-of-need as it does not require any special equipment. In addition, the sample preparation procedure could be applied to juices prepared from different fruits without any noticeable effect in the immunosensor performance by the difference in the sample matrix, which is very important for future on-site analysis of different juices.
