Wide-Scope Screening Method for Multiclass Veterinary Drug Residues in Fish, Shrimp, and Eel Using Liquid Chromatography-Quadrupole High-Resolution Mass Spectrometry

Abstract

A screening method for veterinary drug residues in fish, shrimp, and eel using LC with a high-resolution MS instrument has been developed and validated. The method was optimized for over 70 test compounds representing a variety of veterinary drug classes. Tissues were extracted by vortex mixing with acetonitrile acidified with 2% acetic acid and 0.2% p-toluenesulfonic acid. A centrifuged portion of the extract was passed through a novel solid phase extraction cartridge designed to remove interfering matrix components from tissue extracts. The eluent was then evaporated and reconstituted for analysis. Data were collected with a quadrupole-Orbitrap high-resolution mass spectrometer using both nontargeted and targeted acquisition methods. Residues were detected on the basis of the exact mass of the precursor and a product ion along with isotope pattern and retention time matching. Semiquantitative data analysis compared MS¹ signal to a one-point extracted matrix standard at a target testing level. The test compounds were detected and identified in salmon, tilapia, catfish, shrimp, and eel extracts fortified at the target testing levels. Fish dosed with selected analytes and aquaculture samples previously found to contain residues were also analyzed. The screening method can be expanded to monitor for an additional >260 veterinary drugs on the basis of exact mass measurements and retention times.

INTRODUCTION

Aquaculture is a growing industry anticipated to supply approximately 100 million tons, or >60%, of the fish destined for human consumption by 2030. Many types of veterinary drugs may be administered to fish in an aquaculture environment to treat disease or proactively prevent infection.^1,2 Traditionally, analytical methods were developed to monitor for one residue, or for several analytes from a specific class of drugs, in a single species of fish or shellfish. More recently, multiclass methods have been developed using liquid chromatography (LC) with tandem mass spectrometry (MS),^3–5 but these methods are still limited to targeted analytes.

Using high-resolution mass spectrometry (HRMS) instruments, a virtually unlimited number of compounds can be simultaneously analyzed because full-scan data are collected rather than preselected ion transitions corresponding to specific compounds. Selectivity is achieved by taking advantage of the instrument’s ability to provide very accurate mass measurements. Residue identification can be based on calculated exact masses of protonated molecules and fragment ions, relative isotopic abundances, and retention times. This can lead to the development of methods that can monitor for a wide scope of residues and contaminants, allowing regulatory agencies to be more proactive in discovering possible adulteration of the food supply including aquacultured products.

The use of HRMS technology for detecting drug residues in aquaculture and other foods has been reviewed^6,7 and includes methods that utilize time-of-flight (ToF)^8–11 or orbital ion trap (Orbitrap)^12–14 HRMS. These methods have demonstrated the ability to detect, identify, and quantitate an increasing number of analytes. However, some of these early HRMS methods were sometimes limited by insufficient mass resolution to accurately measure exact mass in complex matrices and a lack of sensitivity to detect residue levels. The Q-Exactive is a hybrid quadrupole-Orbitrap HRMS instrument that has the advantage of operating at increased resolution and the ability to collect product ion spectra with or without initial precursor isolation using the quadrupole filter.

Using a MS detector with this capability further enhances the need for the most universal extractant and cleanup method possible. Ideally, the method should also be simple and quick. Although a large number of veterinary drug residue LC-MS methods and cleanup strategies are available in the literature for fish and shrimp,^5,15 specific lipid cleanup technologies have recently been introduced. Removal of lipids is especially important for high fat-containing matrices such as salmon and other fish. Fats (specifically phospholipids) can be significant HPLC column contaminants and can also contribute to severe matrix suppression in signal response for MS detection. New commercially available cleanup techniques are designed to eliminate these interferences without severe loss of analyte recovery.^16,17 These new sample cleanup products were evaluated along with the choice of initial extraction solution to optimize a procedure suitable for a wide variety of analytes at residue levels.

The aim of this research was to be able to screen and identify a wide scope of veterinary drug residues in several types of fish matrices. An optimized extraction method was used with an LC separation combined with a Q-Exactive (Orbitrap) HRMS. The method was validated according to U.S. Food and Drug Administration (FDA) guidelines^18,19 using the representative compounds listed in Table 1. In addition to monitoring fish extracts for these test compounds, the full scan HRMS data could be compared to a compound database containing about 260 additional (>330 total) veterinary drugs to significantly expand the number of residues that might be detected in any given sample.

Table 1.

Test Compounds

analyte	class	TTL (μg/kg)	RT (min)	formula	MH+		fragment ions
doramectin (DOR)	avermectin	200a	11.3	C50H74O14	921.4971b	449.2298	777.4126
emamectin B1a (EMA)	avermectin	200a	9.6	C49H75NO13	886.5311	82.0651	158.1176	302.1962
ivermectin B1a (IVER)	avermectin	200a	12.0	C48H74O14	897.4971b	609.3398	753.4184
amoxicillin (AMOX)	β-lactam	100	1.9	C16H19N3O5S	366.1118	114.0372	208.0427	349.0853
ampicillin (AMP)	β-lactam	25	4.2	C16H19N3O4S	350.1169	106.0651	114.0372	160.0427
aspoxicillin (ASP)	β-lactam	25	2.6	C21H27N5O7S	494.1074	160.0427	250.1186	366.1118
cloxacillin (CLOX)	β-lactam	25	9.0	C19H18ClN3O5S	436.0729	160.0427	277.0375
dicloxacillin (DICLOX)	β-lactam	25	9.6	C19H17Cl2N3O5S	470.0339	160.0427	310.9985
oxacillin (OXAC)	β-lactam	25	8.5	C19H19N3O5S	402.1118	114.0372	160.0427	243.0764
penicillin G (PEN G)	β-lactam	25	7.5	C16H18N2O4S	335.1060	114.0372	160.0427	176.0706
penillic acid	β-lactam	NAc	4.9	C16H18N2O4S	335.1060	128.0528	160.0427	289.0997
albendazole (ALB)	benzimidazole	50	7.0	C12H15N3O2S	266.0958	159.0427	191.0148	234.0696
albendazole sulfoxide (ALB SULF)	benzimidazole	50	5.0	C12H15N3O3S	282.0907	208.0175	240.0437
fenbendazole (FEN)	benzimidazole	50	8.4	C15H13N3O2S	300.0801	159.0427	268.0539
fenbendazole sulfone (FEN SULF)	benzimidazole	50	7.3	C15H13N3O4S	332.0700	300.0437
cephapirin (CEPH)	cephalosporin	25	3.4	C17H17N3O6S2	424.0632	152.0165	292.0573
brilliant green (BG)	dye	1	9.5	C27H33N2	385.2638d	297.1386	341.2012
crystal violet (CV)	dye	1	9.0	C25H30N3	372.2434d	251.1543	356.2121
leucocrystal violet (LCV)	dye	1	5.6	C25H31N3	374.2591	239.1543	253.1699	358.2278
leucomalachite green (LMG)	dye	1	8.2	C23H26N2	331.2169	194.0964	239.1543	315.1856
malachite green (MG)	dye	1	8.1	C23H25N2	329.2012d	208.1121	313.1699
ciprofloxacin (CIP)	fluoroquinolone	5	4.6	C17H18FN3O3	332.1405	245.1085	288.1507
danofloxacin (DANO)	fluoroquinolone	5	5.0	C19H20FN3O3	358.1562	283.1241	314.1663	338.1499
difloxacin (DIFLOX)	fluoroquinolone	5	5.4	C21H19F2N3O3	400.1467	299.0990	356.1569
enrofloxacin (ENRO)	fluoroquinolone	5	5.1	C19H22FN3O3	360.1718	245.1085	316.1820	342.1612
norfloxacin (NOR)	fluoroquinolone	5	4.7	C16H18FN3O3	320.1405	276.1507	302.1299
sarafloxacin (SAR)	fluoroquinolone	5	5.4	C20H17F2N3O3	386.1311	299.0990	342.1413
methyl testosterone (M TET)	Hhormone	0.8	9.4	C20H30O2	303.2319	97.0648	109.1011
lincomycin (LIN)	lincomycin	50	3.8	C18H34N2O6S	407.2210	126.1277	359.2214
azithromycin (AZI)	macrolide	50	5.4	C38H72N2O12	749.5158	158.1176	591.4210
erythromycin A (ERY)	macrolide	50	6.6	C37H67NO13	734.4685	83.0491	158.1176	576.3742
erythromycin dehydrated	macrolide	NAc	7.2	C37H65NO12	716.4580	158.1176
spiramycin (SPIRO)	macrolide	50	5.5	C43H74N2O14	843.5213	174.1125	540.3136
tilmicosin (TIL)	macrolide	50	6.0	C46H80N2O13	435.2903e	174.1125	522.3789	695.460
tylosin A (TYL)	macrolide	50	7.0	C46H77NO17	916.5264	174.1125
ketoconazole (KETO)	nitromidazole	10	7.2	C26H28Cl2N4O4	531.1560	82.0525	489.1455
metronidazole (MNZ)	nitromidazole	10	2.0	C6H9N3O3	172.0717	82.0525	128.0455
florfenicol amine (FFA)	phenicol	50f	1.33	C10H14FNO3S	248.0751	104.0632	130.0651	230.0646
ormetoprim (ORM)	potentiator	10	4.6	C14H18N4O2	275.1503	123.0665	259.1190
trimethoprim (TRIMETH)	potentiator	10	4.3	C14H18N4O3	291.1452	123.0665	230.1162
ethoxyquin (ETHOX)	preservative	50	7.7	C14H19NO	218.1539	148.0757	176.1070
flumequine (FLU)	quinolone	10	7.82	C14H12FNO3	262.0874	202.0299	244.0768
nalidixic acid (NAL)	quinolone	10	7.6	C12H12N2O3	233.0921	187.0502	215.0815
oxolinic acid (OXO)	quinolone	10	6.5	C13H11NO5	262.0710	216.0291	244.0604
sulfacetamide (SAA)	sulfonamide	10	2.2	C8H10N2O3S	215.0485	92.0495	108.0444	156.0114
sulfachloropyridazine (SCP)	sulfonamide	10	5.8	C10H9ClN4O2S	285.0208	92.0495	108.0444	156.0114
sulfaclozine (SULC)	sulfonamide	10	6.9	C10H9ClN4O2S	285.0208	92.0495	108.0444	156.0114
sulfadiazine (SDZ)	sulfonamide	10	2.9	C10H10N4O2S	251.0597	92.0495	108.0444	156.0114
sulfadimethoxine (SDM)	sulfonamide	10	7.0	C12H14N4O4S	311.0809	108.0444	156.0114	156.0768
sulfadoxine (SDX)	sulfonamide	10	6.1	C12H14N4O4S	311.0809	92.0495	108.0444	156.0114
sulfaethoxypyridazine (SEP)	sulfonamide	10	6.2	C12H14N4O3S	295.0859	92.0495	108.0444	156.0114
sulfamerazine (SMR)	sulfonamide	10	4.1	C11H12N4O2S	265.0754	92.0495	108.0444	156.0114
sulfamethazine (SMZ)	sulfonamide	10	4.8	C12H14N4O2S	279.0910	92.0495	108.0444	156.0114
sulfamethoxazole (SMX)	Sulfonamide	10	6.1	C10H11N3O3S	254.0594	92.0495	108.0444	156.0114
sulfamethoxypyridazine (SMP)	sulfonamide	10	5.1	C11H12N4O3S	281.0703	108.0444	126.0662	156.0114
sulfamonomethoxine (SULFMON)	sulfonamide	10	5.6	C11H12N4O3S	281.0703	92.0495	108.0444	156.0114
sulfapyridine (SPD)	sulfonamide	10	4.0	C11H11N3O2S	250.0645	108.0444	156.0114	184.0869
sulfaquinoxaline (SQX)	sulfonamide	10	7.1	C14H12N4O2S	301.0754	92.0495	108.0444	156.0114
sulfathiazole (STZ)	sulfonamide	10	3.9	C9H9N3O2S2	256.0209	92.0495	108.0444	156.0114
chlortetracycline (CTC)	tetracycline	100g	5.6	C22H23ClN2O8	479.1216	444.0845
doxycycline (DC)	tetracycline	100g	5.9	C22H24N2O8	445.1605	154.0499	410.1234	428.1340
oxytetracycline (OTC)	tetracycline	100g	4.7	C22H24N2O9	461.1555	154.0499	426.1183
tetracycline (TC)	tetracycline	100g	4.8	C22H24N2O8	445.1605	154.0499	410.1234
negative ion analyte	class	TTL (μg/kg)	RT (min)	formula	MH-		fragment ions
chloramphenicol (CAP)	phenicol	0.3	6.5	C11H12Cl2N2O5	321.00505	152.0353	176.0353	257.0335
florfenicol (FF)	phenicol	5	6.0	C12H14Cl2FNO4S	355.99319	185.0278	335.9870
thiamphenicol (THIAM)	phenicol	5	4.6	C12H15Cl2NO5S	353.99752	185.0278	290.0259
toltrazuril (TOLT)	toltrazuril	50	10	C18H14F3N3O4S	424.05843	316.98132	404.97665
toltrazuril sulfone (TOLT SULF)	toltrazuril	50	9.9	C18H14F3N3O6S	456.04826	none
toltrazuril sulfoxide (TOLT SULFX)	toltrazuril	50	9.2	C18H14F3N3O5S	440.05335	371.05781
diflubenzuron (DIFLU)	benzylurea	50	10	C14H9ClN2O2F2	309.02478	242.98601	289.0184
lufenuron (LUF)	benzylurea	50	10.4	C17H8Cl2F8N2O3	508.97115	174.95972	325.9591	488.96492
teflubenzuron (TEFLU)	benzylurea	50	10.3	C14H6Cl2F4N2O2	378.96697	195.95378	338.95451

^aCurrent FDA program recommends TTL of 10 μg/kg.²⁰

^bMNa⁺.

^cNA, not applicable (degradant).

^dM⁺.

^eMH₂²⁺.

^fFDA tolerance is 1 mg/kg for FFA as marker residue in aquaculture.²⁰

^gFDA tolerance is 2 mg/kg for sum of OTC, CTC, and TC in finfish and lobster.²⁰

MATERIALS AND METHODS

Standard Preparation

Individual stock standards were made in methanol, except for β-lactams, which were dissolved in water or acetonitrile/water depending upon solubility. Oxolinic acid was prepared in acetonitrile. All stock standard solutions were made at a concentration of approximately 100 μg/mL as the free base or acid. Stock standards for the β-lactams were stored at −25 °C; all others were stored at 4 °C. Two different spiking standard mixes (“stable” and “nonstable”) for positive ion compounds were made. A 1X target testing level (TTL) spike was made by adding portions (see below) of both mixtures to control tissue. The “nonstable” analyte spiking mix contained the β-lactams, tetracyclines, cephapirin, and dye compounds as listed in Table 1. The nonstable standard spiking mix was prepared by combining an amount (a volume equal to 20 times the TTL level for each compound, corrected for the exact concentration) of each individual stock standard and diluting to 20 mL with acetonitrile. To prepare a 1X TTL spike containing all of the nonstable compounds, 20 μL of this standard mix was added to 2 g of control tissue. This “nonstable” spiking mix standard is stable for 6 months at −25 °C. The “stable” standard spiking mix consists of all remaining positive ion compounds listed in Table 1. Similarly, the stable analyte mix was also prepared by combining an amount (a volume equivalent to 20X the TTL for each compound) of each individual stock standard and diluting to 40 mL with acetonitrile. To prepare a 1X TTL spike containing all of the stable compounds, 40 μL of this standard mix was added to 2 g of control tissue. This “stable” spiking mix standard is stable for 1 year at −25 °C. A separate fortification mixture for the negative ion compounds was made in the same manner as the stable spiking mix. Samples were fortified with negative ion compounds by adding 40 μL of that standard mix to 2 g of control tissue.

To prepare a solvent standard equivalent to 1X TTL in tissue, 100 μL of the stable standard mix and 50 μL of the nonstable standard mix were combined and diluted to 5 mL with 90:10 water/acetonitrile. (This assumes an extraction scheme of 2 g→ 10 mL then 2 mL → 0.4 mL). If an additional acetonitrile injection was needed for salmon, then an equivalent acetonitrile solvent standard was additionally prepared by diluting the previous solvent standard 1:5 with acetonitrile. A negative ion 1X TTL equivalent solvent standard was prepared by diluting 100 μL of the negative mixed standard mix to 5 mL with 90:10 water/acetonitrile.

Sample Extraction

Homogenized tissue (2.0 ± 0.05 g) was weighed into a 50 mL polypropylene tube, and spiking standard mixes were added as appropriate and allowed to mix with the tissue for at least 5 min. The tissue was extracted with 8 mL of extraction solution consisting of 0.2% p-toluenesulfonic acid (p-TSA) monohydrate (w/v) and 2% glacial acetic acid (v/v) in acetonitrile. The samples were vortex mixed for 30 min using a multitube vortex mixer at a setting speed of 2500 rpm and then centrifuged for 7 min (4 °C) at minimum of 17,000 RCF (g). A portion (3 mL) of the extract was transferred to an Oasis PRiME HLB 6 cc (200 mg) extraction cartridge with a 15 mL polypropylene tube underneath. The samples were allowed to gravity drain (ca. 10 min) through the cartridges. The remaining few drops of extractant were gently pushed out through the SPE tube with a pipet bulb. (This should give just over 2 mL of liquid at this point.) If the tissue was salmon, 100 μL of the extract was transferred into an LC vial for an acetonitrile injection. The remaining portion of the extract was taken to near dryness under nitrogen stream at 55 °C (a drop of liquid remaining in the 15 mL tube was acceptable). The extract was then reconstituted with 400 μL of 10% acetonitrile in water (v/v), mixed, and centrifuged at a minimum of 28,900 RCF (g) for 7 min. Finally, an aliquot of 300 μL was carefully removed from the polypropylene tube, leaving particulates behind, and transferred into a LC vial. Overall there was no net dilution of concentration of the sample through the extraction procedure. (A sample fortified at 10 μg/kg = 10 ng/mL in LC vial.)

Incurred Samples

Incurred fish samples were provided by the Center for Veterinary Medicine Office of Research. Two tilapia were dosed with 1 mg/kg body weight of sulfadiazine with 1 day of depuration. Two other tilapia were dosed with 5 mg/kg body weight of sulfadiazine and 1 mg/kg body weight of trimethoprim and were sacrificed at 3 and 4 days. Two catfish were dosed with 5 mg/kg body weight of enrofloxacin with a 6 day withdrawal. Control tilapia and catfish were grown and harvested concurrently with these dosed animals. Catfish and salmon from an earlier (2014) dosing study with triphenylmethane dyes were also tested with this method. These fish were exposed to a water bath containing a mixture of 2 μg/L of malachite green, crystal violet, and brilliant green for 1 h followed by 1 h of depuration in clean water.

Instrument Acquisition Methods

Instrumentation

The instrument used was a Thermo Q-Exactive Orbitrap high-resolution mass spectrometer (HRMS) with a heated electrospray ionization source coupled to a Thermo Ultimate 3000 LC system. Thermo XCalibur software (v. 3.0.63) was used for data acquisition and preliminary data analysis; data analysis was also performed using TraceFinder software (v. 3.2).

MS Acquisition Programs

The instrument was calibrated for mass accuracy according to the manufacturer’s recommendations at least once a week. The tuning method optimized signals for a majority of the test compounds with the LC conditions described below. General MS acquisition parameters were as follows: spray voltage, 4 kV (positive), 2.5 kV (negative); S-Lens RF level, 50; capillary temperature, 350 °C; auxiliary gas temperature, 325 °C; gas flow rates (N₂, arbitrary units), sheath, 50; auxiliary, 10; sweep, 2. Other general MS parameters included acquisition time, 0–12.5 min; lock mass, OFF; and chrom peak, 15 s.

Two different types of acquisition programs were used to analyze the fish extracts. All ion fragmentation (AIF) was used for initial data acquisition. AIF is a nontargeted method in which a full scan MS is followed by a MS² scan where all precursors are allowed into the high collision dissociation (HCD) cell to form product ions simultaneously. The settings for AIF were as follows: MS¹, 70K resolution; 3e⁶ automatic gain control target; maximum inject time, 200 ms; m/z 150–1000 scan range. MS², 70K resolution; 3e⁶ automatic gain control target; maximum inject time, 200 ms; m/z 80–1000 scan range; normalized collision energies of 10, 30, and 50.

A second set of data (separate injection of fish extract) was obtained using Data Dependent MS² (DDMS²) data acquisition. With this program, MS² data were collected when a precursor ion from a predefined “inclusion list” was detected (from a full MS¹ scan) above a set threshold. When that occurred, the quadrupole filtered the precursor ion into the HCD cell using a limited m/z window to produce fragment ions related to that compound. The inclusion list for positive ion analytes contains approximately 290 of the compounds from the larger database, including all test compounds and other analytes for which a retention time was known (1.5 min windows were used in the list). Some of the more obscure analytes in the larger compound database were not included so that the instrument did not waste analysis time triggering spectra for those compounds. Analytes can be added to the inclusion list as needed. The DDMS² inclusion list for negative ion compounds was much smaller, containing only compounds included in the validation (Table 1), although this list can also be expanded as needed. The operating parameters for this data acquisition for MS¹ are 70K resolution, 1e⁵ automatic gain control target, maximum inject time of 200 ms, and m/z 150–1000 scan range. The operating parameters for DDMS² are 17.5K resolution, 1e⁵ automatic gain control target, maximum inject time of 50 ms, loop count of 3, isolation width of 4 m/z, and normalized collision energies of 10, 30, and 50. Other data-dependent settings are as follows: underfill ratio, 0.5%; calculated intensity threshold, 1e⁴; apex trigger, 3–6 s; dynamic exclusion, 6 s. Negative ion data were collected with separate injections using AIF and DDMS² acquisition.

Chromatography

LC separation was performed using a Supelco Ascentis Express C18 (7.5 cm × 2.1 mm, 2.7 μm) fused-core reversed-phase column. The mobile phase consisted of 0.1% formic acid in water (A) and acetonitrile (B) at a flow rate of 0.3 mL/min. The LC gradient program was initialized at 5% B and held for 1.5 min and then ramped to 50% B from 1.5 to 8.5 min, followed by a ramp to 99% B from 8.5 to 9 min, and then was held at 99% B from 9 to 12 min. The mobile phase was returned to 5% B from 12 to 12.5 min, and the column was re-equilibrated for an additional 2 min. The total LC run time was 14.5 min; MS data were collected for 12.5 min (no divert valve was used). The column temperature compartment was kept at 30 °C, and the autosampler tray temperature was maintained at 10 °C. The LC injection volume was 10 or 20 μL for the AIF or DDMS² MS acquisition program, respectively. Multiple injections of a high-level dye standard may be needed to condition a new column to detect low levels of leucocrystal violet and leucomalachite green.

Data Analysis

Several stages of data analysis were performed. First, the AIF data were analyzed to determine if the test compounds were identified and present at concentrations above the threshold cutoff level (>50% TTL). AIF data could also be compared to the larger veterinary drug compound database to determine if additional analytes (beyond the test compounds) were in a sample. The product ion spectra from DDMS² data could also be evaluated. Negative ion data were evaluated using the same processes.

Limit Testing and Confirmation of Identity for Test Compounds Using AIF Data

AIF data from full MS¹ scans were used for initial screening of test compounds. A Thermo TraceFinder “Quantitative Method” was established to provide data for the test compounds listed in Table 1, including a few degradants (e.g., penillic acid and dehydrated erythromycin). The test compounds in the “Quantitative Method” were a subset of analytes imported from the larger compound database (N > 330), which contains information for retention time and exact masses of fragment ions.

To be qualitatively identified, the precursor ions must be present (signal-to-noise > 3) and match theoretical exact mass within a 5 ppm mass tolerance. The data analysis program searched for residues within a time window of 60 s (30 s on each side of specified retention time), but a narrower retention time match (±0.1 min) to a standard injected the same day was typically observed. Fragment detection was also required with at least one fragment ion with 500 count minimum intensity threshold within a 10 ppm maximum mass deviation window. These criteria are consistent with FDA guidance.¹⁹ The isotope match feature was also enabled with a 70% fit threshold, 5 ppm mass deviation, and 10% intensity deviation allowance. A sample would be considered presumptive positive for a test compound if the qualitative criteria were met and the signal was ≥50% as compared to the matrix-extracted standard fortified at the TTL.

Expanded Screening for Additional Veterinary Drug Residues

A Thermo TraceFinder “Screening Method” was used to search for additional residues beyond the test compounds. Data collected using AIF were compared to a compound database containing >330 potential veterinary drug residues, including metabolites and minor components. New compounds are continuously being added to the database, and experimental data for retention time and fragment ions are included for a majority of these compounds. Criteria used in the “Screening Method” were 3 ppm mass tolerance for the precursor ion, >100 signal-to-noise ratio, and a signal of >5000 counts for initial detection. To identify a residue by the screening method, a retention time window match within 60 s and a minimum of one fragment ion with an intensity threshold of >500 counts and a mass tolerance within 10 ppm were required. The retention time and fragment ion criteria could be ignored if not defined in the database. The isotope pattern match option was used to filter out false detects.

Additional Qualitative Data Analysis from DDMS²

The extracts were analyzed in a separate LC-MS injection using DDMS² data acquisition. After a full MS¹ scan, product ion spectra were collected after precursor isolation for analytes in an inclusion list if the signal from the MS¹ ion met data-dependent triggering requirements. Residue findings were evaluated for identification criteria using the same data analysis methods described above for AIF data. Product ion spectra produced by DDMS² data acquisition could also be manually evaluated using XCalibur QualBrowser software for consistency with solvent- and/or matrix-extracted standards. Alternatively, spectra could be compared to commercially available libraries.

Validation

The method was validated according to the FDA Office of Food and Veterinary Medicine Guidelines for Chemical Method Validations v. 2 for “limit testing”¹⁸ with veterinary drugs from a variety of chemical classes in representative matrices. A summary of the fortification samples generated for the validation is included in the Supporting Information. Salmon and tilapia were used for the initial validation. The majority of the replicates were analyzed at the TTL (1X) to generate variance data at that level to set an appropriate threshold cutoff to determine if samples should be considered presumptive positive. Fortification samples at 2X, 0.5X, and 0.1X of the TTL were also analyzed along with blank matrix samples to determine the lowest minimum detectable and confirmation levels, as well as the rates of false-positive and false-negative results. The method was also applied, with fewer overall replicates, to catfish, shrimp, and eel to demonstrate matrix extension to these species. Incurred samples of tilapia (SDZ and SDZ + TMP), catfish (ENRO, dyes), and salmon (dyes), as well as regulatory samples that had been found to contain residues, were also tested. A second set of validation samples was generated for negative ion compounds. Shrimp and salmon were the primary matrices validated for negative ion analytes, as these are the species in which phenicol or benzyl urea residues are expected to be found, respectively.

RESULTS AND DISCUSSION

Compounds and Testing Levels

A list of veterinary drug residues of importance in aquacultured species (along with their target testing levels) is given in Table 1. The target testing level (TTL) or “1X” is not necessarily considered an official tolerance or action level. The majority of these compounds are not approved for use in fish in the United States, so technically any amount found and identified would be violative. These values are meant to provide a reasonable concentration at which a residue may be expected to be consistently detected and identified using this screening method. For ease of sample preparation and data reporting, it is helpful if TTLs for the test compounds are all within a practical range (5–50 μg/kg). Some analytes will have TTLs higher (tetracyclines) or lower (methyl testosterone and dyes) than that range due to previously established levels of interest.²⁰ In addition, for some compounds, such as amoxicillin and the avermectins, the TTLs may be >100 μg/kg due to higher detection limits for these analytes in this method. The minimum detectable limits for all compounds extracted from fish was determined, and for most compounds were found to be considerably lower (0.1X) than the TTL.

Optimization of Extraction Procedure

Extraction Solvent

The list of analytes shown in Table 1 includes triphenylmethane dyes and avermectins. These two classes of compounds greatly increase the difficulty of choosing an acceptable sample extractant and cleanup procedure. Previous work⁴ summarizes the difficulty of including dyes in a multiresidue method. In addition, it is desirable that any chosen extractant be able to extract very polar analytes such as florfenicol amine and metronidazole as well as several nonpolar compounds. The goal was to develop a simple cleanup procedure to apply to as many of the compounds in this list as possible.

Methanol or acetonitrile as the primary organic component in a meat or fish extractant is most widely reported in the literature. Water (with or without buffer salts) is not effective for extracting nonpolar analytes. Acetonitrile is usually preferred because of its ability to precipitate proteins in tissue. The use of acetonitrile only (with no other purification steps) as a veterinary drug extractant for tissue has been published.²¹ Acetonitrile has also been used (in combination with secondary hexane partitioning) to extract residues from catfish, salmon, and trout.²² However, when 100% acetonitrile (without acid modifiers) was tried for our list of analytes, recoveries for crystal violet and fluoroquinolones were very low (<10%). Current veterinary drug residue methods used by the U.S. Department of Agriculture (USDA)^23,24 for bovine muscle use 4:1 acetonitrile/water for extraction. The method has some advantages, including extracting a wide range of compounds and allowing very rapid sample throughput. However, although this extractant appeared to extract avermectins successfully in bovine muscle, the 80% acetonitrile solution was too polar to successfully extract ivermectin from high-fat salmon. In addition, the final extract composition described in the USDA method (70% acetonitrile) gave poor peak shape for polar, early-eluting analytes using the rapid reversed-phase chromatographic gradient developed for this screening method. Further dilution with water of the final extract would improve chromatography, but would sacrifice sensitivity.

Our initial investigations focused on using acetonitrile containing acid modifiers as an extractant. Acetonitrile with 0.1% acetic acid has been used²⁵ to extract residues from salmon and was successful in extracting emamectin and malachite green. Acetonitrile with 1% acetic acid has also been used²⁶ for shrimp with similar results. In initial experiments with salmon, degradation of several β-lactams was observed when 0.2 or 1% formic acid in acetonitrile was used as an extractant. Formic acid in water is known²⁷ to cause rapid degradation of monobasic penicillins. Furthermore, formic acid in acetonitrile did not extract avermectins well from salmon and had an adverse effect on macrolides. For these reasons, acetic acid was chosen to acidify the acetonitrile extractant. By increasing acetic acid from 0.1 to 2% in the acetonitrile extractant, the recoveries for fluoroquinolones, avermectins, and penicillins doubled. Previous work in our laboratory⁴ showed that the addition of p-TSA improved the recovery of dyes and fluoroquinolone compounds, so this acid was added to the 2% acetic acid in acetonitrile extractant. Provided that acetic acid was also present in the extractant, the p-TSA did not seem to appreciably degrade penicillins. The final extraction solvent chosen was acetonitrile with 2% acetic acid with 0.2% p-TSA. A comparison of recoveries for representative analytes using variable amounts of acid modifiers in tilapia (using Oasis PRiME HLB) is shown in Figure 1.

Figure 1.

Comparison of selected analyte recoveries in tilapia at 1X TTL with different levels of acid modifiers (normalized to 0.2% p-TSA and 2% acetic acid (AA) recoveries).

Sample Cleanup

Two new types of cleanup techniques designed specifically for high-fat samples were evaluated. Agilent’s EMR Lipid system is a dispersive SPE technique involving two primary sorbents: a water-activated sorbent designed to specifically trap fats containing >5-carbon aliphatic chains and a second sorbent containing magnesium sulfate and sodium chloride. A published study¹⁶ using this technique for the analysis of veterinary drug residues in bovine liver was initially evaluated for salmon. Although this procedure worked well for many analytes in Table 1, some were not recovered. The 5% formic acid solution used as the extractant in the application note did not work well for some penicillins or for the avermectins in salmon. The magnesium sulfate sorbent also greatly lowered the recovery of some tetracyclines.

A Waters Oasis PRiME HLB SPE cartridge was also evaluated as a cleanup tool. As described in a published method¹⁷ for the analysis of veterinary residues in pork tissue, a portion of tissue extract is introduced into the SPE cartridge (no conditioning is required) and collected by gravity drain. For this method, two changes were made to accommodate the analyte list in Table 1. First, the extractant of formic acid, acetonitrile, and water was changed to the final optimized extractant described above. Second, to improve sensitivity, a portion of the extract was evaporated (instead of diluted per original method) and reconstituted with 9:1 water/acetonitrile to allow for successful reversed-phase chromatography. For the newest triple-quadrupole mass spectrometers, dilution may become less of an issue, but for full scan data acquisition obtained using the Orbitrap, a 10 min evaporation step was included to concentrate the extract. The larger 200 mg SPE sorbent size was used to collect a larger volume of extractant for evaporation and concentration. Various HPLC filters were evaluated for use in filtering the final extract before injection. Although polyvinylidene fluoride filters worked the best (compared to nylon or PTFE), we observed inconsistent or no recovery after filtration in solvent spikes for some analytes. For this reason the final extract was not filtered but was instead centrifuged to help remove any remaining particulates after the evaporation and reconstitution step.

For salmon, some compounds (avermectins, brilliant green, and crystal violet) were lost when extracts were evaporated and reconstituted most likely because they partitioned into insoluble lipid material. This necessitated a separate acetonitrile injection to detect those compounds in salmon as described earlier in the experimental section. Because the relevant compounds for the acetonitrile injection elute late in the reversed phase chromatographic system, good peak shapes and consistent retention times are achieved despite the high organic content. This extra injection was also required for nonpolar negative ion compounds (lufenuron, teflubenzuron, toltrazuril) fortified in salmon. A portion of the acetonitrile extract (before drying) was also tested with the other fish matrices. In general, this step was not necessary with the less fatty fish as residues were adequately recovered in the evaporated and reconstituted extract.

Method Evaluation

Although this method is meant to be used as a screen, some evaluation was made of its quantitative performance (analyte recoveries and method precision). The majority of the test compounds fortified in tilapia had recoveries (as compared to a single solvent standard at the target testing level) of >70%. Recovery values for selected analytes in tilapia are shown in Figure 2. Some compounds such as avermectins gave high recoveries (150–220%) compared to a solvent standard, indicating matrix signal enhancement might occur, especially when monitoring sodiated molecules. A few compounds (such as penillic acid and erythromycin dehydrate) were formed by decomposition during the acidic extraction process, so their apparent recoveries were quite high as compared to a solvent standard. The compounds with lower recoveries tended to be β-lactams, benzimidazoles, and macrolides. No changes were made to the extraction method to accommodate the negative ion compounds.

Figure 2.

Recoveries for selected analytes in tilapia fortified at the target testing level (1X) as compared to a 1X solvent standard.

Strategy for Screening Using HRMS

The strategy for this method development was to optimize and validate sample preparation and MS acquisition methods for representative compounds that are most likely to be used in aquaculture and then use the full scan HRMS capability along with a compound database of veterinary drugs to screen samples for >260 additional compounds. Although very large compound databases with exact mass data for thousands of analytes are commercially available, this method is focused on a more limited set of analytes that are likely to be used as veterinary drugs. In addition, a majority of the compounds in this database have been analyzed with this LC-MS method so that relevant retention time data have been obtained. The chromatographic parameters were also designed to accommodate a large number of compounds with very different polarities in a relatively short analysis time. The retention times of the test compounds, along with exact masses of precursor and product ions, are listed in Table 1. Exact mass data for product ions, which have been verified as the correct theoretical values, are also included.^11,28 Searching against larger commercial databases containing additional compounds such as pesticides, forensic chemicals, and other contaminants can always be performed, but this can lead to a higher percentage of false detects.¹¹

Two types of MS acquisition methods were evaluated. Initially, nontargeted AIF was used to analyze the extracts. With AIF data acquisition, all precursor ions (MS¹) eluting into the MS at a given time are allowed into the HCD cell to form product ions. This results in a “product ion” spectrum that is a mixture of ions from different analytes and matrix components. However, using the HRMS instrument’s ability to extract data within a narrow mass range, a specific product ion in that spectrum can be detected within the retention time window of the precursor ion to identify the compound. AIF proved to be a simple, reliable method of generating screening data as shown below. A more targeted approach, DDMS², was also used to generate product ions specific to selected precursor ion from compounds on an inclusion list. Data independent analysis (which allows subsets of precursors into the HCD cell) was also investigated, but was not found to have any clear advantages over AIF at this time.

Results for Validation Samples

AIF Data

With this screening method, data from fortified samples were analyzed to determine if the test compounds could be detected, identified, and measured at a concentration near the TTL. Area counts from extracted ion chromatograms of the precursor ions were compared to those from a matrix-extracted standard fortified at the TTL (1X). Figure 3 shows examples of extracted ion chromatograms of selected positive ion residues in tilapia fortified at the TTL.

Figure 3.

Extracted ion chromatograms (5 ppm window) from MS¹ AIF data for representative test compounds in tilapia fortified at target testing level.

Semiquantitative limit testing determines if a residue is present at or above the concentration of interest, so it is important to measure the variance of the signals generated from residues present in samples at this concentration. A threshold cutoff value can then be set to determine when to call a sample “presumptive positive”.^4,18 For all matrices tested, 10 replicate samples fortified at the TTL were extracted and analyzed on the same day. One extracted sample was assigned as the standard (100% relative recovery) and the relative recoveries of the other replicates were determined by comparison to that single standard (data included as Table 2 in the Supporting Information). The precision was also determined. For most residues the average recoveries for these data were >90%, and standard deviations were <15%. For example, cloxacillin (CLOX) in salmon samples fortified at the TTL had an average recovery of 91% with a standard deviation of 6%. The limit threshold cutoff value for that compound in salmon calculates to 80%. This means any salmon sample with a signal for CLOX >80% of the signal in a matrix-extracted TTL standard could (with 95% confidence) contain CLOX at a concentration at or above the TTL. For some residues, however, the standard deviations were higher and the threshold cutoff values were therefore lower. To avoid false negatives and simplify data analysis by treating all test compounds the same, a threshold cutoff value of ≥50% TTL seemed reasonable for all analytes to be considered presumptive positive.

In addition to a residue having the correct exact mass precursor ion (5 ppm) and a signal of ≥50% compared to that analyte in a matrix-extracted standard of the same matrix fortified at the TTL, other criteria for a residue to be presumptive positive for the test compounds include retention time matching and the presence of at least one fragment ion with the correct exact mass (within 10 ppm).¹⁹ Isotope matching can also be used for comparison, but is not required in the identification criteria. The allowed retention time window was set fairly wide (30 s each side of specified time) to accommodate drift in the method due to column age or slight changes in mobile phase composition. The measured retention times were generally much tighter than the allowed 30 s (usually within 0.1 min). The data were examined closely to determine if a peak was within the established experimental time when compared to matrix-extracted standards analyzed on the same day. Figure 4 shows selected product ions and the precursor isotope pattern for ciprofloxacin in catfish (5 μg/kg). The ciprofloxacin data show that the software is able to isolate and evaluate product ions generated by AIF to determine if the criteria for identification are met. With these criteria (including signal ≥50% compared to the TTL standard), the number of false negatives for samples fortified at the TTL and the number of false positives (for the test compounds) in matrix blanks could be determined and are shown in Table 2. In general, the performance of these test compounds in this screen was very good using the AIF acquisition program with false positives detected for only one compound and <2% false negatives for a majority of the compounds in all matrices. A few compounds (amoxicillin, florfenicol amine, avermectins, and lufenuron) had higher rates (5–20%) of false negatives at the TTL level. Sulfacetamide performed poorly in the screening method with low variable recovery and high background and was removed from the list of validated test compounds.

Figure 4.

AIF product ions (A) and precursor isotope pattern (B) for CIP (5 μg/kg) in catfish extract compared to theoretical values.

Table 2.

Screening Validation Results

analyte	class	TTL (μg/kg)	lowest level confirmed by AIF (X)	minimum detectable concentration (μg/kg)	% false neg at 1X (AIF), N = 77	% false neg at 1X (DDMS2), N = 77	% false pos in blanks (AIF), N = 47
doramectin	avermectin	200a	0.5	100	13	19	0
emamectin B1a		200a	0.1	20	1	0	0
ivermectin B1a		200a	0.5	100	5	57	0
amoxicillin	β-lactam	100	1	100	6	97	0
ampicillin		25	0.1	2.5	1	13	0
aspoxicillin		25	1	25	0	32	0
cloxacillin		25	0.1	2.5	0	2	0
dicloxacillin		25	0.5	12.5	0	5	0
oxacillin		25	0.1	2.5	0	31	0
penillic acid		25b	0.1	2.5	0	99	0
albendazole	benzimidazole	50	0.1	5	0	0	0
albendazole sulfoxide		50	0.1	5	0	0	0
fenbendazole		50	0.1	5	0	0	0
fenbendazole sulfone		50	0.1	5	0	0	0
cephapirin	cephalosporin	25	0.1	2.5	0	0	0
brilliant green	dye	1	0.5	0.5	1	84	0
crystal violet		1	0.5	0.5	0	14	0
leucocrystal violet		1	0.5	0.5	3	73	0
leucomalachite green		1	0.1	0.1	0	42	0
malachite green		1	0.5	0.5	0	2	0
ciprofloxacin	fluoroquinolone	5	0.5	2.5	0	10	0
danofloxacin		5	0.5	2.5	0	90	0
difloxacin		5	0.1	0.5	0	0	0
enrofloxacin		5	0.1	0.5	0	0	0
norfloxacin		5	0.5	2.5	1	0	0
sarafloxacin		5	0.1	0.5	0	0	0
methyl testosterone	hormone	0.8	0.5	0.4	0	2	0
lincomycin	lincomycin	50	0.1	5	0	0	0
azithromycin	macrolide	50	0.1	5	1	12	0
erythromycin dehyr		50b	0.1	5	0	0	0
spiramycin		50	0.1	5	0	1	0
tilmicosin		50	0.1	5	0	0	0
tylosin A		50	0.1	5	0	0	0
ketoconazole	nitromidazole	10	0.1	1	0	0	0
metronidazole		10	0.1	1	0	13	0
florfenicol amine	phenicol	50c	0.1	5	9	51	0
ormetoprim	potentiator	10	0.1	1	0	0	0
trimethoprim		10	0.1	1	0	0	0
ethoxyquin	preservative	50	0.1	5	10	10	0
flumequine	quinolone	10	0.1	1	0	0	0
nalidixic acid		10	0.1	1	0	0	0
oxolinic acid		10	0.1	1	0	0	0
sulfacetamide	sulfonamide	10	1	10	21	100	9
sulfachloropyridazine		10	0.1	1	0	8	0
sulfaclozine		10	0.1	1	0	0	0
sulfadiazine		10	0.1	1	1	31	0
sulfadimethoxine		10	0.1	1	0	0	0
sulfadoxine		10	0.1	1	0	0	0
sulfaethoxypyridazine		10	0.1	1	0	0	0
sulfamerazine		10	0.1	1	1	17	0
sulfamethazine		10	0.1	1	0	14	0
sulfamethoxazole		10	0.1	1	0	0	0
sulfamethoxypyridazine		10	0.1	1	0	8	0
sulfamonomethoxine		10	0.1	1	0	0	0
sulfapyridine		10	0.1	1	0	4	0
sulfaquinoxaline		10	0.1	1	0	1	0
sulfathiazole		10	0.5	5	6	16	0
chlortetracycline	tetracycline	100d	0.1	10	0	0	0
doxycycline		100d	0.1	10	0	0	0
oxytetracycline		100d	0.1	10	0	0	0
tetracycline		100d	0.1	10	0	2	0
chloramphenicol	phenicol	0.3	0.5	0.15	0f	94f	0f
florfenicol		5	0.1	0.5	0f	0f	0f
thiamphenicol		5	0.1	0.5	0f	34f	0f
toltrazuril	toltrazuril	50	0.1	5	0f	77f	0f
toltrazuril sulfone		50	>2e	>100	100f	100f	0f
toltrazuril sulfoxide		50	0.5	25	0f	8f	0f
diflubenzuron	benzylurea	50	0.5	10	0f	69f	0f
lufenuron		50	0.5	10	8f	34f	0f
teflubenzuron		50	0.5	10	0f	11f	0f

^aCurrent FDA program recommends TTL of 10 μg/kg.²⁰

^bMarker compounds were degradants; amounts were compared to matrix-extracted standard fortified with parent compound at TTL.

^cFDA tolerance is 1 mg/kg for florfenicol amine as marker residue in aquaculture.²⁰

^dFDA tolerance is 2 mg/kg for sum of oxytetracycline, chlortetracycline, and tetracycline in finfish and lobster.²⁰

^eNo fragments were observed for toltrazuril sulfone.

^fFor negative ion analytes: N = 35 fortified at 1X and N = 15 blanks.

Because the TTLs are somewhat arbitrary (any amount of a nonapproved drug identified could be considered violative), the minimum detectable limits using AIF data were also determined by fortifying samples with the test compounds at lower levels. The same criteria, with the exception of the ≥50% of signal compared to TTL, were applied to determine if the residues could be detected and identified. In most cases, analytes were still identified in extracts from fish and shellfish that had been fortified at 0.1X the TTL.

This method is designed to analyze a wide variety of compounds with very different polarities; therefore, the performance for some drugs may not compare well to methods that were developed specifically for those analytes. For example, some of the avermectins do not meet all of the criteria for presumptive positive with this screening method below a fortification level of 100 μg/kg, even though the current testing level for these compounds at the FDA is 10 μg/kg.²⁰ There are several reasons for this including the fact that these compounds are much more nonpolar than most veterinary drugs, so extraction efficiency and low solubility may reduce recoveries. Penicillins, on the other hand, are most soluble in water; to extract these compounds simultaneously with avermectins proved difficult. The extraction presented here was the best compromise. In addition, although using the generic acetoni-trile/0.1% formic acid LC gradient works well for a majority of the veterinary drugs in the >330 compound database, a buffer containing ammonium formate would have facilitated the formation of ammonium adducts for the avermectins (rather than the sodiated ions, which are more difficult to dissociate into product ions). The sodiated precursor ions of ivermectin and doramectin were observed in extracts at levels <200 μg/kg (0.1X or 20 μg/kg), meaning the method could potentially be used to detect these residues at lower levels. The incidence of false positives that can occur when only using the measured exact mass of the precursor ion would be less for these compounds because their m/z values (m/z 897, 921) are above those of most background contaminants. However, the absence of product ions prevented confirmation as presumptive positive under the described criteria. Emamectin, in contrast, forms a protonated molecule, and product ions could be detected for this compound at the 0.1X level.

This HRMS screening method was also validated for several negative ion compounds. It was possible to detect chloramphenicol in shrimp and salmon using negative ion AIF data acquisition below the established TTL (0.3 μg/kg); florfenicol and thiamphenicol were detected at 0.1X of the 5 μg/kg TTL. Although the precursor ion of toltrazuril sulfone was detected at low levels, it did not form consistent product ions and so could not be considered presumptive positive even at the TTL. The more nonpolar negative ion compounds were detected and identified by AIF at low levels from fortified salmon and catfish in the acetonitrile portion (lufenuron, teflubenzuron) or in the final 10% acetonitrile extract (toltrazuril metabolites) or both (diflubenzuron, toltrazuril). All negative ion compounds could be detected using the extraction procedure in fortified shrimp, tilapia, and eel, although the recoveries for lufenuron were variable.

DDMS² Data

Data-dependent MS² acquisition, where a narrow range of precursor ions are allowed through the quadrupole into the HCD cell and fragmented, generated product ion spectra more specific to the precursor. These spectra can be used to generate product ion ratios and/or search against library spectra. With this method, DDMS² worked well for a majority of compounds tested. Figure 5 illustrates how isobaric compounds flumequine and oxolinic acid (FLU and OXO) can be identified on the basis of their difference in precursor ion exact mass, retention times, and unique product ion spectra obtained by DDMS². Some analytes, however, were missed by DDMS² data acquisition because the threshold signal level needed to trigger the product ion spectra was not reached. Generally, this occurred for residues with low TTL and/or poor recoveries. The number of false negatives based on DDMS² acquisition is shown in Table 2. Approximately 20 of the positive ion test compounds had a false negative rate of >10% at the TTL. Many of these analytes did produce product ion spectra using DDMS² in extracts from fish that had been fortified at a slightly higher level (2X TTL). Similarly, some negative ion compounds did not generate DDMS² product ion spectra at the TTL, even though product ions were detected using AIF. Because AIF data gave more reliable results for the presence of confirmatory product ions, DDMS² spectra were collected to obtain additional qualitative information but were not required for initial screening and identification.

Figure 5.

DDMS² data for OXO and FLU in shrimp at target testing level (10 μg/kg): (A) extracted ion chromatograms for MH⁺; (B) chromatograms for MS² of m/z 262; (C) product ion spectra.

Results for Incurred and Regulatory Fish Samples

Test Compounds

In addition to fortified samples, incurred and violative regulatory samples were also tested with this screening method. Initially, data from these samples were evaluated to look for the test compounds that had been validated. The results for residues found in incurred and violative samples are shown in Tables 3 and 4, respectively. As expected, high levels of sulfadiazine (SDZ) were found in tilapia dosed with that compound. An average of 230 μg/kg SDZ was measured in fish dosed at 1 mg/kg body weight (1 day withdrawal), and up to 650 μg/kg SDZ in fish dosed with 5 mg/kg body weight (3 day withdrawal). However, only low levels (<1 μg/kg) of trimethoprim were detected in tilapia dosed with a combination of trimethoprim and SDZ, and DDMS² product ion spectra were triggered for trimethoprim in only one of the two dosed fish. Sulfadiazine was also identified in five regulatory tilapia samples that had previously been found to be violative for that residue. The concentrations of SDZ in these fish estimated using a one-point matrix-extracted TTL standard for comparison were similar to the amounts reported previously using LC-MS/MS methodology.⁴ Sulfamethoxazole and trimethoprim were identified in another sample of tilapia. Sulfamethazine (SMZ) was found in regulatory eel samples; one sample had levels of SMZ > 100 μg/kg, and the other was below the threshold testing level (<50% TTL). The MS¹ data report for SMZ and other compounds found in the first eel sample is shown in Figure 6. Trimethoprim was found to be presumptive positive in both eel samples analyzed.

Table 3.

Screening Results for Incurred Fish Samples

incurred sample	dosing drug (X), mg/kg body weight	depuration time	N	found by QqQ30 (μg/kg)	test compounds presumptive positive AIFa (μg/kg)	additional compounds found by AIF	spectra obtained by DDMS2	comments
tilapia inc 1	SDZ (1)	1 day	2	not analyzed	SDZ (220)	N4 acetyl-SDZ ETHOX dimerb	SDZ N4 acetyl-SDZ	ETHOX found by AIF at <50% TTL
tilapia inc 2	SDZ (1)	1 day	2	not analyzed	SDZ (240)	N4 acetyl-SDZ ETHOX dimerb	SDZ N4 acetyl-SDZ
tilapia inc 3	SDZ (5) TRIMETH (1)	3 days	2	not analyzed	SDZ (650)	N4 acetyl-SDZ ETHOX dimerb	SDZ N4 acetyl-SDZ	TRIMETH found by AIF at <50% TTL
tilapia inc 4	SDZ (5) TRIMETH (1)	4 days	2	not analyzed	SDZ (280)	N4 acetyl-SDZ	SDZ TRIMETH N4 acetyl-SDZ	TRIMETH found by AIF at <50% TTL
catfish inc 1	ENRO (5)	6 days	4	not analyzed	ENRO (620) CIP (30)	des-ENRO	ENRO, CIP ETHOX	ETHOX found by AIF at <50% TTL
catfish inc 2	ENRO (5)	6 days	4	not analyzed	ENRO (601) CIP (41)	des-ENRO	ENRO, CIP ETHOX	ETHOX found by AIF at <50% TTL
catfish inc 3	MG (2)c CV (2)	1 h	1	LCV (4.3) MG (3.1)	LCV (2.7) LMG (0.8)		LCV ETHOX	BG, CV, MG, and ETHOX found by AIF at <50% TTL
	BG (2)			LMG (2.7) BG (1.2)
salmon inc 4	MG (2)c CV (2) BG (2)	1 h	1	BG (1.8) MG (1.8) LMG (0.8) LCV (0.4)	none	ETHOX dimerb	none	BG and MGs found by AIF at <50% TTL

^aAverage amount calculated as compared to one-point matrix-extracted standard with same matrix at 1X target testing level.

^bEthoxyquin dimer standard was obtained, and compound was added to DDMS² inclusion list after these data were collected.

^cFish were exposed to bath containing 2 μg/L MG, CV, and BG for 1 h followed by 1 h in clean water tank.

Table 4.

Screening Results for Violative Regulatory Fish Samples

regulatory sample	N	found by QqQ4 (μg/kg)	test compounds presumptive positive AIFa (μg/kg)	additional compounds found by AIF	spectra obtained by DDMS2	comments
tilapia reg 1	2	SDZ (76)	ETHOX (105) SDZ (69)	N4 acetyl-SDZ ETHOX dimerb	SDZ, ETHOX N4 acetyl-SDZ
tilapia reg 2	2	SDZ (77)	SDZ (52)	N4 acetyl-SDZ ETHOX dimerb	SDZ, ETHOX N4 acetyl-SDZ	ETHOX found by AIF at <50% TTL
tilapia reg 3	2	SDZ (9)	SDZ (6)	none	SDZ, ETHOX	ETHOX found by AIF at <50% TTL
tilapia reg 4	2	SDZ (4)	SDZ (3)	N4 acetyl-SDZ ETHOX dimerb	ETHOX	ETHOX found by AIF at <50% TTL
tilapia reg 5	2	SDZ (5)	SDZ (3)		ETHOX	ETHOX found by AIF at <50% TTL
tilapia reg 6	2	SMX (20) TRIMETH (6)	SMX (15) TRIMETH (6)	N4 acetyl-SMX ETHOX dimer	SMX, TRIMETH ETHOX, ETHOX dimer N4 acetyl-SMX	ETHOX found by AIF at <50% TTL
catfish reg 1	2	LCV (1.1) ENRO (<5)	LCV (0.5)	ETHOX dimer	LCV, ENRO ETHOX dimer	ENRO found by AIF at <50% TTL
catfish reg 2	2	ENRO (10)	ENRO (8)	ETHOX dimer	ENRO, ETHOX ETHOX dimer	ETHOX found by AIF at <50% TTL
eel reg 1	2	TRIMETH (124, 597)c CIP (26) ENRO (8) SMZ (<10)	TRIMETH (270) CIP (48) ENRO (9)	des-ENRO ETHOX dimer N4 acetyl-SMX thiabendazole	TRIMETH, CIP, ENRO ETHOX dimer N4 acetyl-SMX thiabendazole in one	SMZ and SMX found by AIF at <50% TTL
eel reg 2	2	SMZ (120) ENRO (12) CIP (90) TRIMETH (30)	ETHOX (114) SMZ (104) ENRO (72) CIP (52) TRIMETH (23)	des-ENRO ETHOX dimer N4 acetyl-SMZ	ETHOX, SMZ ENRO, CIP, TRIMETH ETHOX dimer N4 acetyl-SMZ LINC, OTC	LINC, OTC found by AIF at <50% TTL

^aAverage amount calculated as compared to one-point matrix-extracted standard with same matrix at 1X target testing level.

^bEthoxyquin dimer standard was obtained and compound was added to DDMS² inclusion list after these data were collected.

^cThe amount of trimethoprim was variable between initial screen and quantification by LC-MS/MS method (tissue was diffcult to homogenize).

Figure 6.

AIF data for regulatory eel sample 1: (A) test compounds that were presumptive positive (signal ≥50% target testing level); (B) test compounds identified, but at concentrations <50% target testing level; (C) other compounds identified when compared to larger veterinary drug database.

Fluoroquinolone residues are often found in aquacultured fish and shellfish. Ciprofloxacin (CIP) is a known metabolite of enrofloxacin (ENRO) in fish, and both compounds were identified in catfish that had been dosed with ENRO. The amount of residues found was consistent with what had been reported earlier for a similar dosing experiment.²⁹ ENRO was identified in two regulatory catfish samples, although at concentrations high enough to be considered presumptive positive in only one fish. The HRMS screening method also found high levels of ENRO and CIP in the eel samples. DDMS² product ion spectra were collected for ENRO and CIP in all of these samples, and the concentrations found were generally consistent with reports from earlier analyses for these regulatory samples using LC-MS/MS.⁴

Fish that had been subjected to a water bath containing triphenylmethane dyes and previously analyzed using an AOAC LC-MS/MS method³⁰ were also tested using this HRMS screening method. The catfish sample was found to be presumptive positive for leucocrystal violet (LCV) and leucomalachite green (LMG). The chromic forms of the dyes (CV, MG, and brilliant green, BG) were detected in the catfish, but the estimated concentrations were below the threshold level (<50% TTL). The dyes previously found (MG and BG) in the incurred salmon sample were identified, but were also not considered presumptive positive because the concentrations were too low. In general, the concentrations for the dyes estimated by this screening method were lower than reported earlier,³⁰ partially because internal standard correction was not utilized. LCV was identified in a regulatory catfish level at the presumptive positive threshold cutoff (0.5 μg/kg = 50% TTL). Earlier analysis of this fish⁴ reported 1.1 μg/kg of LCV.

Ethoxyquin was found in many of these samples. Residues of ethoxyquin are not unexpected as this is an approved preservative in fish feed.³¹ In general, the concentration of ethoxyquin found by this method was relatively low (<50% of TTL at 50 μg/kg), although it was present at high levels (>100 μg/kg) in two samples. The recoveries and precision of ethoxyquin at the TTL were variable depending on the matrix, so it may be difficult to determine the significance of these findings. Other test compounds that were confirmed by both AIF and DDMS² data in one of the regulatory eel samples were lincomycin and oxytetracycline (Figure 6), but the concentrations were well below the threshold cutoff. The DDMS² spectrum for lincomycin in this sample is shown in Figure 7.

Figure 7.

DDMS² product ion spectra for residues in regulatory eel sample 2: (A) lincomycin; (B) ethoxyquin dimer.

Semitargeted Screening

The advantage of full scan HRMS screening is that the data can be searched for additional analytes. Although these compounds may not have been validated in the method using fortified samples, the compound database with >260 additional compounds contains information to match precursor ion and, for many analytes, retention time and fragment ions. When data from the incurred and violative regulatory samples were compared to compounds in the larger database, several other analytes were detected, and these results are also shown in Tables 3 and 4. In fish extracts containing high levels of sulfonamides, the N⁴-acetyl metabolites were also detected. These metabolite products have been described⁸ and were included in the DDMS² inclusion list so that unique product ion spectra could be obtained.

Another additional compound that was found in many of these samples was ethoxyquin dimer, a known byproduct of ethoxyquin in fish.³² When these samples were first analyzed, a reference standard for ethoxyquin dimer was not available for comparison, and the compound was detected by matching the exact m/z of the precursor. Subsequently, a standard was purchased, and it was determined that the retention time and product ions also matched that compound. The metabolite desethylene-enrofloxacin (des-ENRO) that had been found earlier in milk samples from cows dosed with ENRO³³ was also found in the catfish dosed with this drug. Thiabendazole was also identified in one regulatory eel sample when the data were compared to the larger veterinary drug database, although the estimated levels were quite low (<1 μg/kg when compared to a solvent standard). One regulatory eel sample was also analyzed by AIF using negative ion data acquisition, but no compounds were detected.

It is worth noting that the rate of false detections increased when the data were compared to more potential analytes in a larger compound database. Searching by exact mass of precursor ion only can lead to a significant number of false hits (40–50) even with the narrower mass window of 3 ppm. Adding criteria for retention time and product ion match greatly reduced the number of potential positives to a reasonable number (<5). Some compounds were often detected in many samples (including matrix controls) at low abundances (~10⁴ counts) with the correct retention time and fragment ions using the screening data analysis program with AIF data, but DDMS² spectra could not be obtained. These include small mass compounds with common fragment ions (e.g., lidocaine, benzocaine) and hormones (hydrocortisone, prednisolone, etc.).

These results from the analysis of incurred and violative regulatory aquaculture samples demonstrate the ability of the HRMS screening method to confirm the findings of more traditional LC-MS/MS methods for a wide range of veterinary drug residues. In addition, the capability to detect metabolites such as N⁴-acetyl-sulfonamides, des-ENRO, and ethoxyquin dimer provides verification that the veterinary drugs were administered to the fish and are not an artifact of postproduction processing or analysis. The identification of other unexpected drug residues in these samples, even at low levels, also demonstrates the potential for expanding the scope of monitoring for chemical contaminants in aquacultured fish and shellfish using HRMS screening methods.