Reference samples
Meat samples (sized approximately 20 × 10 × 2 cm) from the diaphragm pillars were collected from 184 fattening pigs from 30 different pig herds between October 2016 and January 2017 at the slaughter line of three abattoirs located in an area with high pig density in the Northwest of Germany. The pigs were regularly delivered to the abattoirs at the end of the fattening period and sampling did not affect the release of carcasses for human consumption. The samples were frozen in plastic bags at minus 20 °C immediately after sampling and defrosted at 20 °C for collecting meat juice. Therefore, the plastic bags were hung up and a clamp was put underneath the meat, leaving just enough space on one side of the bag for the meat juice to drip to the bottom of the bag. This method was previously described by Meemken et al. [2] with an elastic plastic band instead of a clamp. After 12 h of defrosting, 6–12 mL meat juice from every bag could be transferred into plastic cups (Eppendorf AG, Hamburg, Germany) and stored at minus 80 °C until further analysis.
In parallel to the meat sampling, 184 blood samples from exactly the same pigs had been taken for the preceding study (ArrayTube platform with serum [5]). These blood serum samples were analysed with ten different ELISA tests (pigtype Toxoplasma Ab, pigtype Trichinella Ab, pigtype Yersinia Ab, pigtype Hepatitis E Virus Ab, pigtype Mycobacterium Ab, pigtype Swine Influenza Virus Ab, pigtype Salmonella Ab, pigtype PRRSV Ab (all Indical Bioscience GmbH, Leipzig, Germany), ID Screen Mycoplasma hyopneumoniae Indirect, ID Screen APP Screening Indirect serotypes 1–12 (both IDvet, Grabels, France)). A selection of 90 reference samples had been made for the preceding study, which optimally covered the measuring range of every ELISA test. The selection of reference samples was complemented with nine Trichinella spp. positive serum samples and individually matched meat juice samples from pig infection trials at the German National Reference Laboratory for Trichinella (German Federal Institute for Risk Assessment, Berlin, Germany), 20 T. gondii positive serum samples from MVZ Diamedis laboratory (MVZ Diamedis GmbH, Bielefeld, Germany) and seven T. gondii positive meat juice samples from infection trials at the Institute for Parasitology, University of Veterinary Medicine, Hannover, Germany. The additionally acquired seropositive samples for Trichinella and T. gondii were only used to characterise the respective Trichinella and T. gondii antigen on the microarray. The total number of negative and positive serum and meat juice samples for every antigen taken into account for the analysis of the different microarray platforms is displayed in the additional files (see Additional file 1).
Microarray production
Two different microarray platforms (ArrayTube, ArrayStrip) manufactured by Abbott (formerly Alere Technologies GmbH) Jena, Germany were used in this study. In both platforms, the same microarray chip as produced for the preceding study [5] was attached to the bottom of the reaction vial (see Figs. 1 and 2).
The antigens were coupled to the glass surface of the microarray chip by means of an epoxy layer. The spotting and manufacturing process were previously described by Ehricht et al. [7]. The following 12 antigens from three different manufacturers were spotted and covalently immobilised as ‘antigen-spots’ on the chip: T. gondii, Y. enterocolitica mix, Salmonella spp. ELISA mix, Salmonella spp. in-house mix, Trichinella spp., M. avium, Hepatitis E virus, Influenza A virus, PRRSV in-house mix (all Indical Bioscience GmbH, Leipzig, Germany), M. hyopneumoniae, A. pleuropneumoniae, (both IDvet, Grabels, France) and Y. enterocolitica Yop O:3 (Institut Virion-Serion GmbH, Wuerzburg, Germany). The antigens offered by Indical Bioscience and IDvet were exactly the same antigens as used in producing the respective ELISA tests. More information on the antigens, the different antigen concentrations that were spotted and the layout of the chip is published in the preceding study [5]. Most antigen concentrations were spotted in quadruplicate and the median of replicated spots was established as test outcome for every antigen concentration. Purified-pig IgG (BIOMOL GmbH, Hamburg, Germany) was spotted to confirm a correct binding of the conjugate and the substrate on every microarray chip.
Microarray test procedures
The test protocol for the analysis of serum on the ArrayTube platform is published in the preceding study [5]. In summary, antibodies that are present in the sample bind to the corresponding antigen spots during incubation time and all spots were antibodies are bound are detected by adding anti-pig-IgG-HRP conjugate, which is made visible by adding HRP-substrate.
In contrast to the test protocol with serum, the washing steps after sample incubation and after adding the conjugate had to be increased from three times to five times for applying meat juice on the microarray. The meat juice itself was centrifuged for 3 min at 4000 rpm immediately before preparing the sample dilution (1:2) from the meat juice supernatant. Different numbers of washing steps and different dilutions of meat juice were preliminarily tested.
For the analysis on the ArrayStrip platform, instead of 500 μL for pre-washing and 350 μL for all other washing steps, only 150 μL protein binding buffer were used in each washing step due to the smaller volume of the wells. Sample preparation for the ArrayStrip platform was identical to the ArrayTube platform: Blood samples had been centrifuged for 10 min at 2000 rpm on the day of sampling and the serum supernatant was diluted of 1:50. Meat juice was prepared as described before. Shaking of ArrayTubes and ArrayStrips was performed with horizontal thermoshakers (BioShake iQ, Quantifoil Instruments GmbH, Jena, Germany or PHMT Thermoshaker, Grant Instruments Ltd., Cambridge, United Kingdom). In order to aspirate liquids from a microarray, it is necessary to carefully approach the side of the tube with the tip of a pipette in order to avoid scratching the surface of the chip. Plastic transfer pipettes were used for this purpose on the ArrayTube platform and multi-channel pipettes on the ArrayStrip platform (see Figs. 1 and 2).
After aspirating the substrate from the microarrays, an image of every microarray was taken by the ArrayMate reading device (Abbott (Alere Technologies GmbH)). This device measures the intensity of staining from every spot on the microarray with a value between 0 (no signal, white spot) and 1 (maximum signal, black spot) as previously described [5]. Signal intensities between 0.1 and 0.7 are within the dynamic range of the test, a value below 0.1 cannot be assumed to show a correct antigen-antibody binding and a value above 0.7 indicates a color saturation of the spot [11]. Disruptive factors such as protein residues, scratches, dust or lint that are visible on the image, can result in an invalid measurement of one or several spots by the Iconoclust software (Abbot (Alere Technologies GmbH)) on the reading device.
Microarray analysis in different laboratories
Three laboratories were involved in this study: The laboratory of the Institute for Food Quality and Food Safety, University of Veterinary Medicine Hannover, Germany, the laboratory of LUFA Nord-West, Oldenburg, Germany (accredited service laboratory affiliated with the Chamber of Agriculture in Lower Saxony, Germany) and the accredited food and veterinary service laboratory LVL Lebensmittel- und Veterinaerlabor GmbH, Emstek, Germany. In the following, the aforementioned laboratories are referred to as laboratories A, B and C. First, 106 meat juice samples were analysed on the ArrayTube platform in laboratory A. This included 90 meat juice samples originating from the same pigs as the serum samples that had been used for the development of the ArrayTube platform in the preceding study [5], 9 Trichinella seropositive and 7 T. gondii seropositive meat juice samples. The analysis of the ArrayStrip platform comprised 95 meat juice and 95 serum samples. This number was determined by the 96-well frame on which 95 samples could be analysed at one time, together with one sample that only contained the sample diluent buffer ‘pigtype blue’. This sample served as a control for false positive signals on antigen spots. The 95 samples consisted of 88 paired serum and meat juice samples from the analysis of the ArrayTube platform plus 7 T. gondii positive meat juice and 7 T. gondii positive serum samples that had also been used on the ArrayTube platform (no paired samples). As the seropositive Trichinella spp. samples did not show positive signals on the ArrayTube platform with serum and meat juice, the Trichinella antigen spots were not considered functional and no further Trichinella spp. seropositive samples were analysed on the ArrayStrip platform. Aliquots from the 95 serum and meat juice samples were sent to laboratories B and C together with all necessary processing liquids and ArrayStrips from the same printing lot as used in laboratory A. Both laboratories were equipped with ArrayMate reading devices and laboratory personnel had received training for microarray analysis from Abbott (Alere Technologies GmbH) together with laboratory A. Laboratories B and C analysed the meat juice and the serum samples on the ArrayStrip platform and submitted the microarray data to laboratory A for statistical analysis.
Statistical analysis
Statistical analysis was performed with R version 3.6.1 [12] and Microsoft Excel 2010. To determine the accuracy of the antigens on the different platforms, receiver operating characteristic (ROC) curve analyses using the ELISA test results as reference were set up with the ‘pROC’ package [13] in R. Area under curve (AUC) confidence intervals were calculated by using the method by De Long [14]. Only antigens that had reached a minimum AUC value of 0.7 (moderate test accuracy [15]) for one of the spotted antigen concentrations were considered for further analysis. For these antigens, cut-off values were set as follows: First, the cut-off value was set to the maximum Youden Index [16]. If this resulted in a cut-off value below a signal intensity of 0.1, the minimum cut-off value 0.1 was chosen according to the dynamic range of the test. If setting the cut-off value to 0.1 implied a sensitivity or specificity below 0.6, the antigen concentration was excluded. Antigen concentrations that met these criteria were considered for agreement analysis between serum and meat juice as sample material. Therefore, Cohen’s kappa coefficients with 95% confident intervals were calculated with the ‘rel’ package [17] in R. In accordance with Landis and Koch [18] and Hunt [19], kappa values between 0.4 and 0.75 represent a fair to good agreement and those values higher than 0.75 an excellent agreement. In addition, Bland-Altman plots [20] were set up for a quantitative comparison between the measured signal intensities for paired serum and meat juice samples. Cohen’s kappa coefficients were also calculated for the level of agreement between the three laboratories.