Optimization and Validation of ELISA for Aflatoxin B1 Detection in Fermented Forages and Feeds

Enzyme-coupled immunosorbent assays (ELISA) methods are usually validated only for homogenous matrixes like corn and wheat. More complex materials like fermented forages and mixed feed are not targeted for mycotoxin measurement. The low number of ELISA methods found in the literature neither contained the pH set for fermented forages nor dealt with the setting of the matrix:solvent ratio. The sample preparation of these matrixes needs to be optimized and validated for aflatoxin B1 analysis from fermented forages (corn silage and rye haylage) and mixed feed for Romer AgraQuant® Aflatoxin B1 ELISA (Romer Labs, Austria). Drying and pH adjustment of fermented forages had high importance before mycotoxin extraction. Because of the matrix swelling, the 1 : 5 ratio of the sample/extraction solute should have been increased to 1 : 8 to gain the highest aflatoxin B1 recovery. The accuracy and repeatability of the analysis were tested and found to be suitable for further application.


Introduction
Aflatoxin B1 (AFL B1) measurement possesses worldwide interest as a mycotoxin presents a real threat. Validation of an ELISA method for AFL B1 detection from some food and feed ingredients [1], mainly grain products, are wellknown, but the detection of AFL B1 by ELISA from fermented forages and feed mixtures by this rapid test is still a neglected area. However, ELISA applications are good options for the agricultural area as the tests are usually suitable for large numbers of samples and are not costly and time-consuming. e data suggest that due to the different matrix composition of the feeding stuff, the degree of recovery may greatly vary even for basic materials like grains (e. g., [2] Percent Recovery AgraQuant ® Aflatoxin B1 ELISA kit Romer Lab; corn: 91-102%, rice: 70-80%, soybeans: 87-114%, wheat: 87-103%). In the case of fermented forage samples, the presence of lactic acid bacteria may cause additional complications, which besides significantly decreasing the pH, may be able to passively bind aflatoxins in feed materials [3,4], thereby removing part of the toxin to be detected from the system. Interferences due to the individual matrix of the feed, e. g., low pH, cross-reaction with other aflatoxins, and excessive water absorption of the matrix can occur. e pH of the sample should be 6-8; hence, excessive alkaline or acidic conditions may affect the antigen-antibody binding in ELISA. erefore, the pH should be adjusted before the test.
We optimized the sample preparation and validated the ELISA method to detect contamination in AFL B1 fermented forages and feed samples using the AgraQuant ® Aflatoxin B1 ELISA test kit (Romer Labs, Austria).

Initial pH and a Corresponding Range of Feed Types.
e initial pH of the matrixes varied between 4.02 and 6.42, which is determined by the chemical composition of the starting material, such as protein and carbohydrate content, and by the presence of metabolites (e. g., volatile organic acids, lactic acid, and acetic acid) originated from microbes that utilize the nutrient sources. For the ELISA kit to function properly, the pH of the sample must be set to pH 6-8. at may affect subsequent ELISA results, as chemical reactions occur slightly in the matrix during pH adjustment, in which products such as salts and precipitates are formed that may interfere with the ELISA system, so we tried to minimize this effect when adjusting the pH.
We tested two chemicals, one commonly used NaOH and the other NaHCO 3 , which was applied in detecting deoxynivalenol and its conjugates in ELISA (Romer AgraQuant ® DON Assay Test Kit; Romer Labs) from beer [5]. Although precipitation was observed with both chemicals, the solution became less cloudy applying 1M NaOH, and a lower volume of the solution had to be added to the system to adjust the correct pH than when 1M NaHCO 3 solution was used at the same concentration. erefore, we decided to continue working with 1M NaOH solution.

Extraction Method.
e recommended protocol for sample preparation requires the matrix to be minced and then mixed or blended with the extraction agent for three minutes and immediately filtered. However, the blender commonly used for sample preparation is unsuitable for effectively homogenizing fibrous forages. erefore, the first crucial step of sample preparation was the homogenization of the sample because silage/haylage samples are shredded to different sizes, often containing coarse, hard-to-grind particles. e samples were dried and then shredded to small particle sizes in an industrial grinder. Finally, the dried sample minced to fine particles was used during the spiking.
After the dry sample had been spiked, we decided to carry out the extraction in lockable Falcon tubes (also to prevent evaporation) with a multifunctional shaker for half hour to provide sufficient time and a more effective protocol for recovering the toxin. All dried and spiked matrixes were tested for AFL B1 as quality control and were suitable for further ELISA process.

Sample-Extraction Solute Ratio.
With the initially recommended 1 : 5 sample: extraction solution ratio for homogenous and less fibrous matrixes such as corn or wheat and low recovery (below 70%) was detected for the fibrous fermented forages and mixed feed due to significant matrix swelling/liquid uptake. However, we wanted to improve the average recovery to higher than 70%. Furthermore, according to the literature, it is possible to reduce the distortion caused by the matrix effect by diluting the sample [6][7][8][9]. erefore, the dilution in the 1 : 5 ratio determined by the manufacturer for homogenous matrixes with low fibre content was increased up to 1 : 10 (Table 1).

Determination of the Limit of Detection and
Quantitation. Prepared sample matrixes were tested in the HPLC method to ensure their mycotoxin-free state. From the repeated ELISA measurement of the blank matrix samples, mean concentrations, standard deviations, and limit of detection and quantitation were calculated ( Table 2). From the calibration curves gained in ELISA, r 2 values were also checked and found between 0.995 and 0.985.

Accuracy.
e mean, standard deviation, and relative standard deviation are calculated by determining the concentration of eight independently measured solutions containing the spiked AFL B1 (Table 3).

Repeatability.
Two different operators determined the AFL B1 concentration of the matrixes from the same sample on different days, making three independent measurements per day. e results are used to determine significant differences between the groups by scattering analysis (Tables 4-6). For all matrixes, there were no significant differences within the analysis.

Discussion
e enzyme-coupled immunosorbent assays (ELISA) methods usually are validated for homogenous matrixes such as corn and wheat, and more complex materials like fermented forages and mixed feed are not targeted because of their complex composition, low pH, and high water content.
e fermented forages need to be dried and ground, and the pH needs to be set before ELISA analysis considering the high water content and the acidic pH of these materials. As the fermented forages and mixed feeds tend to swell in extraction solutions, higher dilution of the samples at a 1 : 8 ratio is recommended to gain a better recovery of the mycotoxin.
Most of the methods we found in the literature neither contained the pH set for fermented forages nor dealt with the setting of the matrix:solvent ratio. e proper ratio is essential to maximise the extraction and avoid the loss of the liquid phase due to matrix swelling, which can be a source of false measurements [10][11][12]. Maggira et al. [13] set the pH of the extract, but they worked with barley and Under modified sample preparation, the AgraQuant ® Aflatoxin B1 ELISA kit was successfully applied for feed and fermented matrixes with more than 70% recovery. Accuracy and repeatability analyses revealed a suitable process for sample preparation. Relative standard deviation (RSD (%)) was lower than 11% for all tested matrixes. e recovery of AFL B1 was the highest in rye haylage and the lowest in milking cow feed concentrate; however, even for that samples, the recovery of AFL B1 was higher than 70%, which is comparable to that of some grains. erefore, the method can be suggested as a quality control method.
is ELISA kit is ideal for measuring multiple samples in parallel, up to 42 samples, with an incubation period of 15 minutes. e kit is manufactured in a ready-to-use format to reduce handling errors. e kit contains five AFL B1 standards (0 ppb, 2 ppb, 5 ppb, 20 ppb, and 50 ppb), antibody-coated plate, conjugate, substrate, and stop solution.

A Detailed Description of the Sample Preparation.
Representative samples (at least 1 kg) were gained at the feed producers. Sample preparation was carried out for matrixes with water content higher than 20%, and the matrix was dried at 60°C ± 1°C (LABORMIM LP-402, Hungary) for 48 hours before sample grinding through a 1 mm grid (SM 100,   Retsch). 20 g of ground sample was weighed into a clean container (Falcon tube). Contamination of the sample with AFL B1 toxin (Biopure) was carried out to a maximum concentration of 40 ppb. e mycotoxin was extracted with 70/30 (v/v) (%) methanol/water extraction solution in a 1 : 5 to 1 : 10 ratio shaking with 360°Multi-Functional Tube Rotator (BIO-SAN) for 30 mins (orbital: 60/109 sec, reciprocal: 5°/5 sec, vibro: 5°/1 min). After settling and filtering the extract through Munktell-Ahlstrom 292 filter paper (5-8 μm), the pH of the filtrate was adjusted to pH 6.5 with 1M NaOH or 1M NaHCO 3 solution.

Detection of Aflatoxin B1 by AgraQuant ® Aflatoxin B1
ELISA Test. 200 μl of the conjugate solution was loaded into the dilution wells, and then 100 μl of each standard and sample were added to the dilution cavities. Next, the conjugate was mixed with the sample, and 100 μl from the dilution wells was transferred to the antibody-covered wells and incubated at room temperature for 15 minutes. Finally, the wells were washed with 250 μl of deionized water five times (a plate washer can be applied), and the washed wells were dried on a paper towel or aspirated with a plate washer.
100 μl of the substrate solution was added to the plate wells covered with the antibody and incubated at room temperature for 5 minutes (competitive ELISA: blue color is more intensive with lower AFL B1 concentration). e enzyme reaction was stopped with 100 μl stop solution (color change to yellow), reading the absorbance at 450 nm and 630 nm differential filter. We used a Synergy HTX multimode reader (BioTek) and Gene5 3.05 software (BioTek) for data collection.

Method of Evaluating the Results.
e dose-effect curve was prepared based on the OD values of the standards. Since the amount of AFL B1 is known in each standard, unknowns can be measured by interpolation based on this standard curve if the Log/Logit regression model is used to interpret the results. Calibration curves were prepared for AFL B1 and coefficients of determination (r 2 ) were calculated in Windows Office 365 Excel software. (1):

Recovery. Recovery calculation
where, R is recovery; C i is measured value from spiked feed; and C ref is measured value from spiked methanol.

Accuracy.
For intra-assay precision within one run, an 8-fold determination of the samples was performed. Mean and standard deviation was calculated (2). e relative standard deviation (RSD) as the absolute value of the coefficient of variation (CV) in (%) was also calculated. e aim was to keep the value below 15 RSD (%) (3).
Performance as the limit of detection (LOD), linear range, and reproducibility of the applied HPLC method was determined with Quality Control Material Aflatoxin B1 in corn Mid-level (Biopure) (n � 8). For AFL B1, the LOD was 0.01 ppb, and the linear ranges were up to 300 ppb. In addition, relative standard deviation (RSD (%)) as an absolute value of the variation coefficient was calculated and found under 10% in all cases.

ELISA Limit of Detection and Quantitation.
e quality of the assay was assured by the limit of detection (LOD), which was determined experimentally from the concentration of 8 blank matrix samples as the mean concentration of blank samples + 3-fold standard deviation of the concentrations of blank samples. e limit of quantification (LOQ) was determined experimentally from the concentration of 8 blank matrix samples as the mean concentration of blank samples + 9-fold standard deviation of the concentrations of blank samples.

Data Availability
e data are available on request.

Conflicts of Interest
e authors declare that they have no conflicts of interest.

Authors' Contributions
I. P. and T. P. conceptualized the study; T. P. elaborated the methods in the study; E. H., C. A., and T. P. validated the study; T. P. presented the formal analysis in the study; E. T. presented the resources in the study; I. P. performed the data curation; E. H. wrote and prepared the original draft; T. P. reviewed and edited this study; E. T. supervised the study; and I. P. helped in funding acquisition. All authors have read and agreed to the published version of the manuscript.