Anal Bioanal Chem DOI 10.1007/s00216-015-8526-4

RESEARCH PAPER

Development of a rapid multiplexed assay for the direct screening of antimicrobial residues in raw milk Terry F. McGrath & Laura McClintock & John S. Dunn & Gregory M. Husar & Michael J. Lochhead & Ronald W. Sarver & Frank E. Klein & Jennifer A. Rice & Katrina Campbell & Christopher T. Elliott

Received: 14 November 2014 / Revised: 19 January 2015 / Accepted: 28 January 2015 # Springer-Verlag Berlin Heidelberg 2015

Abstract Antimicrobial residues found to be present in milk can have both health and economic impacts. For these reasons, the widespread routine testing of milk is required. Due to delays with sample handling and test scheduling, laboratory-based tests are not always suited for making decisions about raw material intake and product release, especially when samples require shipping to a central testing facility. Therefore, rapid on-site screening tests that can produce results within a matter of minutes are required to facilitate rapid intake and product release processes. Such tests must be simple for use by non-technical staff. There is increasing momentum towards the development and implementation of multiplexing tests that can detect a range of important antimicrobial residues simultaneously. A simple in situ multiplexed planar waveguide device that can simultaneously detect chloramphenicol, streptomycin and desfuroylceftiofur in raw dairy milk, without sample preparation, has been developed. Samples are simply mixed with antibody prior to an aliquot being passed through the detection cartridge for 5 min before reading on a field-deployable portable instrument. Multiplexed calibration curves were produced in both buffer and raw milk. Buffer curves, for chloramphenicol, streptomycin and Published in the topical collection on Hormone and Veterinary Drug Residue Analysis with guest editors Siska Croubels, Els Daeseleire, Sarah De Saeger, Peter Van Eenoo, and Lynn Vanhaecke. T. F. McGrath (*) : L. McClintock : K. Campbell : C. T. Elliott Institute for Global Food Security, School of Biological Sciences, Queen’s University, David Keir Building, Stranmillis Road, Belfast BT95AG, UK e-mail: [email protected] J. S. Dunn : G. M. Husar : M. J. Lochhead MBio Diagnostics Inc, 5603 Arapahoe Avenue, Suite 1, Boulder, CO 80303, USA R. W. Sarver : F. E. Klein : J. A. Rice Neogen Corporation, 620 Lesher Place, Lansing, MI 48912, USA

desfuroylceftiofur, showed linear ranges (inhibitory concentration (IC)20–IC80) of 0.1–0.9, 3–129 and 12–26 ng/ml, whilst linear range in milk was 0.13–0.74, 11–376 and 2– 12 ng/ml, respectively, thus meeting European legislated concentration requirements for both chloramphenicol and streptomycin, in milk, without the need for any sample preparation. Desfuroylceftiofur-contaminated samples require only simple sample dilution to bring positive samples within the range of quantification. Assay repeatability and reproducibility were lower than 12 coefficient of variation (%CV), whilst blank raw milk samples (n=9) showed repeatability ranging between 4.2 and 8.1 %CV when measured on all three calibration curves. Keywords Rapid multiplexed screening . Veterinary drug residue milk . No sample preparation . Planar waveguide . On-site screening

Introduction The use of antimicrobial compounds in food-producing animals can lead to the selection and promotion of drug-resistant pathogens. If the food becomes contaminated during processing, then consumption can lead to antibiotic-resistant infections [1]. Furthermore, consumption of meat-containing antibiotic residues can lead to drug sensitisation, allergic reactions and changes to the balance of gut flora [2–4]. Beyond the health risks, there is also a substantial financial risk to food sectors, e.g. antibiotic residues in milk will inhibit the culturing process in cheese and yoghurt production causing a drop in production or ultimately the requirement to dispose of large volumes of contaminated raw materials followed by intense purging processes [5, 6]. Due to the large number of samples and the number of contaminants requiring screening, multiplexing has become a key goal in both laboratory and

T.F. McGrath et al.

on-site testing methods. Multiplexing serves to reduce operator time and reduce costs and the amount of equipment required compared to running tests individually [2]. Multiplexing can be performed within drug families, by using generic binders, or cover several families, by use of multiple binding sites (or channels in laboratory-based tests) [2]. The present study focuses on showing the proof-of-concept and development of an on-site multiplexed test for the detection of antimicrobial residues in raw milk. Desfuroylceftiofur (the marker residue for ceftiofur), streptomycin and chloramphenicol were selected for inclusion due to their importance in terms of monitoring needs in the dairy industries globally. Europe set maximum residue limits (MRLs) for ceftiofur (sum of all residues retaining the beta-lactam structure expressed as desfuroylceftiofur) and streptomycin at 100 and 200 ng/ml in milk, respectively [7], whilst chloramphenicol is a banned substance and has a minimum required performance limit (MRPL) of 0.3 ng/ml [7, 8]. Within Europe, there were 30,748 targeted milk samples tested in 2012, in statutory laboratories. There were 27 non-compliant results (0.09 %), of which one was chloramphenicol (group A6) and nine where antibacterials (group B1) [9]. These figures only cover the official testing plans and fail to reflect the large number of tests carried out by producers or food processing companies who need to perform tests to insure quality control and guarantee that their raw materials comply with product intake specifications. All of these industry-driven tests require rapid and reliable methods to maintain production levels and keep manufacturing plants contaminant free. The majority of these tests will be performed on-site instead of in a laboratory due to time pressures. The late reporting of contamination can result in contaminated product entering the food chain. Commercially, there are several companies providing lab-based testing solutions such as Charm Sciences Inc. (Charm II [10]), whilst others, e.g. Idexx Laboratories Inc. (Snap [11]), Charm Sciences Inc. (ROSA [12]), Neogen Corporation (Penzyme and Betastar [13]), DSM (Delvotest [14]) and Unisensor (4Sensor and Trisensor [15]), provide on-site antibiotic drug residue testing kits. The majority of these on-site tests currently focus on individual families of drug residues or cover one other family therefore requiring the use of multiple testing products. However, some products of note are the 4Sensor and Trisensor test kits from Unisensor that can detect across at least three families of antimicrobials using a dipstick format. Subjective results can be obtained by visual interpretation of the strip whilst incubating in a heating block; however, the use of an additional sensor reader will yield objective results [15]. Also of note is the Charm ROSA QUAD Test, which is a lateral flow test for the detection of four different antimicrobial families. The manufacturer suggests that the use of an additional incubator will assure better line formation for visual interpretation and more accurate quantitation using their reader [12]. All providers claim results can be generated within the region

of 5–10 min [11–15]. A topic search for the last 5 years, using the terms Bdrug residues milk^, in Web of Science produced approximately 680 hits. Within this search, there were examples of multiple residue detection, from different families, within milk. However, almost all require significant sample preparation and complicated laboratory equipment that make them unsuitable for rapid on-site testing [16–21]. Some lateral flow devices were also described, but the focus remained on within family multiplexing or inclusion of one other group [22, 23]. In the present study, a SnapEsi reader with single use disposable cartridge, from MBio Diagnostics Inc., was used for multiplexing three different antimicrobial measurements in buffer and milk samples. Milk samples were analysed without the need for any sample preparation. Both the reader and cartridges employed have been described in detail elsewhere [24–26]. Briefly, the detection principle is based on the measurement of far-red fluorescence. Laser light is coupled to an injection moulded plastic waveguide via an integrated lens moulded as part of the waveguide. This light propagates down the length of the waveguide under total internal reflection forming a uniform illumination field. At an interface between a liquid medium and the waveguide surface, an evanescent field is produced which penetrates into the liquid medium. This field decays rapidly and so only fluorescent dyes held in close proximity to the surface will undergo excitation [24–26]. Signal level is measured in terms of arbitrary units (arb. U). Immunochemical reagents can be accurately immobilised to the surface using nano-spotting technology, thus providing multiplexing capability, whilst their binding partners can be fluorescently labelled and used within the liquid medium. The SnapEsi reader can detect and quantify levels of fluorescence for individual spotted reagents within an array. The reader itself has a very small footprint of approximately 24×11×17 cm and is powered, as well as communicates, via a USB interface to a laptop. These features of the reader allow for portability in analysis having both laboratoryand field-based applications. Lochhead et al. provided proof of concept for human clinical diagnostic applications, simultaneously measuring host antibody response to eight antigens from HIV-1, hepatitis C virus and Treponema pallidum [26]. Meneely et al. then described a method for the detection of a wide range of paralytic shellfish toxins in algal culture. An elegant broad-spectrum assay was achieved by using a highly cross-reactive antibody to detect multiple structurally related compounds from the same family. Their assay was also three steps, but time was reduced to 15 min [24]. Devlin et al. also describe a three-step approach for the detection of a structurally related family of freshwater toxins (microcystins) in both water and algal cultures also using a highly cross-reactive antibody with an assay time of 15 min [24]. The work described in the present study has made a substantial advance on the previous work in that it provides, for

Development of a rapid multiplexed assay

the first time, data on the effects on performance when key assay parameters are manipulated. This information provides an important basis for future assay development on this platform. Advances are also presented with the simultaneous detection of multiple antimicrobial residues, each structurally unrelated and from a different antimicrobial family, using mixed antibodies. Identification of the particular residue present in a sample can clearly be ascertained based on the location of the spot producing the signal. This identification was performed in raw milk without the need for any sample preparation. The general assay protocol was reduced from a threestep to a two-step procedure, (a) mixing of labelled primary antibody with sample and (b) addition of sample to cartridge. The analysis time was further reduced initially to 7.5 min then to 5 min.

Materials and methods Reagents Raw negative milk was obtained from the Agri-Food and Biosciences Institute, Belfast, UK. Phosphate-buffered saline (PBS) ×10 was obtained from Fisher Scientific, Loughborough, UK. Aliquots of chloramphenicol-IgG (8.5 mg/ml), desfuroylceftiofur-bovine serum albumin (BSA) (2.51 mg/ml), desfuroylceftiofur (D-CEFT), mouse anti-ceftiofur antibody (2.91 mg/ml), rabbit anti-chloramphenicol antibody (2.09 mg/ml), rabbit anti-streptomycin antibody (2.69 mg/ml) and streptomycin-BSA (6.7 mg/ml) were supplied as a gift by Neogen Corporation, Lansing, MI, USA. Alexa Fluor 647 conjugated BSA was obtained from Life Technologies, Paisley, UK. Cartridge components and the SnapEsi instrument were supplied by MBio Diagnostics, Inc., Boulder, CO, USA. sciFLEXARRAYER S5 and associated software were obtained from Scienion AG, Berlin, Germany. BSA, casein, chloramphenicol (CAP), dibasic sodium phosphate, monobasic sodium phosphate, sodium chloride, streptomycin sulphate (STREP) and Tween-20 were purchased from Sigma-Aldrich, Dorset, UK. DyLight 650 antibody labelling kit was purchased from Thermo Scientific, UK. Performance and assay characteristics Labelling antibodies Primary antibodies were labelled using a DyLight 650 antibody labelling kit. Briefly, 1 mg of antibody was prepared in 500 μl borate buffer (50 mM) before being added to a DyLight reagent vial. The mixture was left to react at room temperature, in the dark, for 1 h. It was then passed through purification resin before antibody concentration and dye-to-protein ratio determined according to the kit manufacturer’s instructions.

Standardised cartridge build process Protein-conjugated targets were diluted in sterile filtered printing buffer (100 mM sodium phosphate, 50 mM sodium chloride, 100 μg/ml BSA, 0.005 % Tween-20, pH 8.0), to the required concentration, and the Scienion sciFLEXARRA YER S5 used to spot 25 nl of each onto planar waveguides in a 2×22 array format (Fig. 1). Typically, two columns were used for each analyte giving four spots (Fig. 1). BSA (647dye-labelled; 1 μg/ml) was positioned at each of the four corners of the array to aid automatic array alignment within the SnapEsi reader. Print buffer or labelled BSA was used to spot the redundant locations within the array. All spotting was carried out at room temperature and 65 % humidity. Waveguides were left at 65 % humidity for 1 h before being stored at 25 °C and 30 % humidity overnight. Cartridge gaskets were attached to the waveguides before being strenuously agitated in rinse solution (0.5 % casein prepared in 1× PBST (0.05 % Tween20)) for 30 s then left to soak for 5 min. Waveguides were then rinsed in deionised water and spun dry at 87×g for 5 min before insertion into the cartridge housing where the gasket produced a 5-mm-wide, 50-mm-long and 0.145-mm-high flow channel. The cartridge build was completed with the addition of an absorbent pad and top cover. When liquid is added to the cartridge inlet, it moves under capillary action, supported by gravity, through the flow channel with the absorbent pad ensuring that this flow is maintained for the whole sample. Assay setup In the assay, prepared antibody concentrations were mixed 1:1 with appropriate standards/samples before 150 μl of the mixture was pipetted onto a cartridge. Readings were then taken at a fixed time point after sample addition. Raw signals were exported into Microsoft Excel for statistical analysis. Calibration curves were generated using a four-parameter fitting function within BIAEvaluation Software V4.1 before curve characteristics were evaluated using the same software. Repeatability, intra-batch reproducibility and inter-batch reproducibility Repeatability of signal was measured by determining the coefficient of variation (%CV) for the signal produced by the labelled BSA spots (1 μg/ml; n=10) on an individual cartridge after they had been wetted with sample buffer (1× BPS). Intrabatch reproducibility was measured by determining the %CV for signal from all labelled BSA spots across a batch (n=24) of cartridges. Inter-batch reproducibility was measured by determining the %CV for signal from all labelled BSA spots across several different production batches (n=3).

T.F. McGrath et al.

Fig. 1 Actual image showing 2×22 array format, as produced by spot recognition element within the SnapEsi control software, when 25 nl of 1 μg/ml labelled-BSA has been spotted on the surface. Labels indicate orientation of lens and sample inlet ends as well as typical locations of analyte spots

The influence of time on signal levels The influence of reading time was assessed by analysing a series of cartridges at two different time points after sample addition. Labelled anti-desfuroylceftiofur antibody was diluted to 11.3 μg/ml with sample buffer and mixed 1:1 with DCEFT standards, prepared in sample buffer, ranging from 0 to 10 ng/ml. Of these solutions, 150 μl was then added to individual cartridges that had previously been spotted with 25 nl of desfuroylceftiofur-BSA (10 μg/ml), in a manner that would allow readings to be taken 7.5 min after sample addition. Readings were then repeated at 120 min after sample addition. The influence of antibody concentration and spot concentration Antibody concentration and spot concentration were manipulated whilst testing inhibition for each of the model systems, D-CEFT, STREP and CAP, as individual assays. The typical assay setup, described above, was followed. Multiplexed calibration curve in buffer Eight waveguides were spotted, with 1 μg/ml dye-labelled BSA for corner markers, 400 μg/ml desfuroylceftiofur-BSA, 25 μg/ml streptomycin-BSA and 100 μg/ml chloramphenicolIgG, as described above and in Fig. 1. Cartridges were then built in the described fashion. Eight calibration standards were prepared 3× desired concentration for each of the three analytes then mixed 1:1:1 to produce the working standards. Antibodies were prepared 3× desired concentration and mixed, in a similar fashion, to provide the working antibody dilution. This was then mixed 1:1 with each of the eight prepared working standards before 150 μl of each solution was added to a prepared cartridge and fluorescence measured 7.5 min after sample addition. Multiplexed calibration curve in milk Waveguides, cartridges and antibody solutions were prepared in the same way as for the buffer calibration curves. Raw milk, free of drug residues, was spiked with the appropriate concentration of each analyte to produce eight samples containing

multiple drug residues. These milk samples were used as mixed standards. The assay procedure used for the buffer calibration curves was followed to produce multiplexed calibration curves in milk.

Assay optimisation in buffer Several series of eight cartridges were prepared using the standardised procedure described above. Various spotting concentrations of protein conjugated analytes were used. Using the typical assay setup, described above, multiplexed calibration curves using mixed standards and antibodies were ran in buffer at various antibody concentrations and reading times. Mixed calibration curves were initially ran between 0 and 10 ng/ml for CAP, 0 and 800 ng/ml for STREP and 0 and 200 ng/ml for D-CEFT, to take into account their MRL/MRPL tolerances. Stronger standards were removed when the fitting function failed.

Optimised calibration curves in milk with spiked sample analysis Waveguides were prepared with 25 nl labelled-BSA (1 μg/ml), chloramphenicol-IgG (400 μg/ml), desfuroylceftiofur-BSA (200 μg/ml) and streptomycin-BSA (6.25 ng/ml) before the cartridges were built as described previously. Antibodies were mixed to give an antibody solution containing antichloramphenicol (1.8 μg/ml), anti-streptomycin (54.4 μg/ml), labelled anti-desfuroylceftiofur (16.3 μg/ml) and unlabelled anti-desfuroylceftiofur (7.3 μg/ml) antibodies. Raw milk, free of drug residues, was spiked with the appropriate concentration of each analyte to produce eight samples containing multiple drug residues for use as calibrants. Further 12 milk samples were prepared as controls: three spiked with CAP only, at 1.5 times MRPL, MRPL and half MRPL; three spiked with STREP only, at 1.5 times MRL, MRL and half MRL; and further three spiked with D-CEFT only, at 1.5 times MRL, MRL and half MRL. The procedure, described previously, was used to produce multiplexed calibration curves in milk before analysis of the prepared spiked milk samples. All sample readings were taken 5 min after sample addition.

Development of a rapid multiplexed assay

Results and discussion Performance and assay characteristics Labelling antibodies Antibody concentration and dye-to-protein ratio were determined for labelled anti-desfuroylceftiofur, anti-streptomycin and anti-chloramphenicol antibodies. Concentration and dyeto-protein ratio, in parentheses, were found to be 1.63 mg/ml (1.9:1), 1.96 mg/ml (2.4:1) and 1.82 mg/ml (2.4:1), respectively. Repeatability, intra-batch reproducibility and inter-batch reproducibility Repeatability of signal produced by labelled-BSA control spots (n=10) on an individual cartridge was found to range between 3.2 and 11.1 % CV for all cartridges within a single batch with an average %CV of 6.4. The average signal for all labelled-BSA spots within a batch of cartridges (n=24) was determined to be 82 arb. U, standard deviation 7.9, with an intra-batch reproducibility of 9.6 %CV, whilst the inter-batch reproducibility (n=3) was 9.5 %CV. The influence of time on signal level Signal range and inhibitory concentration (IC) values, at various points, on a calibration curve, can be used as indicators during assay development to illustrate the effects of changes in the assay parameters relative to the overall performance of the final assay. IC values are a function of the obtained signal range. IC20–IC80 values are used to define the dynamic range [27–29], whilst IC10–IC90 has been used to describe the usable concentration range [30]. Figure 2 shows that as time increased, the signal level also increased. This effect is most notable at the lower concentrations of standard, e.g. at standard concentrations lower than Fig. 2 Calibration curves showing the effect of assay reading time at fixed antibody concentration, 11.3 μg/ml for DCEFT assay. The same trend was shown for the STREP and CAP assays. Error bars are signal standard deviation (n=4)

7.5 ng/ml, the change in signal level with increased time is more pronounced. Interestingly, although the signal range increases significantly with time, the IC20, IC50 and IC80 remain within the same order of magnitude as the readings at 7.5 min. Increasing signal with time is best explained by mass transfer, with diffusion of the dye-labelled detection reagent to the capture spot still in a kinetic regime at 7.5 min during this static (no fluid movement) assay. Whilst the longer times give more robust signal as equilibrium is approached, the shorter times provide significant advantages for practical assay implementation. Provided that assay times can be controlled, accurate quantitative results can be generated with a kinetic assay. The observation that the IC20, IC50 and IC80 were similar at 7.5 and 120 min, despite significant signal differences, supports this statement. In addition to tuning assay reagent concentrations to increase signal, further assay optimization focused on determining the minimum time required to generate stable IC20, IC50 and IC80 values. It is noted here that MBio Diagnostics’ next-generation cartridge and reader add the ability to run accurately timed assays without any user intervention. The influence of antibody concentration and immobilised antigen concentration Table 1 outlines the effects on signal intensity, range and inhibition for each of the single assay systems, when antibody concentration and spot concentration are varied. Readings were taken at 7.5 min after sample addition. Desfuroylceftiofur assay Anti-desfuroylceftiofur antibody was originally diluted to 11.3 μg/ml and at this concentration gave a very low signal range, 7.6 arb. U. When the antibody concentration was doubled, the range improved to 35 arb. U, as did the level of inhibition, from 58 to 83 %. This improvement in inhibition is contra to what would normally be observed [31] and may purely be a result of the extremely low signal range for the 11.3-μg/ml antibody concentration.

T.F. McGrath et al. Table 1 The effects of altering antibody concentration and spotted antigen concentration on signal levels, range and inhibition for each of the model systems ran as individual assays Model

Antibody concentration (μg/ml)

25 nl of antigen concentration spotted on surface (μg/ml)

Signal: negative sample

Signal: positive sample

Range

Inhibition (%)

D-CEFT

11.3 22.6 22.6 22.6 22.6 22.6 22.6 22.6 27.2 27.2 27.2 27.2 27.2 27.2 27.2 13.6 9.1 6.8 25.3 25.3 25.3 25.3 25.3 25.3 25.3

10 10 25 50 100 200 400 800 10 25 50 100 200 400 800 25 25 25 10 25 50 100 200 400 800

13.1 42.0 59.2 88.5 132.7 124.6 159.2 172.5 251.6 401.9 442.4 476.3 497.7 531.0 438.4 84.2 33.6 16.5 72.5 110.0 221.3 267.9 340.0 354.8 412.6

5.5 7.1 10.7 16.9 29.7 21.0 37.4 47.8 109.4 140.1 187.7 186.1 207.3 215.6 188.3 34.5 11.0 8.1 5.8 8.7 11.6 15.1 17.9 19.9 28.3

7.6 35.0 48.5 71.5 103.0 103.6 121.8 124.7 142.2 261.9 254.7 290.2 290.5 315.4 250.2 49.7 22.6 8.4 66.7 101.3 209.7 252.8 322.0 334.9 384.3

58.0 83.2 81.9 80.8 77.6 83.1 76.5 72.3 56.5 65.2 57.6 60.9 58.4 59.4 57.1 59.0 67.2 51.0 92.0 92.1 94.8 94.4 94.7 94.4 93.1

STREP

CAP

Readings were taken at 7.5 min after sample addition. Positive sample for D-CEFT was 10 ng/ml, whilst 100 ng/ml was used for STREP and CAP

Table 1 shows that, for D-CEFT, as the concentration of solution spotted increased so did the signal level and signal range, e.g. signal levels for negative samples increased from 42 to 172 arb. U, whilst for positive samples, it increased from 7.1 to 47.8 arb. U as the antibody concentration was kept constant but the spotted concentration increased. The signal range also increased from 35 to 124 arb. U under the same conditions. However, inhibition was generally unaffected at the fixed antibody concentration remaining around 79 %. Optimum range was reached when spotting with 25 nl of 100 μg/ml. Using stronger spotting solutions did not greatly increase the range, e.g. an eightfold increase in spot concentration only increased the range by 22 arb. U from 103 arb. U at 100 μg/ml to 125 arb. U at 800 μg/ml spotted concentration. Streptomycin assay The effects of changing spot concentrations for the STREP assay again showed that as the spot concentration increased, the signal intensity recorded also increased, e.g. signal levels for negative samples increase from

252 to 438 arb. U, whilst for positive samples, they ranged from 109 to 188 arb. U. There was also very little change in signal range for spot concentrations above 25 μg/ml, average range was 277 with a %CV of 9.2 (n=6). Inhibition levels obtained (typical value 59 %) were also lower than those for D-CEFT (typical value 81 %). Further dilution of the antistreptomycin antibody did not affect the inhibition level to any great extent as it remained around 59 %. It was also discovered that increasing the concentration of the positive between 100 and 400 ng/ml did not increase the percent inhibition above 68 % (data not shown). From these data, it seems that some degree of binding of the labelled antibody to the conjugated protein was occurring. Chloramphenicol assay The anti-chloramphenicol antibody exhibited a similar trend, in signal levels and inhibition, as the anti-desfuroylceftiofur antibody with increasing spot concentration, e.g. signal levels for negative samples, increased from 73 to 413 arb. U, whilst for positive samples, it went from 5.8 to 28.3 arb. U.

Development of a rapid multiplexed assay

Multiplexed calibration curve in buffer Multiplexed calibration curves (for D-CEFT, STREP and CAP) were produced in buffer using 22.6, 27.2 and 25.3 μg/ml final antibody concentrations (Fig. 3). The D-CEFT calibration curve was produced between 0 and 20 ng/ml standards and gave a signal range of 123 arb. U (Fig. 3a). Inhibition was around 63 % for the highest concentration standard used. In the same run, a STREP calibration curve was produced between 0 and 50 ng/ml and gave a signal range of 217 arb. U (Fig. 3b). The highest inhibition achieved was around 45 %, although there was a clear plateau across the higher concentration standards. The CAP calibration curve, between 0 and 100 ng/ml (Fig. 3c), had a signal range of 589 arb. U and reached 92 % inhibition. Calibration point standard deviations (n=4) were higher in the STREP calibration curve (average SD 27.6 %) than the DCEFT (average SD 5.1 %). When the signals generated for STREP on both spotted columns in the cartridge were plotted as individual curves, instead of combined to make one curve, then standard deviation improved to an average SD of 14.2 % (Fig. 3d). Based on cartridge architecture (Fig. 1), where samples enter at the end furthest from the lens and travel towards it, and the fact that signal levels were consistently higher for column 16 than column 15 (Fig. 3d), it is suggested that there

is an antibody depletion effect on column 15. This depletion does not have a large effect on IC20, IC50 and IC80 values which remain around 2.1, 4.8 and 9.5 ng/ml, respectively (Fig. 3d). Multiplexed calibration curve in milk Figure 4 shows multiplexed calibration curves in milk produced under the same conditions as the calibration curves in buffer. The D-CEFT calibration curve gave a signal range of 200 arb. U (Fig. 4a). Inhibition was approximately 74 % for the highest concentration standard used. In the same run, the STREP calibration curve gave a signal range of 372 arb. U (Fig. 4b). The highest inhibition achieved was approximately 48 %. The CAP calibration curve (Fig. 4c) had a signal range of 886 arb. U and reached 91 % inhibition. When comparing inhibition levels produced by the highest standard in both buffer and milk, the D-CEFT is the only curve that has moved, to any great extent, with inhibition changing from 63 % in buffer to 74 % in milk for the same standard concentration. When comparing the IC50 values for buffer and milk curves, the STREP and CAP values remained constant, approximately 4.8 and 8.9 ng/ml, respectively, whilst the value for DCEFT almost halved going from 4.4 ng/ml in buffer to 2.4 ng/ml in milk. In fact, for D-CEFT, the whole curve appears to have shifted left towards lower concentrations with

a

b

c

d

Fig. 3 Multiplexed calibration curves. a D-CEFT calibration curve in buffer. b STREP calibration curve in buffer. c CAP calibration curve in buffer. d STREP calibration curve in buffer showing curves produced by

the two different rows of spots. For a–c, each point is an average of four spots, whilst for d, each point is an average of two spots. The error bars represent standard deviation

T.F. McGrath et al. Fig. 4 Multiplexed calibration curves in milk. a D-CEFT calibration curve. b STREP calibration curve. c CAP calibration curve. Each point is an average of four spots. The error bars represent standard deviation

a

b

c

IC20, IC50 and IC80 values all being lower than their buffer counterparts. For STREP and CAP, the increased signal range slightly increased the linear range of the calibration curve with IC20 values slightly lower in milk and IC80 values higher in milk compared to buffer. Although calibration curves have been produced for the three analytes simultaneously, the curves were not fit for purpose since the analyte MRLs/MRPL

did not fall within the calibration range. Further assay optimisation was required. Assay optimisation in buffer Table 2 shows the results for multiplexed assay development in buffer. Although the work was performed in a multiplexed

Development of a rapid multiplexed assay Table 2

Calibration curve characteristics for multiplexed assay optimisation for D-CEFT, STREP and CAP in buffer

Model

Assay no.

Reading time (min)

Ab conc (μg/ml)

Spot (25 nl of × μg/ml)

Signal range

IC20 (ng/ml)

IC50 (ng/ml)

IC80 (ng/ml)

CAP

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9

7.5 7.5 7.5 7.5 7.5 7.5 7.5 7.5 5 5 5 5 5 5 5 7.5 7.5 7.5 7.5 7.5 7.5 7.5 7.5 5 5 7.5 7.5 7.5 7.5 7.5 7.5 7.5 7.5 5

100 50 25 12.5 100 100 100 100 100 100 100 200 400 400 400 25.0 12.5 6.25 3.125 6.25 6.25 1.00 1.00 6.25 6.25 400 200 100 50 25 10 5 1 100

561.0 354.1 165.2 100.7 630.4 533.6 301.9 267.1 214.2 154.1 127 177 83.4 24 55.8 496.3 415.9 340.3 161.6 350.8 552.6 72.3 107.2 436.6 504.4 327.6 288.3 232.1 175.6 117.3 99.5 35.5 19.4 112.8

2.5 2.4 2.4 2.9 2.7 1.6 1.3 0.4 0.5 0.3 0.4 0.4 0.1 0.02 0.2 1.1 0.8 1.0 3.7 1.6 1.6 2.3 1.1 3.1 9.7 2.1 2.3 1.8 2.1 1 1 1.3 0.01 3.1

4.7 4.8 4.9 4.9 5 3 2.4 1.4 1.2 0.7 0.8 0.8 0.4 0.09 0.3 5.9 6.0 13.6 23.5 16.8 17.7 22.4 14.7 21.1 39.1 3.8 3.8 3.4 6.7 2.5 2.4 2.5 0.3 5.4

8.9 9.4 9.8 8.1 9.6 5.8 4.7 2.8 2.7 1.7 1.6 1.6 0.9 0.4 0.4 24.1 36.2 120.7 123.5 116.9 130.2 170.6 128.6 129.1 144.4 6.9 6.0 6.1 6.3 6.5 5.3 5.1 4.5 9.5

10

5

100

90.8

11.9

17.1

24.4

11

5

100

142.4

7.2

11.0

16.4

12

5

100

144.1

4.6

7.2

11.1

13

5

200

66.0

12.2

17.8

26.0

14

5

400

60.6

11.9

16.4

22.0

15

5

200

64.6

5.8

10.2

18.0

16

5

12.6 12.6 12.6 12.6 37.9 9.1 6.1 4.5 4.5 3 2.3 2.3 1.8 1.5 2.3 27.2 27.2 27.2 27.2 27.2 54.4 27.2 54.4 54.4 98.0 22.6 22.6 22.6 22.6 22.6 22.6 22.6 22.6 16.3 labelled 7.3 unlabelled 16.3 labelled 29.1 unlabelled 20.4 labelled 14.6 unlabelled 20.4 labelled 9.7 unlabelled 16.3 labelled 7.3 unlabelled 16.3 labelled 29.1 unlabelled 16.3 labelled 14.6 unlabelled 16.3 labelled 9.7 unlabelled

200

59.0

9.3

9.7

10.1

STREP

D-CEFT

IC90 (ng/ml)

43.5 77.4 287.2 264.5 252.1 288.7 411.6 300.7 300.1 284.6

Assays with same assay no. were multiplexed on the same cartridges. Antibody concentrations are those on mixing of all antibodies in the multiplexed experiment before 1:1 mixing with samples. Spot concentration is the concentration of solution used to spot with

T.F. McGrath et al.

fashion, it is easier to describe the effects by dealing with the model assays on an individual basis. In relation to the CAP results (Table 2), with a reading time of 7.5 min, antibody concentration fixed at 12.6 μg/ml and spot concentrations decreasing (assay nos. 1–4), the expected trend of decreased signal range as spot concentration decreases was observed with signal ranges going from 561 to 101 arb. U as spot concentration goes from 100 to 12.5 μg/ml. Interestingly, there was no real change to IC20, IC50 and IC80 values which remained around 2.5, 4.8 and 9.1 ng/ml, respectively. Comparing CAP assay no. 5 to no. 1, where the antibody concentration had tripled in no. 5 whilst the spot concentration and timing remained the same, signal range increased, from 561 to 630 arb. U, as expected. However, there were still no changes on other curve characteristics with IC20, IC50 and IC80 values remaining around 2.6, 4.9 and 9.3 ng/ml, respectively. It was only when the spot concentration was kept high enough not to be a limiting factor and the antibody concentration decreased that changes to IC20, IC50 and IC80 values are observed (assays nos. 5–8), where IC values continually decreased from 2.7, 5 and 9.6 ng/ml to 0.4, 1.4 and 2.8 ng/ml, respectively, with the MRPL of 0.3 ng/ml almost detectible at the IC20 position. As previously observed, when the antibody concentration was decreased, there was a drop in signal range, e.g. from 630 to 267 arb. U as antibody concentration dropped from 37.9 to 4.5 μg/ml. Comparing CAP assays nos. 8 and 9 where the only difference was reading time, assay no. 9 had the reading that was 2.5 min earlier than no. 8, there was a small decrease in signal range (53 arb. U), as expected from the findings of the timing experiment. Based on the observations with assay nos. 5–8, where antibody concentration was decreased, and looking at CAP assays nos. 9– 11, where antibody concentration was again decreasing but using a shorter reading time, we would have expected to see larger decreases for the IC20, IC50 and IC80 values, but the drop appears to have levelled out, especially between assay no. 10 and no. 11 where IC20, IC50 and IC80 values remained around 0.3, 0.7 and 1.7 ng/ml, respectively. The signal range continued to decrease, going from 214 arb. U for assay no. 9 to 127 arb. U for assay no. 11 where antibody concentration was lower. In an attempt to continue to shift the curve left whilst keeping a high signal range, the spotted concentration was increased in assay no. 12. Comparing these curve indicators to no. 11, the signal range had increased from 127 to 177 arb. U, but the IC20, IC50 and IC80 values were unchanged at 0.4, 0.8 and 1.6 ng/ml, respectively. Assay no. 13, where the spotted concentration was further doubled but the antibody concentration reduced, does show changes to the IC20, IC50 and IC80 values which decreased to 0.1, 0.4 and 0.9 ng/ml, respectively, but the signal range had dropped below 100 arb. U. Chloramphenicol’s MRPL was sitting close to the IC50 value indicating that this curve could potentially be used to screen for CAP at levels around the permitted level. Assay no. 14 and

no. 15, where antibody concentration was further reduced, all produced curves that could detect CAP at the MRPL, but the signal range was greatly reduced. With regards to the STREP results (Table 2), the initial curve needed to be desensitised to pick up the streptomycin MRL of 200 μg/ml. In assay nos. 1–4, the antibody concentration was kept high but spotted concentration was reduced. As expected, as the surface concentration decreased so did the signal range which went from 496 to 162 arb. U as the solution used to spot the surface went from 25 to 3.125 μg/ml. It was not until assay no. 3, where the surface was spotted with 25 nl of 6.25 μg/ml, that any effect on IC20, IC50 and IC80 values was observed when they shifted to 1, 13.6 and 120.7 ng/ml, respectively. These values further increased in assay no. 4 which used 3.125 μg/ml solution to spot the surface. However, these changes did not desensitise the curve sufficiently to bring the MRL into range. In an attempt to evaluate where the MRL would be positioned on the STREP curve, it was decided to add an additional curve indicator, IC90, for STREP. IC90 values for assay no. 3 (287 ng/ml) and no. 4 (265 ng/ml) indicate that levels around the MRL could potentially be detected. Further attempts were made to try and desensitise the curve and get the MRL within the linear range. The most useful curve from the first four assays would appear to be no. 3 with a signal range above 300 and an IC90 above the MRL. This was repeated in assay no. 5, with similar findings before assay nos. 6–8 attempted to improve on those results by increasing antibody concentration and decreasing spot concentration. Assay no. 6 pushes up the signal range to 553 arb. U and maintained curve indicators, whereas decreasing the spot concentration in no. 7 caused the range to drop to 72 arb. U but would appear to bring the MRL close to the linear range, as indicated by an IC80 of 171 and IC90 of 412 ng/ml. Assay no. 8 shows that even with a very low spot concentration, the signal range can be further increased by increasing the antibody concentration, although doubling the antibody concentration has only added approximately 30 arb. U to the signal range. Because of these low signal ranges produced when using 1 μg/ml spotting solution, it would be prudent to work at a minimum of 6.25 μg/ml spotting solutions. STREP assay nos. 8 and 9 were run with increasing antibody concentration and had their readings taken earlier in an attempt to further desensitise the curve. When comparing assay no. 6 (7.5 min reading) and assay no. 9 with the same conditions except for a 5-min reading, the range had dropped from 553 to 437 arb. U as expected with the earlier reading time, but other indicators were mostly unchanged. Assay no. 10 utilised the strongest antibody concentration tested and increased the range slightly to 504 from 437 arb. U but did not bring the MRL onto the linear range of the curve. In the case of the D-CEFT results (Table 2), it is clear from assay nos. 1–8 that no matter what variation was introduced to proportionately increase the antibody concentration, it was

Development of a rapid multiplexed assay

impossible to shift the calibration curve in any meaningful way towards the MRL of 100 μg/ml, e.g. IC80 values could not be increased above 7 ng/ml. The anti-D-CEFT antibody would appear to be very sensitive to low residue concentrations. A different approach was attempted with assay nos. 9– 16 in that unlabelled antibody was also introduced in the hope that it would bind to analyte in solution thus leaving free labelled antibody to bind to the surface giving increased signal Fig. 5 Optimised multiplexed calibration curves in milk. a DCEFT calibration curve. b STREP calibration curve. c CAP calibration curve. Each point is an average of two spots. The error bars represent standard deviation

a

b

c

range and desensitise the assay. Assay no. 9 showed a slight increase in IC20 to IC80 values compared to the conventional approach of assay nos. 1–8. However, when the concentration of unlabelled antibody was increased in assay no. 10, the signal range was constricted from 113 to 91 arb. U, counteracting any observed desensitisation in the other curve markers. This would suggest that the concentration of unlabelled antibody had reached a significant enough level

Negative 0.11 0.10 0.11 Negative 0 0 2 Negative 1 1 1 Negative Low 0.02 0.08 Negative Low Low Low Negative Low Low Low Negative 0.04 0.02 0.09 Negative 0 Low 0 Negative 1 1 1 Negative 0.04 0.01 0.09 Negative Low Low 0 50 High High High D-CEFT

Row 19 Row 20 Combined

Row 15 Row 16 Combined STREP

CAP

Row 11 Row 12 Combined

Expected (ng/ml) Observed (ng/ml) Observed (ng/ml) Observed (ng/ml) Expected (ng/ml) Observed (ng/ml) Observed (ng/ml) Observed (ng/ml) Expected (ng/ml) Observed (ng/ml) Observed (ng/ml) Observed (ng/ml)

Negative 0.02 0.04 0.08 300 281 259 279 Negative Low Low Low

Negative 0.09 0.07 0.10 200 315 230 287 Negative 1 1 1

Negative 0.07 0.06 0.13 100 120 97 115 Negative Low 1 0

0.45 0.53 0.50 0.51 Negative 0 Low 0 Negative 0 0 1

0.3 0.41 0.42 0.41 Negative 1 1 2 Negative Low 1 Low

0.15 0.14 0.14 0.15 Negative 0 0 2 Negative Low Low 0

Negative Low Low 0.01 Negative Low Low Low 200 High High High

Negative Low Low 0.07 Negative 1 0 2 100 High High High

k j d c b a Sample quantitative

Table 3 Quantitative results for spiked milk samples when read against the multiplexed calibration curves. Negative samples were verified as negative milk samples by source. A value described as low has been returned by the software as being lower than the lowest concentration of

e

f

g

h

i

l

the four-parameter fit calibration curve. A value described as high has a concentration beyond the highest standard used in the calibration curve

T.F. McGrath et al.

to start competing for binding to the spot, and since only labelled antibody bound to the surface produced a signal, this will have a large impact on signal levels. Changing the labelled and unlabelled antibody concentrations as well as altering the spot concentration could not increase the signal range, assay nos. 11–16. Optimised calibration curves in milk with spiked sample analysis Figure 5 shows the optimised multiplexed calibration curves in milk used to analyse the negative and spiked milk samples. Both spotted columns for each analyte have been plotted separately. However, when all points (n=4) for each standard were combined, the curve indicators for each curve remained generally unchanged. Combined curve for D-CEFT had a range of 61 arb. U, with IC20, IC50 and IC80 of 2.2, 4.8 and 11.3 ng/ml, respectively. The combined streptomycin curve had a range of 96.0 arb. U, with IC20, IC50, IC80 and IC90 of 13, 97, 383 and 576 ng/ml, respectively. For CAP, the range was 113.0 arb. U, with IC20, IC50 and IC80 of 0.13, 0.31 and 0.71 ng/ml, respectively. The D-CEFT calibration curve (Fig. 5a) showed a similar range, to the buffer curve produced under the same assay conditions (Table 2). However, in milk, the curve was sensitive to lower concentrations of the antibiotic. When the values were compared to the original milk calibration curve (Fig. 4a), all the adjustments made to the assay parameters desensitised the curve by approximately a factor of 2, whilst the signal range had decreased by a factor of 3 from 200 to 61 arb. U. The STREP calibration curve (Fig. 5b) had its range suppressed by a factor of 4, from 440 to 96 arb. U, when compared to the buffer curve produced under the same assay conditions (Table 2). However, unlike the D-CEFT curve, there had also been a desensitisation, towards the MRL, of the other curve indicators. Comparing this milk curve to the original (Fig. 4b), it become clear that the assay manipulation had successfully desensitised the curve with IC20, IC50 and IC80 values shifting from 2, 5 and 15 ng/ml for the original milk curve to 13, 97 and 383 ng/ml for the optimised curve. This had come at the expense of signal range which decreased from 372 to 96 arb. U following curve optimisation. The CAP calibration curve (Fig. 5c) showed a slight increase in range, from 83.4 to 113 arb. U, when comparing the milk to the buffer calibration curve produced under the same conditions (Table 2). The IC20, IC50 and IC80 values remained within the same order of magnitude. Comparing this milk curve to the original (Fig. 4c), it was clear that the assay manipulation had successfully sensitised the curve towards lower concentrations by a factor of approximately 28 with IC 20 , IC 50 and IC 80 values shifting from 3.6, 9.0 and 20.3 ng/ml to 0.13, 0.31 and 0.71 ng/ml, respectively. However, once again, this was at the expense of signal range which has decreased by a factor of 8.

Development of a rapid multiplexed assay

All 12 spiked milk samples were correctly identified for chloramphenicol and streptomycin when read against individual spotted column curves as well as when all data was combined (Table 3). Ten samples were identified as compliant with values below the MRL/MRPL, and two samples were non-compliant. There were no false positives and no false negatives identified. The result at half MRL/MRPL had a recovery of between 93 and 100 % for chloramphenicol and between 97 and 120 % for streptomycin. Values were higher at the MRL/MRPL but clearly indicated a positive result whilst 1.5 times MRL/MRPL once again indicate positive samples with recoveries between 111 and 118 % for chloramphenicol and 86 to 93 % for streptomycin. The nine known negative samples for desfuroylceftiofur have been identified correctly when read against the calibration curve; however, the results for the three spiked samples were outside of the range of the calibration curve. This was to be expected with the linear range of the curve being between 2 and 11 ng/ml and the lowest spiked sample being at 50 ng/ml (half MRL). These high results can only indicate that desfuroylceftiofur is present at a concentration above 11 ng/ml. To determine if the concentration is above the MRL, a further quick screening test would be required on the milk sample after it had been diluted using a known negative milk. For the nine known negative samples, repeatability was calculated based on signal level since not all results returned a concentration value (Table 3). CVs ranged from 4.2 to 8.1 % when compared against individual spotted columns and 6.5 to 11.7 % when the two sets of column curves and sample signals were combined.

Conclusion For the first time, the development and optimisation for the SnapEsi assays have been demonstrated in both single and multiplexed forms, using a two-step approach of mixing labelled primary antibodies with untreated samples before placement onto sample cartridges. Multiplexed calibration curves were obtained for all three structurally unrelated analytes in both buffer and milk. Both streptomycin and chloramphenicol could be detected at their relevant MRL/MRPL in both buffer and raw milk, without the need for sample preparation. In its current form, the desfuroylceftiofur assay requires an additional test following a dilution step for samples contaminated above 11 ng/ml. An alternative approach would be to replace the antibody with another less sensitive binding protein. The ability to detect across a wide sensitivity range, 0.3 to 200 ng/ml, covering chloramphenicol MRPL and streptomycin MRL, within a single multiplexed assay without any sample preparation is a significant achievement and shows the versatility of the equipment and reagents used. A combination of the sciFLEXARRAYER S5 and SnapEsi instrument resulted in excellent repeatability and reproducibility values for

assay performance. Because of its small footprint, laptop power source and low %CV values, this instrument could potentially be used, with batch specific built-in calibration curves, for on-site rapid screening of contaminated milk samples, without the need for any sample preparation. Analysis time had been reduced to 5 min, which is an improvement on Lochhead et al., Meneely et al. and Devlin et al. [24–26]. Future developments could be undertaken to reduce the assay to a single step, by eliminating the need to mix antibody with sample. The detection of three structurally unrelated compounds that pose a risk to milk safety and quality could be further advanced by the inclusion of additional drug residue targets, on the same cartridge, such as β-lactams, sulphonamides and tetracyclines. The screening for multiple compounds within an individual family would be assured by evaluating the cross-reactivity profiles of the selected binding partners and using the most appropriate binder. This could complete the development of a milk testing cartridge before focus is shifted towards the development of cartridges specifically targeted towards the needs of other food industry sectors that require rapid antibiotic testing capabilities, e.g. honey and pork.

References 1. Centers for Disease Control and Prevention (2014) Antibiotic use in food-producing animals—tracking and reducing the public health impact. Available via http://www.cdc.gov/narms/animals.html. Accessed 16/01 2015 2. McGrath TF, Elliott CT, Fodey TL (2012) Biosensors for the analysis of microbiological and chemical contaminants in food. Anal Bioanal Chem 403:75–92 3. Cerniglia C, Kotarski S (1999) Evaluation of veterinary drug residues in food for their potential to affect human intestinal microflora. Regul Toxicol Pharmacol 29:238–261 4. Kaklamanos G, Vincent U, von Holst C (2013) Multi-residue method for the detection of veterinary drugs in distillers grains by liquid chromatography-Orbitrap high resolution mass spectrometry. J Chromatogr A 1322:38–48 5. FDA Grade BA^ Pasteurized Milk Ordinance (Includes provisions from the Grade BA^ Condensed and Dry Milk Products and Condensed and Dry Whey—Supplement I to the Grade BA^ PMO) 2009 Revision U.S. Department of Health and Human Services. Public Health Service. Food and Drug Administration. Available via http://www.fda.gov/downloads/Food/GuidanceRegulation/ UCM209789.pdf. Accessed 09/29 2014 6. G. M. Jones (2009) On-farm tests for drug residues in milk communications and marketing. College of Agriculture and Life Sciences, Virginia Polytechnic Institute and State University. Available via http://pubs.ext.vt.edu/404/404-401/404-401_pdf.pdf. Accessed 09/ 29 2014 7. Commission Regulation (EU) No 37/2010 of 22 December 2009 on pharmacologically active substances and their classification regarding maximum residue limits in foodstuffs of animal origin (text with EEA relevance) (OJ L 15, 20.1.2010, p. 1), consolidated version 2014-08-19

T.F. McGrath et al. 8. Commission Decision 2002/657/EC of 12 August 2002 implementing Council Directive 96/23/EC concerning the performance of analytical methods and the interpretation of results (notified under document number C(2002) 3044) (Text with EEA relevance) (OJ L 221, 17.8.2002, p. 8) consolidated version 2004-01-10 9. European Commission. Commission Staff Working Document on the Implementation of National Residue Monitoring Plans in the Member States in 2012 (Council Directive 96/23/EC). Available via http://ec.europa.eu/food/food/chemicalsafety/residues/docs/ workdoc_2012_en.pdf. Accessed 10/13 2014 10. Charm Sciences Inc (2014) Charm II. Available via http://www. charm.com/en/products/charm-ii.html. Accessed 09/30 2014 11. IDEXX Laboratories Inc (2014) Dairy testing. Available via http:// www.idexx.co.uk/dairy/dairy-testing.html. Accessed 09/30 2014 12. Charm Sciences Inc (2014) Charm ROSA milk tests. Available via http://www.charm.com/en/products/rosa-milk.html. Accessed 09/30 2014 13. Neogen Corporation Food Safety Dairy Analysis Test Kits. Available via http://www.neogen.com/FoodSafety/FS_DA_Index.html. Accessed 09/30 2014 14. DSM (2014) Milk tests. Available via http://www.dsm.com/markets/ foodandbeverages/en_US/products/tests/delvotest.html#. Accessed 09/30 2014 15. Unisensor (2012) Unisensor home page. Available via http://www. unisensor.be/. Accessed 01/13 2015 16. Bohm DA, Stachel CS, Gowik P (2009) Multi-method for the determination of antibiotics of different substance groups in milk and validation in accordance with Commission Decision 2002/657/EC. J Chromatogr A 1216:8217–8223 17. Deng X, Yang H, Li J et al (2011) Multiclass residues screening of 105 veterinary drugs in meat, milk, and egg using ultra high performance liquid chromatography tandem quadrupole time-of-flight mass spectrometry. J Liq Chromatogr Rel Technol 34:2286–2303 18. Freitas A, Barbosa J, Ramos F (2013) Development and validation of a multi-residue and multiclass ultra-high-pressure liquid chromatography-tandem mass spectrometry screening of antibiotics in milk. Int Dairy J 33:38–43 19. Hou X, Chen G, Zhu L et al (2014) Development and validation of an ultra high performance liquid chromatography tandem mass spectrometry method for simultaneous determination of sulfonamides, quinolones and benzimidazoles in bovine milk. J Chromatogr B 962:20–29

20. Toaldo IM, Gamba GZ, Picinin LA et al (2012) Multiclass analysis of antibacterial residues in milk using RP-liquid chromatography with photodiode array and fluorescence detection and tandem mass spectrometer confirmation. Talanta 99:616–624 21. Raz SR, Bremer MGEG, Haasnoot W et al (2009) Label-free and multiplex detection of antibiotic residues in milk using imaging surface plasmon resonance-based immunosensor. Anal Chem 81:7743– 7749 22. Douglas D, Banaszewski K, Juskelis R et al (2012) Validation of a rapid lateral flow test for the simultaneous determination of betalactam drugs and flunixin in raw milk. J Food Prot 75:1270–1277 23. Zhi A, Li B, Liu Q et al (2010) Development of a lateral-flow immunochromatographic test device for the rapid detection of difloxacin residues. Food Agric Immunol 21:335–345 24. Meneely JP, Campbell K, Greef C et al (2013) Development and validation of an ultrasensitive fluorescence planar waveguide biosensor for the detection of paralytic shellfish toxins in marine algae. Biosens Bioelectron 41:691–697 25. Devlin S, Meneely JP, Greer B et al (2013) Next generation planar waveguide detection of microcystins in freshwater and cyanobacterial extracts, utilising a novel lysis method for portable sample preparation and analysis. Anal Chim Acta 769:108–113 26. Lochhead MJ, Todorof K, Delaney M et al (2011) Rapid multiplexed immunoassay for simultaneous serodiagnosis of HIV-1 and coinfections. J Clin Microbiol 49:3584–3590 27. Burkin M, Galvidis I (2013) Immunochemical detection of apramycin as a contaminant in tissues of edible animals. Food Control 34:408–413 28. McNamee SE, Elliott CT, Delahaut P et al (2013) Multiplex biotoxin surface plasmon resonance method for marine biotoxins in algal and seawater samples. Environ Sci Pollut Res 20:6794–6807 29. Vdovenko MM, Lu C, Yu F et al (2014) Development of ultrasensitive direct chemiluminescent enzyme immunoassay for determination of aflatoxin M1 in milk. Food Chem 158:310–314 30. Reimer G, Gee S, Hammock B (1998) Comparison of a timeresolved fluorescence immunoassay and an enzyme-linked immunosorbent assay for the analysis of atrazine in water. J Agric Food Chem 46:3353–3358 31. McGrath TF, Campbell K, Fodey TL et al (2013) An evaluation of the capability of a biolayer interferometry biosensor to detect lowmolecular-weight food contaminants. Anal Bioanal Chem 405:2535– 2544