Target screening of 105 veterinary drug residues in milk using UHPLC/ESI Q-Orbitrap multiplexing data independent acquisition

Transcription

1 Analytical and Bioanalytical Chemistry (2018) 410: PAPER IN FOREFRONT Target screening of 105 veterinary drug residues in milk using UHPLC/ESI Q-Orbitrap multiplexing data independent acquisition Jian Wang 1 & Daniel Leung 1 & Willis Chow 1 & James Chang 2 & Jon W. Wong 3 Received: 8 November 2017 /Revised: 11 December 2017 /Accepted: 21 December 2017 /Published online: 5 February 2018 # Crown 2018 Abstract This paper presents a multi-class target screening method for the detection of 105 veterinary drug residues from 11 classes in milk using ultra-high performance liquid chromatography electrospray ionization quadrupole Orbitrap mass spectrometry (UHPLC/ ESI Q-Orbitrap). The method is based on a non-target approach of full mass scan and multiplexing data-independent acquisition (Full MS/mDIA). The veterinary drugs include endectocides, fluoroquinolones, ionophores, macrolides, nitroimidazole, NSAIDs, β-lactams, penicillins, phenicols, sulfonamides, and tetracyclines. Veterinary drug residues were extracted from milk using a salting-out and solid-phase extraction (SOSPE) procedure, which entailed the precipitation of milk proteins by an extraction buffer (oxalic acid and EDTA, ph 3) and acetonitrile, a salting-out acetonitrile/water phase separation using ammonium sulfate, and solid-phase extraction for clean-up using polymeric reversed-phase sorbent cartridges. The Q-Orbitrap Full MS/dd-MS 2 (data-dependent acquisition) was used to acquire product-ion spectra of individual veterinary drugs to build a compound database and a mass spectral library, whereas its Full MS/mDIA was utilized to acquire sample data from milk for target screening of veterinary drugs fortified at 1.0 or 10.0 μg/kg. The in-spectrum mass correction or solvent background lockmass correction was used to minimize mass error when building the compound database from experimental dd-ms 2 accurate mass data. Retention time alignment and response threshold adjustment were used to eliminate or reduce false negatives and/or false positive rates. The validated method was capable of screening 58% and 96% of 105 veterinary drugs at 1.0 and 10.0 μg/kg, respectively, without manually evaluating every compound during data processing, which will reduce the workload in routine practice. Keywords UHPLC/ESI Q-Orbitrap. Veterinary drug residues. Compound database. Target screening. Milk. Multiplexing data-independent acquisition Introduction Veterinary drugs have been widely used in veterinary medicine and animal production to treat and prevent animal diseases and to enhance growth rate and feed efficiency. Incorrect Published in the topical collection Food Safety Analysis with guest editor Steven J. Lehotay. * Jian Wang jian.wang@inspection.gc.ca Calgary Laboratory, Canadian Food Inspection Agency, th Street N.W, Calgary, Alberta T2L 2L1, Canada ThermoFisher Scientific, 355 River Oaks Parkway, San Jose, CA 95134, USA Center for Food Safety and Applied Nutrition, US Food and Drug Administration, 5100 Campus Drive, College Park, MD 20740, USA administration of the drug or improper withdrawal time after treatment could lead to the presence of drug residues in foods of animal origin. The residues may provoke allergic reactions in some hypersensitive individuals or encourage the spread of drug-resistant pathogenic bacterial strains [1 3]. Furthermore, veterinary drugs present in milk can have negative implications on microbial processes (for example cheese and yogurt production [4, 5]). Therefore, the residue levels of veterinary drugs should not exceed the maximum residue limits set by regulatory agencies in milk to ensure the safety of food supply. This justifies the need for analytical methods that are capable of quantifying and screening an increasingly large number of veterinary drug residues that are potentially used for food production in routine monitoring programs [6]. Veterinary drug residues in food can be determined through biological screening methods such as microbial inhibition tests, immunochemical methods etc., and quantitative and confirmatory methods such as liquid chromatography coupled

2 5374 Wang J. et al. to mass spectrometry (LC-MS). Historically, veterinary drug residues, which were typically in a group of less than 20 compounds, were analyzed by a single-class or related families using LC/MS/MS. A single-class method was relatively easy to optimize for both extraction and instrument parameters because of the similar physical and chemical properties of veterinary drugs. In the last decade, the use and development of high resolution mass spectrometric instrumentation such as LC-Orbitrap and LC-TOF (Time of Flight) have increased rapidly. Their applications for quantitation and target screening of chemical residues in food are advantageous because of their high massresolving power and accurate mass measurement, multifunction data acquisition, and powerful data processing capacity. The Orbitrap and TOF mass spectrometers (i.e., high-end models) offer high mass-resolution (>20,000 FWHM), accurate mass measurement (<5, excellent Full MS scan sensitivity, and complete mass spectrometric information. The Full MS scan data allow for the screening of targeted analytes, quantitation of selected compounds, and retrospective analysis (knowns or unknowns) even when appropriate standards are not available at the time of analysis. When operated in Full MS scan mode with generic instrument parameter settings, these instruments have been increasingly used in target approach based on accurate masses to quantify or screen veterinary residues in food. Examples of quantitative and/or screening methods based on Full MS scan that have been reported for multi-veterinary drugs at various concentration levels include screening of 150 veterinary drugs in milk using TOF [7]; 100 veterinary drugs in egg, milk, fish, and animal tissues using TOF [8, 9]; 54 veterinary drugs in honey and 59 veterinary drugs in milk using Q-TOF [10]; 100 veterinary drugs in milk using Q-Orbitrap [6]; 105 veterinary drugs in milk using Q-Orbitrap [11]; 90 veterinary drugs in royal jelly using Q-TOF [12]. Those applications focused on a generic sample preparation and a full set of calibration standards for quantifying or screening of as many as possible veterinary drugs in food. Recently, target screening based on non-target approach of data acquisition and retrospective data processing has become one type of promising application for the purpose of screening using high resolution mass spectrometers such as TOF or Orbitrap [13 16]. Data were acquired using a combination of Full MS scan, all in fragmentation (AIF) and dataindependent acquisition (DIA), etc. The screening of targeted chemical contaminants, for example veterinary drugs and pesticides, was accomplished using accurate masses of the targeted precursor ion (i.e., pseudomolecular ion) along with retention time or characteristic fragment ion by specific software for automated data mining and exploitation. These types of analyses are often referred to as screening methods or semiautomated mass spectrometry-based detections. The screening concept offers laboratories an effective means to extend their analytical scope to chemical contaminants, which potentially have a low probability of being present in the samples. The residues that occur more frequently would continue to be monitored using validated quantitative multi-residue methods [17]. This study was designed to develop a systematic and detailed protocol for the development of a compound database (CDB) of the retention time and the exact or accurate mass of a precursor and its fragments and a mass spectral (MS) library of a product ion spectrum, and their applications for target screening of 105 veterinary drugs in milk based on multiplexing data independent acquisition using a Q- Orbitrap. Full MS/dd-MS 2 (data-dependent acquisition) was utilized to acquire product-ion spectra of veterinary drugs for individual standards to obtain exact masses of fragments for a CDB and a MS library. The CDB was built based on theoretical and experimental mass data. Accurate mass, retention time, and response threshold were three key parameters that were used to build a functional and working CDB, and they were optimized or adjusted to reduce false negatives and/or false positives. Full MS/mDIA (multiplexing dataindependent acquisition) was used to acquire sample data from milk for target screening of veterinary drugs fortified at 1.0 or 10.0 μg/kg. The method was validated according to SANTE/11945/2015 [17], which provided guidance for target screening validation. This qualitative method significantly reduced the workload for data processing in routine practice. The entire procedure, including sample extraction and data processing, allowed for high-throughput testing of routine samples, which could benefit veterinary drug residue monitoring programs. Material and methods Chemicals and reagents Ten batches of whole milk (4 or 8L/batch) were collected from 10 different local farms. All milk samples, which were tested free of veterinary drug residues using the previously developed UHPLC/ESI Q-Orbitrap method [11], were kept at 20 C. Pierce LTQ ESI positive ion calibration solution (10 ml) was purchased from ThermoFisher Scientific (Rockford, IL, USA). The calibration solution, which includes n-butylamine (m/z 74), caffeine (m/z 195 and its fragment m/z 138), Ultramark 1621 (m/z 1022, 1122, 1222, 1322, 1422, 1522, 1622, 1722, 1822), and MRFA (m/z 524), was used to tune and calibrate the Q- Orbitrap. Ammonium acetate (reagent grade), ammonium sulfate (reagent plus, >99.0%), formic acid (LC-MS grade, ~98%), ammonium formate (for mass spectrometry, >99.0%), oxalic acid (reagent plus, >99.0%), ammonium hydroxide (28% 30%), and LC-MS acetonitrile (Chromasolv, 2.5 L) were purchased from Sigma-Aldrich Corp. (Oakville, Canada). Acetic

3 Target screening of 105 veterinary drug residues in milk using UHPLC/ESI Q-Orbitrap multiplexing data acid (glacial acetic acid, reagent grade, 99.7%), acetonitrile (distilled in glass), methanol (distilled in glass), and EDTA disodium salt were obtained from Caledon Laboratories Ltd. (Georgetown, ON, Canada). Water (18.2 MΩ cm) used for reagent and sample preparation was obtained from a Barnstead Nanopure system (Thermo Scientific, Marietta, OH, USA). Veterinary drug standards were obtained from Sigma-Aldrich Corp. or Toronto Research Chemicals (ON, Canada). A LC vial was a 0.45 μm PVDF Syringeless Filter Device Mini-UniPrep with polypropylene housing (GE Healthcare UK Limited, Little Chalfont, UK). OASIS HLB Plus 225 mg cartridges were purchased from Waters Corporation (Milford, MA, USA). Preparation of standard solutions Individual veterinary drug standard stock solutions were generally prepared at a concentration of 1000 or 2000 μg/ml in methanol, acetonitrile, or water. Intermediate veterinary drug working solutions were prepared at 5.0 μg/ml in methanol from stock solutions. Stock and intermediate solutions were stored at 20 C. Two-level veterinary drug standard mix working solutions for sample fortification were prepared by transferring 0.4 and 4.0 ml of 5.0 μg/ml into two separate 100 ml volumetric flasks for their respective concentration levels, and then making up to volume with acetonitrile. The resulting concentrations were and μg/ml, which were used for sample fortification at 1.0 or 10.0 μg/kg. All working solutions were stored at 4 C. UHPLC/ESI Q-Orbitrap parameters UHPLC/ESI Q-Orbitrap system consisted of an Accela 1250 LC pump and an Accela open autosampler coupled with a Q- Exactive mass spectrometer (ThermoFisher Scientific, Germany). The system was controlled by Xcalibur 3.1 with Tune 2.8 SP1 Build The UHPLC/ESI Q-Orbitrap instrument parameter settings were the same as those we used for the target screening of 448 pesticide residues in fruits and vegetables [15]. (a) Ultra-high pressure liquid chromatography UHPLC mobile phases A and B consisted of 4 mm ammonium formate and 0.10% formic acid in water and methanol, respectively. The UHPLC column utilized was a Hypersil Gold, 100 mm 2.1 mm, 1.9 μm column (Thermo Scientific, Marietta, OH, USA). The UHPLC guard column was an Accucore aq mm, 2.6 μm Defender cartridge (Thermo Scientific, Marietta, OH, USA). The UHPLC gradient profile and flow rate are shown in Table 1. Columnoven temperature was set at 45 C and auto-sampler temperature was set at 5 C. Injection volume was 5 μl and the total runtime was 14 min. Table 1 Steps Time (b) UHPLC gradient profile Mobile phase A (%) Q-Orbitrap parameters Mobile phase B (%) Flow rate (μl/ The Q-Exactive ion source was equipped with a heated electrospray ionization (HESI) probe and the Q-Orbitrap was tuned and calibrated using positive LTQ calibration solution once per week. The workflow of target screening is shown in Fig. 1. To obtain the product-ion spectra that were used to build an inhouse compound database and a mass spectral library, the Q- Exactive was operated in Full MS/dd-MS 2. During the Full MS scan, the Q-Exactive mass-resolution was set at 70,000 FWHM; AGC target: 1.0E6; maximum IT: 250 ms; and scan range m/z 80 to If the targeted mass of a compound from the inclusion list was detected within ±5 ppm mass tolerance, the precursor ion was isolated by the quadrupole and sent to the HCD (higher energy collisional dissociation) cell for fragmentation via the C-trap. The inclusion list consisted of the masses of precursor ions in forms of [M+H] +,[M+NH 4 ] +,and [M+Na] + for veterinary drugs. The precursor ion was fragmented with a stepped normalized collision energy (NCE) to generate product-ion spectra. For dd-ms 2,theQ- Exactive mass-resolution was set at 35,000 FWHM; AGC Compound Database Workflow Individual veterinary drug standards by UHPLC Full MS/dd-MS 2 or samples by UHPLC Full MS/mDIA Semi-automated Qualitative Analysis for Target Screening Mass Spectral Library Fig. 1 Workflow for building Compound database and mass spectral library using UHPLC/ESI Full MS/dd-MS 2 and sample analysis by UHPLC/ESI Full MS/mDIA, and data process for target screening

4 5376 Wang J. et al. target: 2E5; maximum IT: 120 ms; isolation window: m/z 1.0; NCE/stepped NCE: 20, 40 and 60; underfill ratio: 10%; intensity threshold: 1.7E5; apex trigger: 2 to 4 s; and dynamic exclusion: 10.0 s. Each veterinary drug standard was injected twice into the system at 50 μg/l (ppb) or higher concentration. For target screening of veterinary drugs in milk samples, data acquisition was achieved through Full MS (m/z ) and mdia (m/z ) (Fig. 2). For Full MS, the Q-Exactive mass-resolution was set at 70,000 FWHM; AGC target: 3.0E6; maximum IT: 200 ms; scan range: m/z 100 to For mdia with segments of mass ranges covering from m/z 100 to 900, the Q-Exactive mass-resolution was set at 17,500 FWHM; AGC target: 1.0E6; maximum IT: auto; loop count: 10; MSX (multiplexing) count: 4; isolation window: m/z 52.0; and NCE/stepped NCE 20, 40, 60. Thus, for loop counts from 1 to 8, a collection of ions isolated by the quadruple in every m/z 50 mass increment from m/z 100 to 500 were sent to HCD cell for fragmentation, and then to Orbitrap via the C-trap in eight steps. For loop count 9 with MSX (multiplexing) count 4, a collection of ions isolated by the quadruple for mass ranges in segments of m/z , , , and were sent to HCD cell sequentially for fragmentation, and then all fragments stored in HCD were sent to Orbitrap via the C-trap in one step. The same procedure was used for loop count 10 with MSX count 4 for mass ranges in segments of m/z , , , and The inclusion list and MSX ID for mdia are shown in Table 2 and Fig. 2A. The masses in the list were centered with each isolation window of m/z 52. Other Q- Exactive generic parameters were: sheath gas flow rate set at 60; Aux gas flow rate: 30; Sweep gas flow rate: 2; Spray voltage (KV): 3.50; Capillary temperature ( C): 350; S-lens a Sixteen isolation window at 50 Da each from m/z IN IL IW Multiplexing 1 Multiplexing 4 b c Fig. 2 (A) Q-Orbitrap multiplexing data independent acquisition. IN: isolation window number. IL: inclusion list. IW: isolation window; (B) target screening for oleandomycin in milk at 10.0 μg/kg by compound database; (C) target screening for oleandomycin in milk at 10.0 μg/kg by mass spectral library

5 Target screening of 105 veterinary drug residues in milk using UHPLC/ESI Q-Orbitrap multiplexing data Table 2 level: 55.0 and Heater temperature ( C): 350 as reported elsewhere [15, 18]. Sample extraction and clean-up The sample preparation was based on a two-step process of salting-out acetonitrile/water extraction (Step 1) and solid-phase extraction clean-up (Step 2) known as salting-out and solid-phase extraction (SOSPE) reported elsewhere [11]. (a) Step 1 Extraction A milk sample (5 g) was weighed into 50 ml centrifuge tubes (VWR International, Edmonton, Alberta, Canada). Two hundred fifty μl per two-level veterinary drug standard working solution for sample fortification was added into each centrifuge tube to provide 1.0 or 10.0 μg/kg equivalent residue in sample. After 15 min, 5 ml of extraction buffer and 10 ml of acetonitrile were added to the sample. The extraction buffer contained 0.86% oxalic acid and 0.74% EDTA disodium salt, and its ph was adjusted to 3.0 using ammonium hydroxide. The sample mixture was capped and shaken for 30 s by hand and then centrifuged at 3000 rpm (~2100 g) for 5 min using a centrifuge. The supernatant was transferred into another 50 mlcentrifugetube,towhich1gofammoniumsulfatewas added. The sample mixture was shaken by hand to mix for 2 min, and then was left to stand for 2 min. A phase separation was observed from the mixture. (b) Method editor - Inclusion List Mss (m/z) Polarity MSX ID Positive Positive Positive Positive Positive Positive Positive Positive Positive Positive Positive Positive Positive Positive Positive Positive 10 Step 2 SPE clean-up Polymeric reversed-phase sorbent OASIS HLB Plus cartridges were used for SPE clean-up. OASIS HLB Plus 225 mg cartridges were attached to 25 ml syringe barrels and set up for solid-phase extraction using a Visiprep 24-port SPE Vacuum Manifold (Sigma-Aldrich Corp). The cartridges were preconditioned sequentially with 10 ml of methanol, 10 ml of water, and 2 ml of extraction buffer. The sample extracts from Step 1 were separated into three layers after centrifugation (3000 rpm or ~2100 g, 3. The top acetonitrile layer (~10 ml) was transferred to a mm disposable test tube and retained for later to be loaded onto OASIS HLB cartridges. The lower aqueous layer was transferred onto the preconditioned Oasis HLB cartridges under vacuum at 2 to 3 inhg with a flow rate of ~1 ml/min. The very thin middle white layer (~1 or 2 mm) was discarded and should be not transferred onto the cartridges. Oasis HLB cartridges were rinsed further with 2 ml of extraction buffer and allowed to run dry. The previously retained acetonitrile layer, which served as eluting solvent as well, was loaded onto the OASIS HLB cartridges through syringe barrels. The flow (~1 ml/ of the eluting solvent was maintained under vacuum at 2to 3 inch Hg and the eluent was collected into a mm disposable glass test tube. The OASIS HLB cartridges were run dry under vacuum. Then an additional 5 ml of methanol was dispensed onto the cartridges to elute further. The eluent was collected into the same test tube as above (acetonitrile layer). The OASIS HLB cartridges were run dry under vacuum. The eluent (~15 ml) was capped and inverted to thoroughly mix. Three ml of eluent was transferred into individual 5 ml PYREX brand centrifuge tubes, which was pre-calibrated with 1 ml volume accuracy (VWR International, Edmonton, Alberta, Canada). The sample extracts were evaporated to ml, which took approximately 20 min, using an N-EVAP nitrogen evaporator (Organomation Associates Inc., Berlin, MA, USA) at 50 C under a stream of nitrogen. Then the extracts were reconstituted by making up to 0.5 ml with acetonitrile and then to 1.0 ml with 0.1 M ammonium acetate. The extracts were vortexed for 30 s. Five hundred μl ofsampleextracts was transferred into a 0.45 μm PVDF Syringeless Filter Device Mini-UniPrep vial and pressed to filter. Sample extracts were ready to be injected to UHPLC/ESI Q-Orbitrap MS for analysis. Experimental design and method validation The method was validated at 1.0 and 10.0 μg/kg using a total of 10 different raw milk blank samples. For each matrix, samples were spiked at 1.0 or 10.0 μg/kg, in duplicate. The experiment was repeated on three different days. The positive screen of a veterinary drug should meet the criterion that it can be detected in at least 95% of the 20 samples (i.e., an acceptable false-negative rate of 5%) in the batch [17].

6 5378 Wang J. et al. Data processing The Target Screening function of TraceFinder 3.3 (ThermoFisher Scientific, USA) was used for data processing based on the compound database developed in-house. For target screening parameter settings, response threshold was set individually for each veterinary drug (Table 3, column 5); mass accuracy: 5 ppm for both precursor ion and fragment; retention time window: 60 s; minimum number of fragments: 1; and MS order: MS 2. Results and discussion Target screening workflow The workflow of semi-automated qualitative analysis for target screening is shown in Fig. 1.FullMS/dd-MS 2 was used to acquire product ion spectra of individual veterinary drugs. The mass-resolution for Full MS was set at 70,000 FWHM, the resolution for dd-ms 2 was at 35,000 FWHM, and the isolation window was at 1.0 m/z. The product-ion spectra provided the accurate masses of fragments that were used to build both the compound database (CDB) and the mass spectral (MS) library. The retention times were obtained from the extracted chromatograms of Full MS using the exact masses of individual veterinary drugs. Both CDB and MS library were tested for their applicability for target screening. The development and application of a CDB based on exact (or accurate mass) and retention time are the main focus in the current study since the mass accuracy and retention time tolerance are well accepted criteria for identification and are associated with the identity of a veterinary drug for target screening. The CDB, which included compound name, elemental composition, exact or accurate mass, retention time, and response threshold, was first organized in Microsoft Excel and then imported into TraceFinder. MS library was created using Thermo Library Manager 2.0 by importing individual product ion spectra of veterinary drugs into the library. For sample data acquisition, Full MS/mDIA (multiplexing data independent acquisition), which acquired spectra of precursor ions as well as their fragments from per defined mass range in multiple steps or events, was utilized. The Q-Orbitrap first performed one Full MS at 70,000 FWHM, followed by mdia at 17,500 FWHM in 10 steps or loops (Fig. 2A, B,and Table 2). For loop counts from 1 to 8 or MSX ID 1 to 8, a collection of ions isolated by the quadruple in every m/z 50 mass increment from m/z 100 to 500 were sent to the HCD cell for fragmentation, and then fragments were sent to the Orbitrap via the C-trap in eight steps. However, for loop counts 9 and 10, fragments were obtained using the multiplexing function of Q-Orbitrap, i.e., MSX count 4. For example, for loop count 9, a collection of ions, which were first isolated by the quadruple for mass ranges in segments of m/z , , , and , were sent to the HCD cell sequentially (namely four events) for fragmentation. The fragments from the four events, which were stored in the HCD, were then sent to Orbitrap as one packet of ions via the C-trap in one step. The spectrum obtained in this case was also called a multiplexed spectrum. The same multiplexing is used for loop count 10 for mass ranges in segments of m/z , , , and Multiplexing was used for the mass range of m/z to speed up the cycle time. In this mass range, less selectivity (small isolation window) is needed, compared with the range of m/z Therefore, the extracted chromatograms of Full MS provided the accurate mass and retention time of a precursor, and mdia likely offered the accurate masses of the fragments generated in the HCD from a precursor for target screening. Target screening parameters and criteria The target screening or semi-automated qualitative analysis was performed using TraceFinder 3.3 Target Screening function. The screening parameters and criteria were based on either Retention Time (±0.5 and mass accuracy ( 5 of a Precursor (RTP by Full MS), or Retention Time (±0.5 and mass accuracy ( 5 of the precursor and a Fragment Ion (RTFI by Full MS/mDIA). A method performance acceptability criterion was set at an acceptable falsenegative rate of 5%. Screening results of incurred residues by RTP approach were considered as tentative positive findings, whereas those by the RTFI were taken as confirmative positive findings. Compound database The accurate masses, retention times, and response thresholds were three key parameters in the compound database (CDB). The mass correction, retention time alignment, and response threshold adjustment were based on a protocol reported elsewhere [15]. Retention time alignment should be made for every batch of samples but mass correction and response threshold should be corrected and set in the beginning when building the CDB. Furthermore, response threshold can be adjusted accordingly based on routine experiences to reduce false positives if a high constant matrix background occurs. The three parameters were first organized in a Microsoft ExcelCDBtemplateandthenimportedtoTracefinder3.3to create an executable compound database (ecdb). The exact masses of precursors were calculated theoretically from the elemental compositions while the accurate masses of their fragments were obtained from the product-ion spectra of individual standards (50 μg/l) acquired using Full MS/dd-MS 2 (Table 3). The Q-Orbitrap performed a three-step normalized collision, i.e.,

7 Target screening of 105 veterinary drug residues in milk using UHPLC/ESI Q-Orbitrap multiplexing data Table 3 UHPLC/ESI Q-Orbitrap compound database of veterinary drug for target screening Compound name Class Total number Chemical formula Response threshold a Adduct Retention time ( Mass b Exact Accurate Accurate Accurate Accurate Precursor Fragment 1 Fragment 2 Fragment 3 Fragment 4 1 c Abamectin B1a Endectocides 7 C48H72O [M+Na] Doramectin Endectocides C50H74O [M+Na] Emamectin B1a Endectocides C49H75NO [M+H] Eprinomectin B1a Endectocides C50H75NO [M+Na] Ivermectin Endectocides C48H74O [M+Na] Moxidectin Endectocides C37H53NO [M+Na] Selamectin Endectocides C43H63NO [M+H] Cinoxacin Fluoroquinolones 17 C12H10N2O [M+H] Ciprofloxacin Fluoroquinolones C17H18FN3O [M+H] Danofloxacin Fluoroquinolones C19H20FN3O [M+H] Difloxacin Fluoroquinolones C21H19F2N3O [M+H] Enoxacin Fluoroquinolones C15H17FN4O [M+H] Enrofloxacin Fluoroquinolones C19H22FN3O [M+H] Flumequine Fluoroquinolones C14H12FNO [M+H] Lomefloxacin Fluoroquinolones C17H19F2N3O [M+H] Marbofloxacin Fluoroquinolones C17H19FN4O [M+H] Nalidixic Acid Fluoroquinolones C12H12N2O [M+H] Norfloxacin Fluoroquinolones C16H18FN3O [M+H] Ofloxacin Fluoroquinolones C18H20FN3O [M+H] Orbifloxacin Fluoroquinolones C19H20F3N3O [M+H] Oxolinic Acid Fluoroquinolones C13H11NO [M+H] Pipemidic Acid Fluoroquinolones C14H17N5O [M+H] Sarafloxacin Fluoroquinolones C20H17F2N3O [M+H] Sparfloxacin Fluoroquinolones C19H22F2N4O [M+H] Lasalocid Ionophores 5 C34H54O [M+NH 4 ] Monensin Ionophores C36H62O [M+NH4] Narasin Ionophores C43H72O [M+NH4] Nigericin Ionophores C40H68O [M+NH 4 ] Salinomycin Ionophores C42H70O M+NH Erythromycin Macrolides 8 C37H67NO [M+H] Neospiramycin I Macrolides C36H62N2O [M+2H] Oleandomycin Macrolides C35H61NO [M+H] Roxithromycin Macrolides C41H76N2O [M+2H] Spiramycin I Macrolides C43H74N2O [M+2H] Tilmicosin Macrolides C46H80N2O [M+2H] Tylosin A Macrolides C46H77NO [M+H] Tylosin B Macrolides C39H65NO [M+H] methyl-4 Nitroimidazoles 13 C4H5N3O [M+H] (5)-nitroimidazole Dimetridazole Nitroimidazoles C5H7N3O [M+H] Etanidazole Nitroimidazoles C7H10N4O [M+H] HMMNI Nitroimidazoles C5H7N3O [M+H] Ipronidazole Nitroimidazoles C7H11N3O [M+H] Ipronidazole-OH Nitroimidazoles C7H11N3O [M+H] Metronidazole Nitroimidazoles C6H9N3O [M+H] Metronidazole-OH Nitroimidazoles C6H9N3O [M+H] Nimorazole Nitroimidazoles C9H14N4O [M+H] Ornidazole Nitroimidazoles C7H10N3O3Cl [M+H] Ronidazole Nitroimidazoles C6H8N4O [M+H] Ternidazole Nitroimidazoles C7H11N3O [M+H]

8 5380 Wang J. et al. Table 3 (continued) Compound name Class Total number Chemical formula Response threshold a Adduct Retention time ( Mass b Exact Accurate Accurate Accurate Accurate Precursor Fragment 1 Fragment 2 Fragment 3 Fragment 4 Tinidazole Nitroimidazoles C8H13N3O4S [M+H] hydroxyflunixin NSAIDS 3 C14H11F3N2O [M+H] Flunixin NSAIDS C14H11F3N2O [M+H] Phenylbutazone NSAIDS C19H20N2O [M+H] Ampicillin Penicillins 6 C16H19N3O4S [M+H] Cloxacillin Penicillins C19H18ClN3O5S [M+H] Dicloxacillin Penicillins C19H17Cl2N3O5S [M+H] Oxacillin Penicillins C19H19N3O5S [M+H] Penicillin G Penicillins C16H18N2O4S [M+H] Penicillin V Penicillins C16H18N2O5S [M+H] Florfenicol Phenicols 2 C12H14Cl2FNO4S [M+NH4] Thiamphenicol Phenicols C12H15Cl2NO5S [M+Na] Dapsone Sulfonamides 26 C12H12N2O2S [M+H] Sulfabenzamide Sulfonamides C13H12N2O3S [M+H] Sulfacetamide Sulfonamides C8H10N2O3S [M+H] Sulfachloropyridazine Sulfonamides C10H9ClN4O2S [M+H] Sulfadiazine Sulfonamides C10H10N4O2S [M+H] Sulfadimethoxine Sulfonamides C12H14N4O4S [M+H] Sulfadoxine Sulfonamides C12H14N4O4S [M+H] Sulfaethoxypyridazine Sulfonamides C12H14N4O3S [M+H] Sulfaguanidine Sulfonamides C7H10N4O2S [M+H] Sulfamerazine Sulfonamides C11H12N4O2S [M+H] Sulfameter Sulfonamides C11H12N4O3S [M+H] Sulfamethazine Sulfonamides C12H14N4O2S [M+H] Sulfamethizole Sulfonamides C9H10N4O2S [M+H] Sulfamethoxazole Sulfonamides C10H11N3O3S [M+H] Sulfamethoxypyridazine Sulfonamides C11H12N4O3S [M+H] Sulfamonomethoxine Sulfonamides C11H12N4O3S [M+H] Sulfamoxole Sulfonamides C11H13N3O3S [M+H] Sulfanilamide Sulfonamides C6H8N2O2S [M+H] Sulfanitran Sulfonamides C14H13N3O5S [M+H] Sulfaphenazole Sulfonamides C15H14N4O2S [M+H] Sulfapyridine Sulfonamides C11H11N3O2S [M+H] Sulfaquinoxaline Sulfonamides C14H12N4O2S [M+H] Sulfathiazole Sulfonamides C9H9N3O2S [M+H] Sulfisomidine Sulfonamides C12H14N4O2S [M+H] Sulfisoxazole Sulfonamides C11H13N3O3S [M+H] Trimethoprim Sulfonamides C14H18N4O [M+H] epitetracycline Tetracyclines 5 C22H24N2O [M+H] Chlortetracycline Tetracyclines C22H23ClN2O [M+H] Doxycycline Tetracyclines C22H24N2O [M+H] Oxytetracycline Tetracyclines C22H24N2O [M+H] Tetracycline Tetracyclines C22H24N2O [M+H] Cefamandole β-lactams 13 C18H18N6O5S [M+H] Cefapirin β-lactams C17H17N3O6S [M+H] Cefazolin β-lactams C14H14N8O4S [M+H] Cefoperazone β-lactams C25H27N9O8S [M+H] Cefotaxime β-lactams C16H17N5O7S [M+H] Cefoxitin β-lactams C16H17N3O7S [M+NH4] Cefquinome β-lactams C23H24N6O5S [M+H] Ceftiofur β-lactams C19H17N5O7S [M+H]

9 Target screening of 105 veterinary drug residues in milk using UHPLC/ESI Q-Orbitrap multiplexing data Table 3 (continued) Compound name Class Total number Chemical formula Response threshold a Adduct Retention time ( Mass b Exact Accurate Accurate Accurate Accurate Precursor Fragment 1 Fragment 2 Fragment 3 Fragment 4 Cefuroxime β-lactams C16H16N4O8S [M+NH4] Cephacetrile β-lactams C13H13N3O6S [M+NH4] Cephalothin β-lactams C16H16N2O6S [M+Na] Cephradine β-lactams C16H19N3O4S [M+H] Desacetyl cephapirin β-lactams C15H15N3O5S [M+H] Total 105 Percentage (%) a Numbers that are underlined are response thresholds set to default of due to low sensitivity of a veterinary drug in UHLPC/ESI Q-Orbitrap b The masses of fragments are corrected based on the mass accuracy of a precursor or c Column number NCE at 20, 40, and 60, to induce fragmentation. The fragments, which were generated sequentially in three NCEs and collected in the HCD, were sent altogether to the Orbitrap analyzer via C- Trap for a single scan detection. Since the stepped NCE was not optimized for each individual veterinary drug, the obtained product-ion spectra did not represent the best scenario in terms of sensitivity for identification, especially for low abundance fragments. In general, the developed CDB served the purpose for target screening of veterinary drugs at the required concentration 1.0 or 10.0 μg/kg levels as indicated in the validated results (Table 4). A veterinary drug can be ionized in the form of [M+H] +, [M+NH 4 ] +,or[m+na] +. Each exact mass of a precursor (protonated or adduct) was selected and entered into Xcalibur Qual Browser to determine which yielded the highest extracted ion peak and the most intense fragment spectrum. Often compound s protonated form showed the highest abundance (Table 3, column 6). The four most abundant fragments plus the precursor were selected to build the CDB (Table 3,columns8 12). The accurate mass of a precursor, which was eventually replaced by the exact mass, was always placed in the first position (Table 3, column 8) in the Excel template, followed by the masses of four fragments arranged according to ion abundance from high to low (Table 3, columns 9 12). (a) Mass correction Mass correction was based on either in-spectral mass correction or solvent background lock-mass correction [15]. In brief, this was achieved using the following equations: A ¼ B þ B ðδme=1; 000; 000Þ ΔMe ¼ C B 1; 000; 000 C where ΔMe : mass accuracy or mass error in ppm; A: corrected mass; B: measured mass; C: exact mass or theoretical value. When the mass error of a precursor was less than 2.5 ppm, the mass correction was made according to the mass accuracy of the precursor in the dd-ms 2 product-ion spectrum and this is called in-spectral mass correction. When a precursor was not observed in the dd-ms 2 product-ion spectrum or its mass error was greater than 2.5 ppm, the mass correction was made according to the mass accuracy of an ion in the background of a Full MS scan spectrum, i.e., m/z , which is n-butyl benzenesulfonamide, and this is the solvent background lockmass correction. The 2.5 ppm cutoff was arbitrarily chosen to reduce the probability of mass overcorrection. Columns 9 12 in Table 3 present the corrected masses for fragments and column 8 lists the exact masses of the precursors. The masses in columns 8 12 in Table 3 were used to build a compound

10 5382 Wang J. et al. Table 4 UHPLC/ESI Q-Orbitrap Full MS/mDIA target screening results of veterinary drug residues spiked in milk Compound name Class Total number Target screening (Scenario I) a Target screening (Scenario II) a 1.0 μg/kg 10.0 μg/kg 1.0 μg/kg 10.0 μg/kg tr(± 0.5 trand FI LS10 tr(± 0.5 trand FI LST10 tr(± 0.5 trand FI LS10 tr(± 0.5 trand FI LST10 1 b Abamectin B1a Endectocides 7 No No No No Yes Yes Yes Yes No No No No Yes No No Yes Doramectin Endectocides No No No No Yes No No No No No No No Yes No No No Emamectin B1a Endectocides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Eprinomectin B1a Endectocides No No No No Yes No No No No No No No Yes No No No Ivermectin Endectocides No No No No Yes No No No No No No No Yes No No No Moxidectin Endectocides No No No No Yes Yes Yes Yes No No No No Yes No No No Selamectin Endectocides No No No No Yes No No Yes No No No No Yes No No Yes Cinoxacin Fluoroquinolones 17 Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Ciprofloxacin Fluoroquinolones Yes Yes Yes Yes Yes Yes Yes Yes No No No No Yes Yes Yes Yes Danofloxacin Fluoroquinolones Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Difloxacin Fluoroquinolones Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Enoxacin Fluoroquinolones No No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes Enrofloxacin Fluoroquinolones Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Flumequine Fluoroquinolones Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Lomefloxacin Fluoroquinolones Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Marbofloxacin Fluoroquinolones Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Nalidixic Acid Fluoroquinolones Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Norfloxacin Fluoroquinolones Yes Yes Yes Yes Yes Yes Yes Yes No No No No Yes Yes Yes Yes Ofloxacin Fluoroquinolones Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Orbifloxacin Fluoroquinolones Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Oxolinic Acid Fluoroquinolones Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Pipemidic Acid Fluoroquinolones Yes No No Yes Yes Yes Yes Yes No No No No Yes Yes Yes Yes Sarafloxacin Fluoroquinolones Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sparfloxacin Fluoroquinolones Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Lasalocid Ionophores 5 Yes No No Yes Yes Yes Yes Yes No No No No Yes Yes Yes Yes Monensin Ionophores Yes No No Yes Yes Yes Yes Yes Yes No No Yes Yes Yes Yes Yes Narasin Ionophores No No No No Yes Yes Yes Yes No No No No No No No No Nigericin Ionophores Yes No No Yes Yes Yes Yes Yes No No No No No No No No Salinomycin Ionophores Yes No No Yes Yes Yes Yes Yes No No No No Yes Yes Yes Yes Erythromycin Macrolides 8 No No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes Neospiramycin I Macrolides No No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes Oleandomycin Macrolides Yes Yes Yes Yes Yes Yes Yes Yes Yes No No Yes Yes Yes Yes Yes Roxithromycin Macrolides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Spiramycin I Macrolides No No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes Tilmicosin Macrolides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Tylosin A Macrolides No No No No Yes No No No No No No No Yes No No No Tylosin B Macrolides No No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes 2-methyl-4(5)-nitroimidazole Nitroimidazoles 13 Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Dimetridazole Nitroimidazoles Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Etanidazole Nitroimidazoles No No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes HMMNI Nitroimidazoles Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Ipronidazole Nitroimidazoles Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Ipronidazole-OH Nitroimidazoles Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Metronidazole Nitroimidazoles Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Metronidazole-OH Nitroimidazoles Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Nimorazole Nitroimidazoles Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Ornidazole Nitroimidazoles Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Ronidazole Nitroimidazoles Yes No No No Yes Yes Yes Yes Yes No No No Yes Yes Yes Yes

11 Target screening of 105 veterinary drug residues in milk using UHPLC/ESI Q-Orbitrap multiplexing data Table 4 (continued) Compound name Class Total number Target screening (Scenario I) a Target screening (Scenario II) a 1.0 μg/kg 10.0 μg/kg 1.0 μg/kg 10.0 μg/kg t R (± 0.5 t R and FI LS10 t R (± 0.5 t R and FI LST10 t R (± 0.5 t R and FI LS10 t R (± 0.5 t R and FI LST10 Ternidazole Nitroimidazoles Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Tinidazole Nitroimidazoles Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes 5-hydroxyflunixin NSAIDS 3 No No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes Flunixin NSAIDS Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Phenylbutazone NSAIDS Yes Yes Yes Yes Yes Yes Yes Yes No No No No Yes Yes Yes Yes Ampicillin Penicillins 6 No No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes Cloxacillin Penicillins Yes Yes Yes Yes Yes Yes Yes Yes Yes No No No Yes Yes Yes Yes Dicloxacillin Penicillins No No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes Oxacillin Penicillins Yes No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes Penicillin G Penicillins Yes Yes Yes Yes Yes Yes Yes Yes No No No No Yes Yes Yes Yes Penicillin V Penicillins No No No Yes Yes No No Yes No No No No Yes No No Yes Florfenicol Phenicols 2 Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Thiamphenicol Phenicols No No No No Yes Yes Yes No No No No No Yes Yes Yes No Dapsone Sulfonamides 26 Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfabenzamide Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfacetamide Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfachloropyridazine Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfadiazine Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfadimethoxine Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfadoxine Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfaethoxypyridazine Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfaguanidine Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes No No No No Yes Yes Yes Yes Sulfamerazine Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfameter Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfamethazine Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfamethizole Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfamethoxazole Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfamethoxypyridazine Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfamonomethoxine Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfamoxole Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfanilamide Sulfonamides Yes Yes No Yes Yes Yes Yes Yes No Yes No Yes Yes Yes Yes Yes Sulfanitran Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes No No Yes Yes Yes Yes Yes Sulfaphenazole Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfapyridine Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfaquinoxaline Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfathiazole Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfisomidine Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sulfisoxazole Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Trimethoprim Sulfonamides Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes 4-epitetracycline Tetracyclines 5 Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Chlortetracycline Tetracyclines Yes No No Yes Yes Yes Yes Yes No No No No Yes Yes Yes Yes Doxycycline Tetracyclines No No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes Oxytetracycline Tetracyclines Yes No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes Tetracycline Tetracyclines Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Cefamandole β-lactams 13 No No No No Yes No No No No No No No No No No No Cefapirin β-lactams Yes No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes Cefazolin β-lactams No No No No Yes Yes Yes No No No No No Yes No No No Cefoperazone β-lactams Yes No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes Cefotaxime β-lactams Yes No No No Yes Yes Yes Yes Yes No No No Yes Yes Yes Yes

12 5384 Wang J. et al. Table 4 (continued) Compound name Class Total number Target screening (Scenario I) a Target screening (Scenario II) a 1.0 μg/kg 10.0 μg/kg 1.0 μg/kg 10.0 μg/kg t R and FI LST10 t R and FI LS10 t R (± 0.5 t R and FI LST10 t R (± 0.5 t R and FI LS10 t R (± 0.5 t R (± 0.5 Cefoxitin β-lactams Yes Yes Yes Yes Yes Yes Yes Yes No No No No Yes Yes Yes Yes Cefquinome β-lactams No No No No Yes Yes Yes No No No No No Yes No No No Ceftiofur β-lactams Yes No No No Yes Yes Yes Yes Yes No No No Yes No No No Cefuroxime β-lactams No No No No Yes No No No No No No No Yes No No No Cephacetrile β-lactams No No No No Yes No No No No No No No No No No No Cephalothin β-lactams No No No No Yes No No No No No No No Yes No No No Cephradine β-lactams No No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes Desacetyl cephapirin β-lactams No No No No Yes Yes Yes Yes No No No No Yes Yes Yes Yes Total Percentage (%) a t R: retention time; FI: fragment ion; LST10: library search score threshold set at 10. b Column number. database for target screening. There were a total of 105 veterinary drugs from 11 groups or classes (Table 3,columns1and 2) in the in-house developed compound database. Since ecdb was based on corrected masses (not exact masses), it can only be used for screening, not for identification. (b) Retention time alignment Retention time (t R ) alignment is critical for the software to identify an incurred compound of interest in order to match fragments in a CDB (Fig. 2B) or a MS library (Fig. 2C)soas to reduce false positive or negative rates [15]. The individual t R of the veterinary drugs were obtained from their extracted ion chromatograms using the exact masses and were input into the CDB. The t R alignment was achieved using a single stable and well-characterized compound to calculate a correction constant, which was applied to all veterinary drugs in the CDB for its current batch of samples. This correction effectively aligned the retention times to the reference and this process was termed as Bt R alignment^. For example, ipronidazole, which eluted at ~4.64 min, was chosen as the reference for t R alignment. If the t R of ipronidazole was 4.64 min in the CDB while the observed t R was 4.70 min in a current batch, the time difference would be 0.06 min. Then, 0.06 min was added to the t R of each veterinary drug in the CDB. (c) Response threshold adjustment Response threshold is a parameter that can be assigned a value for each compound in the CDB [15]. The software integrates peaks when a sample peak response surpasses the threshold. When a generic default is set as the response threshold, this could result in an increased number of false positives. For example, when the response threshold was set at 30,000, the number of false veterinary drug positives was in a range of 8 to 29 in 30 blank samples (Fig. 3A).Thedetectedpeaksof false positives resulted from small adjacent peaks, interferences, and noisy baselines, or peaks from a trace amount of veterinary drug residues present in the Bblank samples^ that were above the response thresholds. It is essential to set appropriate response thresholds to reduce false positives and/or false negatives. It is well known that the ionization efficiency is compounddependent and matrix effects contribute to the responses of individual veterinary drugs in the electrospray ion source as a result of ion suppression and enhancement. For example, fluoroquinolones have much higher ionization efficiency than endectocides, penicillins, tetracyclines, and β-lactams (Fig. 3B). This is observed from the differences in peak areas of the veterinary drugs injected to the UHPLC/ESI Q-Orbitrap at the same concentration. Therefore, response thresholds should be set accordingly for each veterinary drug.

13 Target screening of 105 veterinary drug residues in milk using UHPLC/ESI Q-Orbitrap multiplexing data Fig. 3 UHPLC/ESI Q-Orbitrap target screening response threshold adjustment. (A) A number of false positives from blank samples of 10 milk blank matrices prepared in triplicate. Data acquisition: Full MS/mDIA. OPA: original peak area. (B) Response thresholds of individual veterinary drugs calculated as 2% original peak areas from 20 injections at 10 μg/l or ppb in solvent To set appropriate response thresholds for individual veterinary drugs, a mixture of standards, which was prepared at 10.0 μg/l in solvent, was injected 20 times on to the UHPL/ESI Q-Orbitrap, and data were acquired by Full MS/mDIA. Ten μg/l in solvent was equivalent to 10.0 μg/kg in sample when the final sample extracts were concentrated in 3:1. Note that 1.0 μg/l was not used since it was too low for some of the veterinary drugs to generate a quantifiable peak area or to be detected. After data acquisition and processing, the peak areas of the individual veterinary drugs were averaged. Due to potential matrix effects (mainly ion suppression) and instrument sensitivity variation from day-to-day, the values of 2%, 10%, 20%, or 40% Original Peak Areas (OPA) at 10.0 μg/l were tested for their applicability to determine the appropriate response thresholds to be used in this study. When response thresholds were set at 2%, 10%, 20%, or 40% OPA, the number of false positives was reduced significantly compared with those from the default of 30,000 (Fig. 3A). However, in order to detect

14 5386 Wang J. et al. veterinary drugs at 1.0 μg/kg, 2% OPA was used to set response thresholds for target screening (Fig. 3B). To reduce false negatives, certain response thresholds were set to 10,000 (Table 3, column 5) to help compensate for the low sensitivity of a veterinary drug in UHLPC/ESI Q-Orbitrap. Mass spectral library Mass spectral (MS) library was built by importing individual dd-ms 2 product ion spectra of veterinary drugs using Thermo Library Manger 2.0. The MS library matching was achieved using the BReverse Search^ function. Spectrum tolerance was 5 ppm. Score threshold was set at 10, 20, or 40 to explore its applicability for screening. The MS library matching served to provide additional information for confirmative screening (Fig. 2C). Validation and results The target screening method was validated using SANTE/ 11945/2015 [17] as a reference. The validation of a screening method based on targeted concentration levels (TCL) focused on detectability. The validation involved analysis of at least 20 samples spiked per TCL. Since the screening method is only intended to be used as a qualitative method, there are no requirements with regard to recovery of the analytes. In addition, there is no need for a strict criterion for the number of false positives detected. Ten blank milk samples were spiked at 1.0 μg/kg and 10.0 μg/kg in duplicate, respectively, and the experiment was repeated on three different days. For every veterinary drug, there were a total of 20 samples per batch that were used to validate the method and to meet the criterion that a veterinary drug had to be detected in at least 95% of the samples (i.e., an acceptable false-negative rate of 5%). After data acquisition, the results from samples spiked at 1.0 μg/kg and 10.0 μg/kg were processed by TraceFinder 3.3, and reviewed accordingly. Table 4 and Fig. 4 present the final validated results based on the criteria of mass accuracy ±5 ppm and t R ±0.5 min by either RTP or RTFI using 2% OPA as response thresholds. Taking Day 1 at 10.0 μg/kg as an example, 103 veterinary drugs were detected by RTP and 88 by RTFI (Fig. 4A, Data Table rows 1 and 3). The method proved to be reproducible as indicated by the repeatability of the results from 3 different days at the concentration level of 10.0 μg/kg but some variations at 1.0 μg/kg. In general, there were more veterinary drugs detected for both tentative (by RTP) and confirmative (by RTFI) screening at 10.0 μg/kg than at 1.0 μg/kg as a result of higher peak intensities at the higher concentration (Fig. 4A). At 10.0 μg/kg, almost all 105 veterinary drugs were found by RTP approach, which served the purpose of target screening. Detectability and its reproducibility According to both Commission Decision 2002/657/EC [19] and SANTE/11945/2015 [17], a qualitative screening method should be able to detect an analyte with a false compliant rate of <5% at the level of interest [19], or the screening method should be able to detect an analyte at the screening detection limit or reporting limit in at least 95% of the samples (i.e., an acceptable false-negative rate of 5%) [17]. The validation of detectability can be achieved using a batch of 20 samples that are analyzed in the same batch. However, it is not clear how the Breproducibility^ (between-day) of detectability should be evaluated for a qualitative or screening method. In our study, we repeated the detectability experiment on three different days (20 samples per batch). Within-day the criterion, that is, an analyte was detected in at least 95% of the samples (i.e., an acceptable false-negative rate of 5%), was followed. There were some issues with Breproducibility^ from the 3-day experiments. Some veterinary drugs were detected on day-1 but not in all 3 days, especially at a low concentration level such as 1.0 μg/kg. Therefore, the final list of veterinary drugs that are included in the method can be consolidated in two ways. One represents the best detection scenario to minimize false negatives (Scenario I). Another represents the assured detection scenario of non-compliant results (Scenario II). For Scenario I, when a veterinary drug was detected in at least 95% of the samples on one of the 3-day experiments, it was included in the method. The results in Fig. 4A (the last 4 th and 3 rd columns) and Table 4 (columns 4 6, 8 10) represent the best detection scenario for the method. In other words, the method was able to tentatively screen 77 veterinary drugs and to confirmatively screen 63 at 1.0 μg/kg (Fig. 4A, thelast4 th column); and to tentatively screen 105 veterinary drugs and to confirmatively screen 95 at 10.0 μg/kg (Fig. 4A,thelast3 rd column). For Scenario II, only when a veterinary drug was detected in at least 95% of the samples on all 3-day experiments, it was included in the method. This was the worst case scenario in terms of the number of veterinary drugs to be detected with assured certainty for target screening. Table 4 (columns 12 14, 16 18) and Fig. 4A (the last two columns) also indicated the veterinary drugs that were detected on all three days. In this case, the method was able to tentativelyscreen61veterinarydrugsandtoconfirmativelyscreen 54 at 1.0 μg/kg (Fig. 4A,thelast2 nd column); and to tentatively screen 101 veterinary drugs and to confirmatively screen 88 at 10.0 μg/kg (Fig. 4A, the last column). The number of veterinary drugs that are included in the final list aredifferentbasedonthedetectabilitybetweendays.sincethetarget screening method focused on detectability, Scenario II is recommended to be adopted to consolidate the final list for the method. As for MS library search, the BReverse Search^ was selected, and the spectrum tolerance was set at 5 ppm as a standard measure related to mass accuracy. The other parameter that was applicable for target screening was the Library Search score

15 Target screening of 105 veterinary drug residues in milk using UHPLC/ESI Q-Orbitrap multiplexing data Fig. 4 (A) UHPLC/ESI Q-Orbitrap target screening results based on compound database with tr tolerance ±0.5 min and mass accuracy 5 ppm; (B) target screening results based on mass spectral library with a score threshold at 10, 20, and 40, respectively. Samples were spiked at 1.0 and 10.0 μg/kg, respectively. Data acquisition: Full MS/mDIA. OPA: original peak area. RTP: retention time and precursor. FI: fragments. RTFI: retention time and fragment ion. LST: library search threshold. S I: Scenario I. S II: Scenario II. Threshold (LST). The LST can be selected arbitrarily between 0 and 100. In this study, we tested its value at 10, 20, and 40, and the results are shown in Fig. 4B. Apparently, the higher score threshold, the lower of numbers matched. The results from LST10 were similar to the RTP, FI, and RTFI values as observed in Fig. 4A, and specifically for each veterinary drug in Table 4. Therefore, library search with score threshold of 10 can potentially be used to provide additional information, likely visual spectra, for the certainty of target screening.

16 5388 Wang J. et al. Conclusions UHPLC/ESI Q-Orbitrap Full MS and mdia along with a compound database can serve as a practical approach for target screening. Full MS/dd-MS 2 acquired data of individual veterinary drugs to build a compound database and a MS library of 105 veterinary drugs. Accurate mass, retention time, and response threshold were three key parameters that needed to be corrected, aligned, and adjusted, respectively, when optimizing a compound database to reduce false negatives and/or false positives. Full MS/mDIA was used to acquire data for target screening of veterinary drug residues in sample. The screening parameters and criteria were based on either retention time (±0.5 and mass accuracy ( 5 of a precursor (RTP by Full MS), or retention time (±0.5 and mass accuracy ( 5 of the precursor and a fragment ion (RTFI by Full MS/mDIA). The screening method performance acceptability criterion was set at an acceptable false-negative rate of 5%, repeated on three different days. Based on the criteria and parameter settings used in this study, the RTP approach tentatively found at least 58% and 96% of the 105 veterinary drugs, whereas the RTFI confirmatively screened 51% and 84% in milk at 1.0 and 10.0 μg/kg, respectively. The RTP approach may be the preferred choice for target screening to avoid possible false negative results, whereas the RTFI approach increases the likelihood of screening a true positive of incurred residue in a sample. MS library can provide additional information or visual mass spectrum for target screen using appropriate score threshold. For future work, the exact masses of all fragments need to be determined for identification purpose in addition to target screening. Compliance with ethical standards Conflict of interest study. References There is no potential conflict of interest in current 1. Florea NF, Nightingale CH. Review of the pharmacodynamics of antibiotic use in animal food production. Diagn Microbiol Infect Dis. 2004;49: Mathur S, Singh R. Antibiotic resistance in food lactic acid bacteria a review. Int J Food Microbiol. 2005;105: Stolker AAM, Brinkman UAT. Analytical strategies for residue analysis of veterinary drugs and growth-promoting agents in food-producing animals a review. J Chromatogr A. 2005;1067: Berruga MI, Novés B, Molina MP, Román AM. Influence of cephalosphorins on the coagulation time of yogurt made from ewes' milk. Int J Dairy Tech. 2008;61: Schiffmann AP, Schütz M, Wiesner HU. False negative and positive results in testing for inhibitory substances in milk. Milchwissenschaft. 1992;47: Kaufmann A, Butcher P, Maden K, Walker S, Widmer M. Multiresidue quantification of veterinary drugs in milk with a novel extraction and cleanup technique: salting out supported liquid extraction (SOSLE). Anal Chim Acta. 2014;820: Ortelli D, Cognard E, Jan P, Edder P. Comprehensive fast multiresidue screening of 150 veterinary drugs in milk by ultraperformance liquid chromatography coupled to time of flight mass spectrometry. J Chromatogr B. 2009;877: Peters RJB, Bolck YJC, Rutgers P, Stolker AAM, Nielen MWF. Multi-residue screening of veterinary drugs in egg, fish and meat using high-resolution liquid chromatography accurate mass timeof-flight mass spectrometry. J Chromatogr A. 2009;1216: Stolker AA, Rutgers P, Oosterink E, Lasaroms JJ, Peters RJ, van Rhijn JA, et al. Comprehensive screening and quantification of veterinary drugs in milk using UPLC-ToF-MS. Anal Bioanal Chem. 2008;391: Wang J, Leung D. The challenges of developing a generic extraction procedure to analyze multi-class veterinary drug residues in milk and honey using ultra-high pressure liquid chromatography quadrupole time-of-flight mass spectrometry. Drug Test Anal. 2012;1(4 Suppl): Wang J, Leung D, Chow W, Chang J, Wong JW. Development and validation of a multiclass method for analysis of veterinary drug residues in milk using ultrahigh performance liquid chromatography electrospray ionization quadrupole Orbitrap mass spectrometry. J Agric Food Chem. 2015;63: Zhang Y, Liu X, Li X, Zhang J, Cao Y, Su M, et al. Rapid screening and quantification of multi-class multi-residue veterinary drugs in royal jelly by ultra performance liquid chromatography coupled to quadrupole time-of-flight mass spectrometry. Food Control. 2016;60: Reinholds I, Pugajeva I, Bartkevics V. A reliable screening of mycotoxins and pesticide residues in paprika using ultra-high performance liquid chromatography coupled to high resolution Orbitrap mass spectrometry. Food Control. 2016;60: Zomer P, Mol HGJ. Simultaneous quantitative determination, identification, and qualitative screening of pesticides in fruits and vegetables using LC-Q-Orbitrap-MS. Food Addit Contam A. 2015;32: Wang J, Chow W, Chang J, Wong JW. Development and validation of a qualitative method for target screening of 448 pesticide residues in fruits and vegetables using UHPLC/ESI Q-Orbitrap based on data-independent acquisition and compound database. J Agric Food Chem. 2017;65: Pérez-Ortega P, Lara-Ortega FJ, Gilbert-López B, Moreno- González D, García-Reyes JF, Molina-Díaz A. Screening of over 600 pesticides, veterinary drugs, food-packaging contaminants, mycotoxins, and other chemicals in food by ultra-high performance liquid chromatography quadrupole time-of-flight mass spectrometry (UHPLC-QTOF MS). Food Anal Method. 2017;10: European Union. Guidance document on analytical quality control and method validation procedures for pesticides residues analysis in food and feed. SANTE/11945/2015 Supersedes SANCO/12571/2013 Implemented by 01/01/ pesticides_mrl_guidelines_wrkdoc_11945.pdf. Accessed 10 Oct Wang J, Chow W, Chang J, Wong JW. Ultrahigh-performance liquid chromatography electrospray ionization Q-Orbitrap mass spectrometry for the analysis of 451 pesticide residues in fruits and vegetables: method development and validation. J Agric Food Chem. 2014;62: Commission Decision of 12 August 2002 implementing Council Directive 96/23/EC concerning the performance of analytical methods and the interpretation of results. 2002/657/EC. Off J Eur Communities 2002;L221:8 36

17 Target screening of 105 veterinary drug residues in milk using UHPLC/ESI Q-Orbitrap multiplexing data Jian Wang hasbeenworkingasa research scientist at the Calgary Laboratory with the Canadian Food Inspection Agency (CFIA) since His work focuses on method development and validation on analysis of chemical contaminant residues, including pesticides and veterinary drugs in food using emerging mass spectrometric technologies. He also develops statistical approaches to estimate measurement uncertainty based on method validation and quality control data using SAS program. James S. Chang has worked with ThermoFisher Scientific since His duties include application and development of laboratory instruments (gas, high performance liquid chromatography, liquid chromatography-tandem mass spectrometry, ultra-fast GC, multidimensional GC, micro-flow ultrahigh performance liquid chromatography, tandem mass spectrometry, and high resolution/high accuracy mass spectrometry) for food and environmental analysis. He has had more than 25 years of experience in the operation of environmental laboratories as a manager and a Director. Daniel Leung is a science laboratory evaluator at the Calgary Laboratory, Canadian Food Inspection Agency, Canada. He studied and obtained his B.Sc. in Chemistry at the University of Alberta. Soon after graduation, he joined the Government of Canada. He has worked in the field of pesticide residue analysis foralmost30years.whenheis not working in the lab, he likes to ride his bike all over Canada and Europe. Jon W. Wong is a research chemist at the Center for Food Safety and Applied Nutrition, United States Food and Drug Administration in College Park, Maryland, USA. He has been with the FDA since 2002 and is involved in the development of multi-residue analytical methods for foods and other agricultural commodities. Willis Chow is a science laboratory evaluator working in the Calgary Laboratory Research and Development Unit of the Canadian Food Inspection Agency. His work focuses on the detection and quantification of pesticide residues in food by both unit mass and high resolution mass spectrometry. He has been with the agency since 2003.