concentration. Each data point represents the mean ± SD (*n* = 2). **3. Discussion**

reduced sensitivity. [21].

**3. Discussion**  In this study, we demonstrated for the first time, a competitive magnetic immunodetection assay for efficient detection and quantification of aflatoxin B1 with comparable sensitivity to described laboratory-based cELISA. Especially due to the handheld reader device which can be operated using a portable battery and without additional laboratory equipment, the method described here is suitable for sensitive on-site aflatoxin B1 testing. Commonly used analytical methods for detection of mycotoxins either exhibit high sensitivity in a laboratory-based setting or can be used on-site with In this study, we demonstrated for the first time, a competitive magnetic immunodetection assay for efficient detection and quantification of aflatoxin B1 with comparable sensitivity to described laboratory-based cELISA. Especially due to the handheld reader device which can be operated using a portable battery and without additional laboratory equipment, the method described here is suitable for sensitive on-site aflatoxin B1 testing. Commonly used analytical methods for detection of mycotoxins either exhibit high sensitivity in a laboratory-based setting or can be used on-site with reduced sensitivity [21].

To establish the novel cMID method, several concentration settings of a cELISA as reference method were tested and optimized using checkerboard titration (Figure A1). Here, the most suitable combination of antigen coating concentration and the amount of used antibody was determined. As explained in Section 2.1.1, in all tested antibody concentrations, for a coating of 0.4 <sup>µ</sup>g·mL−<sup>1</sup> and onwards, a saturation in measuring signal could be detected (Figure A1). A clear difference to this can be seen within the checkerboard optimization experiment for cMID (Figure A2). Here, a saturation-like performance can either be seen when using antibody concentrations up to 2.5 <sup>µ</sup>g·mL−<sup>1</sup> or, if further increasing antibody concentrations, saturation cannot be reached. This could be due to the more than 40-fold higher protein binding surface of immunofiltration columns in comparison to the binding surface of an ELISA microtiter plate well. Especially due to the highly porous membrane of the immunofiltration column, a higher protein binding capacity can be reached. High binding 96-well microtiter plates (Greiner Bio-One, article number 655061) have a maximum binding capacity of 600 ng·cm<sup>2</sup> , whereas ABICAP immunofiltration columns can bind up to 24,000 ng·cm<sup>2</sup> , which in fact explains the absence of signal saturation at the used conditions. A further difference when using checkerboard titration for assay optimization of cELISA or cMID can be seen in fluctuating readout signals, whereas in ELISA, concentration-dependent signals were obtained with almost no variability between consecutive measuring points. In contrast, signals generated in MID checkerboard titration varied, especially at high antibody concentrations. This could be caused by an overdose of magnetic particles, which might result in sterical hindrances and a partly erratic aggregation of particles within the assay matrix. Based on this hypothesis, a magnetic-particle dose adaptation was made by reducing the applied concentration towards robustly detectable saturated measuring signals (Figure 5).

However, the resulting cMID assay did not lead to the expected results when using 700 nm magnetic beads. Switching to 70 nm magnetic particles instead resulted in a tenfold increase of measuring signal combined with reduced variability and provides a basis for a highly sensitive and reliable assay (Figure 6). The results presented here correlate with data published by Achtsnicht et al. in 2019, where the authors used magnetic frequency mixing for the detection of cholera toxin subunit B (CTB) in a sandwich-based manner [27]. In their experiments, using magnetic particles with hydrodynamic diameters of 75 nm and 1010 nm, they observed an eight-fold higher response signal with smaller particles compared to the larger beads when detecting a concentration of 750 ng·mL−<sup>1</sup> CTB. The most simple explanation for this striking difference would be the higher absolute number of small particles compared to larger particles when used at identical concentrations (80 <sup>µ</sup>g·mL−<sup>1</sup> ) since the molar concentration is proportional to the molecular weight of the particles. By applying a higher ratio of small particles onto the column, a higher amount of small MPs could bind to antigens coated on the column surface. Another explanation could be the occurrence of steric hindrance within the ABICAP immunofiltration matrix when using large particles. Especially if there is a low concentration of free mycotoxin within the sample, a high amount of biotinylated antibody will be retained within the matrix in a competitive assay approach, resulting in a high antibody density on the matrix surface. When large MPs are applied, steric hindrance can occur at the surface. Either multiple, closely located biotinylated antibodies could be linked to one MP, or multiple antibodies could be blocked due to the size of large MPs. Here, especially the thousand-fold increased volume of 700 nm diameter MPs in comparison to 70 nm MPs in diameter might have the most important role. By the greatly increased size and volume, several antibodies are covered by one single MP, so that they are no longer accessible for binding of further MPs. By using smaller particles, shielding effects and sterical hindrances at the surface are reduced, resulting in a higher magnetic particle density within the column, and finally, a higher readout signal. A similar effect observed by Achtsnicht et al. (2019) when using small 75 nm instead of large 1010 nm particles in their MID experiments was also explained by the blocking of remaining binding sites [27]. The authors also found similar high standard deviations when large particles remain in a high concentration within the column (Figure 6A). With such a high SD, sensitive detection of either CTB or aflatoxin B1 is almost impossible. In their study, a LOD of 3.1 ng·mL−<sup>1</sup> was found when using the large particles, but a more than 15-fold lower LOD of 0.2 ng·mL−<sup>1</sup> could be

reached using the small-sized MPs. In our study, especially due to the high standard deviations at high measuring signals when exciting the 700 nm MPs for aflatoxin B1 detection, a reliable calculation of LOD or IC<sup>50</sup> values was not possible. However, by using 70 nm particles, high and stable readout signals in combination with a perfectly matching nonlinear-fit (R<sup>2</sup> = 0.9938) enabled the reliable calculation of LOD and IC<sup>50</sup> values of 1.1 ng·mL−<sup>1</sup> and 5.4 ng·mL−<sup>1</sup> , respectively, which are in the same sensitivity range as our comparative lab-based cELISA

Although the best performance of the cMID assay was achieved with 70 nm small magnetic particles, the general applicability of 700 nm large particles in cMID assays should not be completely neglected since, in contrast to small particles, larger particles can be separated in a gradient magnetic field. However, the possibility of magnetic separation and thereby enrichment of magnetic particles might play a crucial role in sample preparation where cleanup and concentrating of an analyte is required to achieve sufficient assay sensitivity, as it was used for example by Lee et al. (2013) or Xuan and colleagues (2019) [30,31]. In the case of mycotoxin detection, magnetic separation-based sample cleanup could be beneficial, especially when working with partially soluble or insoluble, or roughly homogenized, e.g., grain samples. After magnetic beads captured the mycotoxin molecules in the sample, a magnetic separation step can be applied in which magnetic particles are retained in the magnetic field, while sample debris can be discarded. Additionally, separated mycotoxin-loaded magnetic particles could be resuspended in a much smaller volume, resulting in an enrichment of analyte. As a consequence, this could lead to increased sensitivity when samples are finally analyzed by cMID.

The 70 nm particles used in this study cannot be efficiently separated in a magnetic gradient field, since their magnetic attraction is low, which was shown by Achtsnicht and colleagues (2018) [32]. They correlated the magnetophoretic velocity of superparamagnetic particles when applying a magnetic field and found an increased velocity with increasing particle size [32]. Typically only bigger particles above 700 nm are used for magnetic separation experiments [30,31]. Further testing of different beads or other strategies as combining big and small-sized particles will be addressed in further studies to enable an optimized pairing of separation and cMID.

In conclusion, we successfully demonstrated the development and implementation of a competitive magnetic immunodetection assay for the detection and quantification of aflatoxin B1 with a LOD of 1.1 ng·mL−<sup>1</sup> . Based on our findings, it can be concluded that the competitive magnetic immunodetection is a powerful tool for portable, easy-to-use on-site monitoring of mycotoxin contamination in various matrices to reduce the risk of processing contaminated food and agricultural products. In future work, we will focus on the further optimization of the cMID procedure using the 70 nm MPs. Especially, a significant reduction of the assay time from currently approximately 4.5 h to less than one hour should be addressed, similar to that described by Rettcher et al. (2015) [26]. There a sandwich-based MID assay lasting less than 30 min was achieved by a more than 50% reduction of initially needed assay time [26]. By using MPs, pre-conjugated with specific anti-mycotoxin antibodies, ready-to-use coated and blocked immunofiltration columns, and testing successive reductions of incubation times of sample-pre-incubation and competitive binding reaction within the matrix, an additional increase of applicability for on-site testing should be possible. The applicability of the described cMID approach will be further studied regarding the detection and quantification of other mycotoxins, as well as a combination of those toxins within a food matrix of one sample. Especially the multiplex detection of various mycotoxins within one sample will be addressed as shown by the multiplex detection of several antibodies in stacked sample matrices by Achtsnicht et al. (2019) [33]. Here the authors detected two different target molecules, namely antibodies, within one sample solution by stacking 3D printed immunofiltration columns coated with different capture antibodies in a sandwich-based MID approach. By adapting this procedure, multiple individually coated matrices could be used for the specific retention of corresponding anti-mycotoxin antibodies in a cMID assay, and with this, a multiplex detection in a food matrix could be obtained.
