Automated dual-cartridge solid-phase extraction method for multi-residue determination of veterinary drugs in bovine muscle, liver, fat, and milk

Abstract

Solid-phase extraction (SPE) is a widely used sample preparation technique for the determination of veterinary drug residues in foods due to its effectiveness in removing matrix components. However, conventional SPE methods are often time consuming and labor intensive. In this study, we developed an automated sample preparation method employing a dual-cartridge SPE system for the multi-residue determination of veterinary drugs in bovine-derived food matrices. The method integrates two C18 cartridges connected in series, with controlled water addition between cartridges to improve cleanup efficiency. Analytical performance was evaluated for 52 veterinary drugs in bovine muscle, liver, fat, and milk at a concentration of 0.01 mg kg⁻¹. Minimal matrix effects were observed, allowing for accurate quantification using solvent-based calibration without the need for matrix-matched standards or isotope-labeled internal standards. Satisfactory analytical performance was obtained for approximately 80% of the analyte–matrix combinations, with trueness values ranging from 70% to 120% and intra- and inter-day precision values within 25% and 30%, respectively, although some combinations fell outside these criteria due to degradation or matrix effects. The method also demonstrated high selectivity, with no interfering peaks detected near the retention times of target analytes. Overall, the developed automated SPE method provides a robust and reliable platform for quantification of veterinary drug residues in complex bovine-derived food products, supporting its suitability for use in routine food safety monitoring and regulatory surveillance.

---

Introduction

Veterinary drugs play a critical role in the field of animal husbandry, where they are used to prevent and treat diseases and enhance productivity. Drugs such as antibiotics, including sulfonamides, macrolides, and quinolones, are extensively used for disease prevention and treatment, whereas hormonal agents are employed to regulate reproductive cycles or promote animal growth. However, the presence of these compounds in edible tissues may pose health risks to humans.¹ To safeguard human health, the Codex Alimentarius Commission, the European Union, and national regulatory authorities in countries such as Japan and the United States have established maximum residue limits (MRLs) for veterinary drugs in foods. Ensuring compliance with these standards requires analytical methods that are highly sensitive and selective and enable accurate quantification of trace-level residues in complex food matrices.

Liquid chromatography-tandem mass spectrometry (LC-MS/MS) is currently the method of choice for determining veterinary drug residues in foods due to its high sensitivity and selectivity.^2–6 However, the complex composition of food matrices can induce significant matrix effects in LC-MS/MS analyses, such as ion suppression or enhancement, which may lead to inaccurate quantification when solvent-based calibration is employed. Thus, matrix-matched calibration is frequently used to compensate for matrix effects.^7–9 However, this approach necessitates the availability of drug-free matrices identical to the test samples in order to accurately compensate for matrix effect, which is often impractical. Although stable isotope-labeled internal standards offer an alternative approach for correction, their high cost and limited availability for a wide range of analytes restrict routine application in multi-residue analyses.^10,11 Additionally, insufficient removal of co-extracted matrix components can interfere with chromatographic separation, thus further compromising analytical accuracy. Therefore, effective sample cleanup procedures are essential for the reliable quantification of trace-level residues in complex food matrices.

Solid-phase extraction (SPE) and dispersive SPE are widely employed to prepare sample solutions for veterinary drug residue analysis.^2–11 Although SPE is generally more effective for the removal of matrix components, it is often time consuming and labor intensive. To address these limitations, automated SPE systems have garnered increasing attention, particularly for pesticide and veterinary drug residue analyses.

Several automated SPE methods have been reported, particularly for the analysis of pesticide residues in crops and veterinary drug residues in muscle of pork, chicken, and fish, demonstrating the feasibility and advantages of automated sample preparation approaches.^12–21 Recently, automated micro-SPE (μSPE), a miniaturized version of conventional SPE, has also attracted attention for the determination of pesticides.^12,13,15 This technique can be directly coupled to LC, enabling for efficient and streamlined analysis.^18,19 However, most existing methods, including μSPE, utilize a single-cartridge format, despite the potential benefits of dual-cartridge configurations in increasing sample cleanup efficiency. To date, no study has reported the implementation of a fully automated dual-cartridge SPE system specifically developed for comprehensive cleanup and simultaneous multi-residue determination of veterinary drugs in complex animal-derived food matrices.

To fill this knowledge gap, the present study developed a robust automated SPE method incorporating a dual-cartridge configuration. The method was designed for the simultaneous determination of 52 veterinary drugs, including sulfonamides, macrolides, quinolones, and hormonal agents, in bovine muscle, liver, fat, and milk. The dual-cartridge SPE system employs two C18 cartridges connected in series, with a solvent delivery nozzle positioned between the cartridges. This configuration enables the introduction of a distinct solvent composition between the two cartridges, thereby enhancing matrix removal efficiency through sequential and selective extraction.

Experimental

Reagents

Acetic acid (guaranteed reagent grade), anhydrous sodium sulfate (pesticide residue and PCB analysis grade), and formic acid (LC/MS grade) were obtained from FUJIFILM Wako Pure Chemical (Osaka, Japan). Acetonitrile (LC/MS grade), distilled water (LC/MS grade), and hexane (pesticide residue and PCB analysis grade) were obtained from Kanto Chemical (Tokyo, Japan). Veterinary drug mixture standard solutions—macrolides (20 μg mL⁻¹ each in acetonitrile) and hormones (20 μg mL⁻¹ each in acetonitrile)—were obtained from FUJIFILM Wako Pure Chemical. PL Veterinary LC/MS Mix 1 and Mix 2 (20 μg mL⁻¹ each in acetonitrile) were purchased from Hayashi Pure Chemical (Osaka, Japan). Equal volumes of the four mixtures were combined to prepare a composite working standard solution for use in analyses.

SPE cartridges

The following SPE cartridges were used in this study: silica-based octadecyl (C18) cartridges—Smart-SPE C18-30 (30 mg sorbent) and C18-50 (50 mg sorbent); polymer-based mixed-mode cartridges—Smart-SPE PBX-20 and PLS3-20 (each with 20 mg sorbent); a polymer-based anion-exchange cartridge—Smart-SPE AXi3-20 (20 mg sorbent); and a silica-based anion-exchange (primary and secondary amine [PSA]) cartridge—Smart-SPE PSA-30 (30 mg sorbent). All cartridges were purchased from AiSTI Science (Wakayama, Japan).

Food samples

Samples of bovine muscle, liver, fat and milk were purchased from retail outlets in Japan. Muscle, liver, and fat samples were homogenized using a knife mill (Grindomix GM200, Retsch, Haan, Germany).

Extraction

An overview of the extraction procedure is shown in Fig. S1. Sample (10.0 g) was weighed in a glass tube and homogenized using a Polytron PT 10-35 GT homogenizer (Kinematica, Lucerne, Switzerland) for 1 min with a mixture of 50 mL of hexane-saturated acetonitrile, 50 mL of hexane, and 1 mL of acetic acid. Subsequently, 20 g of anhydrous sodium sulfate was added, and the mixture was homogenized for an additional 1 min. The resulting homogenate was centrifuged at 1932 × g for 5 min using an S700FR centrifuge (Kubota, Tokyo, Japan). After removal of the upper hexane layer, the lower acetonitrile layer was collected. The residue was extracted again by adding 40 mL of acetonitrile and homogenizing for 1 min, followed by centrifugation at 1932 × g for 5 min. The resulting supernatant was collected and combined with the first extract. The combined extract was then adjusted to a final volume of 100 mL using acetonitrile.

Automated SPE procedure

SPE was performed using an automated SPE system (ST-L400, AiSTI Science, Wakayama, Japan). Fig. 1 provides an overview of the finalized automated SPE procedure. Two identical Smart-SPE C18-50 cartridges (each containing 50 mg of sorbent) were installed in series, with a solvent delivery nozzle positioned between the cartridges. The cartridges were conditioned sequentially using 1 mL of acetonitrile followed by 1 mL of acetonitrile/water (9 : 1, v/v) (Fig. 1A and B). During the acetonitrile/water conditioning step, 0.2 mL of water was introduced through the nozzle to dilute the effluent from the first cartridge before it passed into the second cartridge (Fig. 1B). Subsequently, 2 mL of sample extract was loaded onto the first cartridge. An additional 0.4 mL of water was introduced through the nozzle to dilute the eluate before transfer to the second cartridge (Fig. 1C). Finally, 0.5 mL of acetonitrile/water (9 : 1, v/v) was applied to the first cartridge, and 0.2 mL of water was again delivered through the nozzle to dilute the eluate as it flowed into the second cartridge for final elution (Fig. 1D). The combined eluate was adjusted to a final volume of 4 mL using 0.1% (v/v) formic acid.

Fig. 1 Schematic overview of the developed automated SPE system. (A) First conditioning step with acetonitrile. (B) Second conditioning step with acetonitrile/water (9 : 1, v/v). (C) Sample loading and first elution step. (D) Second elution step with acetonitrile/water (9 : 1, v/v).

LC-MS/MS analysis

LC-MS/MS analyses were performed using a Shimadzu Nexera X3 system (Shimadzu, Kyoto, Japan) coupled to a Triple Quad 7500 mass spectrometer (AB Sciex, Framingham, MA, USA). Chromatographic separation was achieved on an InertSustain AQ-C18 column (2.1 × 100 mm, 2 μm; GL Sciences, Tokyo, Japan) maintained at 40 °C. The mobile phases consisted of 0.1% (v/v) formic acid in water (solvent A) and 0.1% (v/v) formic acid in acetonitrile (solvent B). The gradient elution program was as follows: 2–70% B (0–15 min), 95% B (15–20 min), and 2% B (20–25 min). The flow rate was set to 0.3 mL min⁻¹, and the injection volume was 3 μL (2 μL was used for SPE cartridge optimization). Both positive and negative electrospray ionization (ESI) modes were employed with the following parameters: ion source gas 1, 70 psi; ion source gas 2, 80 psi; curtain gas, 35 psi; CAD gas, 7 arbitrary units; heater temperature, 450 °C; ion spray voltage, 2000 V. Selected reaction monitoring was used for analyte quantification, with the transitions and associated parameters for each compound summarized in Table S1. SCIEX OS software (version 2.0.1.48692; AB Sciex) was used for data acquisition and processing.

Method validation

Method validation was performed for 52 target compounds in bovine muscle, fat, liver, and milk. The validation procedure was conducted in accordance with the Japanese method validation guideline²² for pesticide and veterinary drug residue analysis, with performance evaluated in terms of trueness, intra-day precision, inter-day precision, and selectivity. Quantification was based on external calibration using solvent-based standard solutions prepared at concentrations ranging from 0.000125 to 0.00075 μg mL⁻¹, corresponding to 0.0025 to 0.015 mg kg⁻¹ in samples. Trueness and precision were assessed at a spiking level of 0.01 mg kg⁻¹ by analyzing two replicates per day over five separate days. Matrix effect values (ME, %) were calculated using the following equation:
where A_MSTD represents the average peak area of the analyte in the matrix-matched standard and A_SSTD the peak area of the analyte in the solvent standard. Both peak areas corresponded to a concentration of 0.01 mg per kg sample equivalents.

Optimization of the automated SPE procedure

SPE conditions were optimized by evaluating different second cartridges with Smart-SPE C18-50 fixed as the first cartridge. Various second cartridges were evaluated, including Smart-SPE C18-30, C18-50, PBX-20, PLS3-20, AXi3-20, and PSA-30. Spiked samples were prepared by adding veterinary drug mixture standard solution to blank bovine extracts, corresponding to 0.1 mg kg⁻¹ in the sample. The automated SPE procedure was conducted as described in the “Automated SPE procedure” section. When either AXi3-20 or PSA-30 was used as the second cartridge, 0.1% (v/v) formic acid was introduced through the nozzle instead of water during both the conditioning and dilution steps.

Results and discussion

Strategy for automated cleanup optimization

This study targeted 52 veterinary drugs with diverse physicochemical characteristics (log P_ow values ranging between approximately −1 and 5), including sulfonamide, quinolone, and macrolide antibiotics, as well as various hormones. Given this chemical diversity, an extraction method capable of extracting both polar and low-polarity analytes was required. Samples were therefore first extracted using an acetonitrile–hexane mixture, with the hexane layer discarded to remove the bulk of the fat (Fig. S1). However, this step alone was insufficient to eliminate residual co-extracted components such as phospholipids, pigments, and other low-polarity components. To further reduce these co-extracted components, an automated SPE cleanup step was incorporated prior to LC-MS/MS analysis.

In the SPE cleanup step, C18 and HLB cartridges are frequently used in the analysis of veterinary drugs^23–27 because many of these veterinary drugs are moderately to highly polar, whereas animal-derived matrices such as muscle, fat, liver, and milk contain substantial amounts of low-polarity components, including triglycerides, cholesterol, phospholipids, and pigments. Such low-polar matrix components can be effectively removed via a pass-through SPE approach, in which relatively polar analytes elute without retention.^23–26 In this study, C18 was selected to remove low-polarity components while allowing the efficient recovery of a broad range of veterinary drugs.

To facilitate automation, the acetonitrile extract was loaded directly onto the cartridge without prior dilution, solvent change, or concentration. This direct-loading strategy minimized handling steps and helped avoid clogging, while maintaining effective removal of low-polarity matrix components. However, because the extract consisted almost entirely of acetonitrile, its polarity was insufficient to exclude certain matrix components, resulting in limited cleanup efficiency. Therefore, a second-stage cleanup step was introduced to enhance matrix removal.

The developed automated SPE workflow enabled high-throughput operation, with an overall run time of approximately 9 min per sample. No consumables other than SPE cartridges were required. The SPE step consumed only 1 mL of acetonitrile, 1.5 mL of acetonitrile/water (9 : 1, v/v), and 0.8 mL of water per sample. Routine maintenance was minimal, as the system automatically rinsed the SPE flow path with acetone, acetonitrile, acetonitrile/water (9 : 1, v/v), and water after each run.

Selection of the second-stage cleanup cartridge

Six different SPE cartridges (C18-30, C18-50, PBX-20, PLS3-20, AXi3-20, and PSA-30) were evaluated to optimize the second-stage cleanup, using bovine muscle as the sample matrix. Fig. 2A provides an overview of recovery values, whereas Fig. 2B summarizes the MEs. Overall, the C18-30 and C18-50 cartridges yielded the highest number of compounds meeting both the recovery (70–120%) and ME (±20%) criteria. The AXi3-20 and PSA-30 cartridges also provided satisfactory recoveries but exhibited stronger matrix effects, particularly for macrolide antibiotics such as erythromycin A, leucomycin A5, and neospiramycin I (Fig. 2C and D). By contrast, the polymer-based cartridges PBX-20 and PLS3-20 generally showed more pronounced MEs, suggesting insufficient matrix removal under the tested conditions.

Fig. 2 Comparison of the cleanup efficiency of various cartridges for bovine muscle. (A) Number of compounds for which recovery was <70%, 70–120%, and >120%. (B) Number of compounds exhibiting MEs of less than −20%, −20% to +20%, and greater than +20%. (C) Recoveries of erythromycin A, leucomycin A5, and neospiramycin I. (D) MEs for erythromycin A, leucomycin A5, and neospiramycin I with different cartridges.

Based on these findings, C18-30, C18-50, AXi3-20, and PSA-30 were identified as promising candidate cartridges and further evaluated for performance in liver, fat, and milk. Across all tested matrices, these cartridges yielded acceptable recovery values. The C18-50 cartridge showed the highest matrix removal efficiency in fat, one of the most complex matrices (Fig. 3). These results indicate that the C18-50 cartridge provides robust and consistent cleanup efficiency across a diverse range of matrices and C18-50 was therefore selected for both stages in the final automated SPE method.

Fig. 3 Comparison of the cleanup efficiency of various cartridges for bovine fat. (A) Number of compounds for which recovery was <70%, 70–120%, and >120%. (B) Number of compounds with MEs less than −20%, −20% to +20%, and greater than +20%.

Effect of the volume of nozzle-introduced aqueous solvent

In our system, a nozzle was integrated between the first and second cartridges, allowing for modification of the solvent composition by introducing an aqueous solvent prior to loading onto the second cartridge. The efficiency of matrix removal in the second cleanup stage depends not only on the type of cartridge used but also on the volume of aqueous solvent introduced between the two cartridges. To optimize this parameter, we evaluated the effect of the volume of water introduced via the nozzle during two operations: loading and elution of the extract (Fig. 1C) and further elution (Fig. 1D).

During the extract loading and elution step (Fig. 1C), 0.4 mL of water was introduced between the two cartridges. As a result, the eluate from the first-stage C18-50 cartridge was diluted before being transferred to the second cartridge. To determine whether increasing the polarity of the eluate further would improve matrix removal in the second stage, the volume of water added was increased to 0.8 mL. Increasing the water volume did not compromise the recovery of any of the compounds, including those with relatively low polarity. However, no significant improvement in matrix effects was observed when compared with the addition of 0.4 mL of water (Fig. S2). These findings suggest that 0.4 mL of water is sufficient to induce the desired polarity shift and achieve efficient matrix removal. Therefore, 0.4 mL was adopted as the optimal volume of water to introduce during extract loading.

In the subsequent step (Fig. 1D), during elution from the first cartridge using 0.5 mL of acetonitrile/water (9 : 1), an additional volume of water was introduced via the nozzle before the eluate was loaded onto the second C18-50 cartridge. To optimize this step, we compared the matrix removal efficiency between the addition of 0.2 mL versus 0.4 mL of water. Similar to the previous experiment, no appreciable improvement in recoveries or matrix effects were observed when the water volume was increased (Fig. S3). This result suggests that 0.2 mL of water provides sufficient dilution and polarity enhancement for matrix removal. Thus, a volume of 0.2 mL was selected for the final procedure.

Finalized automated cleanup method

Details of the finalized automated SPE cleanup method are illustrated in Fig. 1. The method employs identical C18-50 cartridges for both the first and second cleanup stages. The acetonitrile extract is directly loaded onto the first cartridge, followed by elution with acetonitrile/water (9 : 1). This procedure effectively removes low-polarity matrix components, including lipids, without cartridge clogging. The presence of a nozzle between the cartridges enables the precise introduction of an aqueous solvent to increase the polarity of the eluate and thereby enhance matrix removal by the second cartridge. This method provides greater cleanup efficiency compared with the use of a single SPE cartridge. Additionally, this stepwise cleanup process—requiring elution, dilution, and reloading between cartridges as a continuous-flow operation—would be difficult to perform manually.

Method validation

The developed method was validated for the simultaneous determination of 52 veterinary drugs at a concentration of 0.01 mg kg⁻¹, the default MRL in Japan for veterinary drugs without established MRLs, in bovine muscle, liver, fat, and milk, following the Japanese method validation guideline.²² No interfering peaks exceeding 10% of the analyte peak area were observed in any matrix, demonstrating the method's high selectivity.

Matrix effects were evaluated for each compound–matrix combination (Fig. 4). Although matrix-matched calibration is commonly used to compensate for matrix effects,^7–9 it requires drug-free matrices identical to the test samples, which is often impractical and does not guarantee accurate quantification across different matrices. Isotope-labeled internal standards provide an alternative approach; however, their high cost and limited availability for a wide range of analytes restrict their routine use in multi-residue analyses.^10,11 Moreover, using non-identical internal standards may compromise quantification accuracy. For these reasons, we optimized the sample cleanup procedure to effectively remove matrix interferences, enabling accurate quantification using solvent-based standards without the need for matrix-matched calibration or isotope-labeled internal standards. As shown in Fig. 4, all compounds in fat showed MEs within ±20%, whereas in muscle, liver, and milk, only two compounds per matrix exceeded this range. These results indicate that the optimized cleanup sufficiently minimized matrix effects for most analytes, demonstrating the suitability of solvent-based calibration for this method. Nevertheless, a few compounds still exhibited matrix effects and were therefore excluded from the scope of this analytical method. Accurate quantification of these compounds would require additional correction, for example by the standard-addition method, although it is more labor-intensive.

Fig. 4 MEs of the developed method for bovine muscle, liver, fat, and milk. Red bars indicate values outside the ±20% range.

Trueness and intra-/inter-day precision were evaluated via recovery tests at 0.01 mg kg⁻¹. As shown in Table 1, trueness values ranging from 70–120% were obtained for 45 compounds in muscle, 41 in liver, 48 in fat, and 47 in milk. Overall, approximately 80% of the compound–matrix combinations evaluated satisfied the trueness criteria, although a few compounds fell outside the acceptable range. In previous studies employing a single SPE cartridge for the determination of veterinary drugs in animal-derived matrices, matrix effects exceeding ±20% were observed for many target compounds,²³ and recoveries for approximately one-third of the target compounds were below 50% when quantified using solvent-based calibration.²⁴ In another study utilizing QuEChERS extraction followed by dispersive SPE for the analysis of veterinary drugs in beef, pronounced matrix effects were observed for most of the tested compounds, with ME values significantly exceeding ±20%.² These results indicate that the previously reported methods did not sufficiently remove matrix interferences. In contrast, the present dual-cartridge automated SPE method achieved markedly higher cleanup efficiency, effectively minimizing matrix effects and enabling accurate quantification using solvent-based calibration for most compound–matrix combinations.

Table 1 Trueness and intra-/inter-day precision of the developed method at 0.01 mg kg⁻¹ (%)

Compound	Muscle			Liver			Fat			Milk
	Trueness	Precision		Trueness	Precision		Trueness	Precision		Trueness	Precision
		Intra-day	Inter-day		Intra-day	Inter-day		Intra-day	Inter-day		Intra-day	Inter-day
Sulfabenzamide	88.6	3.1	4.6	87.8	3.1	5.5	89.2	4.0	7.2	90.5	2.3	2.5
Sulfabromomethazine	88.7	2.0	3.6	102.5	2.5	13.7	87.6	3.5	6.1	89.0	2.9	2.9
Sulfachloropyridazine	89.8	5.2	5.2	88.8	3.1	3.9	88.6	3.8	6.3	87.6	3.2	3.2
Sulfadiazine	89.9	3.5	4.8	91.5	3.5	3.5	91.3	3.7	7.5	97.6	1.8	1.8
Sulfadimethoxine	89.9	3.3	5.1	89.4	4.2	4.2	90.2	4.1	8.3	88.4	2.4	2.5
Sulfadimidine	88.0	2.7	5.9	92.9	2.8	5.5	90.4	0.5	7.0	90.1	5.1	5.1
Sulfadoxine	90.0	2.5	4.0	90.0	2.3	3.2	90.1	2.6	7.3	88.3	3.2	3.4
Sulfaethoxypyridazine	90.8	2.6	5.1	89.1	2.6	3.5	89.9	2.8	6.2	86.6	3.1	3.2
Sulfaguanidine	88.6	10.1	12.8	106.6	8.3	16.7	90.6	8.4	8.4	87.7	9.2	9.7
Sulfamerazine	91.5	4.5	5.2	88.3	1.8	3.7	91.1	3.3	6.1	89.1	2.6	2.6
Sulfamethoxazole	90.3	4.5	6.9	87.3	2.7	4.2	90.7	4.0	6.5	88.0	3.7	3.8
Sulfamethoxypyridazine	74.0	2.3	10.7	91.2	3.2	3.2	91.1	2.6	6.6	88.7	4.3	4.5
Sulfamonomethoxine	90.2	2.3	5.4	89.4	3.7	4.7	89.7	3.0	5.9	89.4	3.8	3.9
Sulfanitran	89.0	13.3	17.2	79.7	10.0	18.8	88.0	10.3	25.1	85.3	14.6	15.4
Sulfapyridine	89.8	2.7	4.4	91.1	3.4	3.4	88.7	5.3	7.6	104.4	5.9	6.2
Sulfaquinoxaline	88.2	2.9	4.8	10.3	5.8	42.6	88.3	4.4	7.7	85.3	3.7	3.8
Sulfathiazole	88.2	2.6	4.1	91.7	3.6	4.3	89.6	4.2	6.2	90.5	3.5	3.7
Sulfatroxazole	90.4	2.8	4.7	90.0	2.9	4.7	90.6	2.5	7.0	88.1	2.6	3.0
Sulfisomidine	89.7	2.4	4.0	94.4	3.0	3.0	90.5	2.3	6.4	93.5	3.2	3.2
Sulfisoxazole	88.2	2.7	5.1	84.3	2.6	4.9	89.9	4.1	7.6	88.2	2.6	2.7
Sulfisozole	89.8	5.1	5.6	85.5	3.7	4.9	90.9	2.6	4.1	89.0	4.5	4.6
Sulfacetamide	88.6	4.8	6.6	77.6	1.8	4.6	89.4	4.2	7.9	91.2	3.5	3.6
Erythromycin A	92.1	6.7	18.5	63.4	4.6	51.9	71.8	5.1	24.7	64.6	31.1	32.9
Leucomycin A5	62.5	2.9	8.3	14.5	4.6	41.3	76.1	2.8	15.1	79.9	4.8	5.1
Neospiramycin I	63.3	1.3	7.9	6.9	8.4	103.8	60.5	5.1	11.2	105.8	9.8	10.4
Spiramycin I	63.6	5.9	7.1	9.7	5.9	60.4	60.3	3.8	11.7	92.3	3.8	4.0
Tilmicosin	124.1	6.2	14.9	129.7	3.2	24.2	89.5	2.8	9.3	128.6	22.5	23.8
Tylosin A	56.3	3.5	8.9	13.0	7.0	61.7	65.7	2.3	16.2	80.3	8.4	8.8
Ciprofloxacin	88.1	7.0	7.6	80.7	4.5	6.3	74.5	7.6	10.0	105.6	3.8	4.3
Danofloxacin	136.4	3.1	5.5	146.8	5.6	15.6	120.6	5.5	5.5	148.3	7.4	7.8
Difloxacin	105.6	6.3	8.0	104.8	3.7	5.6	101.0	3.7	7.8	107.4	4.3	4.4
Enrofloxacin	106.6	3.6	4.8	110.9	2.2	6.1	104.0	5.5	6.6	109.7	3.2	3.2
Flumequine	101.8	3.9	8.0	103.5	3.1	10.2	97.4	4.9	10.4	105.0	6.1	6.4
Marbofloxacin	123.2	3.9	5.1	112.7	4.2	5.7	110.6	2.8	6.8	141.8	4.9	5.1
Miloxacin	101.6	3.7	11.0	94.8	3.3	16.1	97.5	3.9	9.6	109.8	8.8	9.3
Nalidixic acid	104.0	2.7	9.8	105.2	2.2	8.2	95.3	4.5	9.9	106.7	6.9	7.3
Norfloxacin	86.3	4.0	5.6	78.8	2.9	7.6	79.4	5.2	12.8	103.8	4.7	4.8
Ofloxacin	113.6	3.8	3.8	114.8	1.1	1.9	110.6	4.1	7.3	123.5	3.9	4.1
Orbifloxacin	99.7	3.9	4.9	95.4	4.5	4.5	96.8	3.2	7.0	101.7	3.2	3.3
Oxolinic acid	98.5	2.7	12.9	100.8	2.3	11.5	97.2	5.4	12.5	108.9	9.9	10.4
Piromidic acid	104.4	5.3	8.0	106.4	3.0	7.2	97.0	4.9	9.7	106.4	6.4	6.8
Sarafloxacin	98.2	5.9	11.5	88.7	5.8	8.8	83.1	3.5	7.1	104.4	5.1	5.2
Ormetoprim	86.0	3.6	3.7	79.9	3.0	3.0	85.3	3.0	5.1	87.2	3.1	3.2
Tiamulin	89.1	3.4	3.8	82.2	2.0	5.9	86.7	2.7	6.1	88.0	1.4	1.4
Trimethoprim	85.5	3.7	4.0	85.2	3.8	4.3	87.2	1.3	6.2	97.6	5.5	5.7
Diaveridine	82.1	2.8	3.1	79.8	2.0	2.0	85.5	2.7	4.7	89.7	2.6	2.8
Pyrimethamine	80.7	2.0	3.7	75.7	4.1	5.1	84.3	2.8	5.8	87.3	1.9	1.9
Clostebol	89.8	3.7	4.0	60.5	2.2	11.9	83.8	3.2	6.2	89.2	2.5	2.7
Methylprednisolone	91.0	6.5	6.5	80.6	4.7	4.7	89.2	4.1	7.3	90.8	3.8	3.8
Prednisolone	90.2	3.4	3.4	79.3	5.5	5.5	88.4	2.3	8.0	90.0	5.4	5.7
α-Trenbolone	85.8	5.7	5.7	84.0	8.1	10.0	85.0	4.1	6.8	88.0	2.4	2.4
β-Trenbolone	84.3	3.6	3.9	70.5	4.9	14.4	84.1	2.8	5.9	88.5	3.5	3.6

---

The fluoroquinolone antibiotics danofloxacin, marbofloxacin, and ofloxacin exhibited recovery values > 110%. Since these compounds showed relatively large MEs in most matrices, the trueness values exceeding 110% were likely due to ionization enhancement during ESI. Previous studies have shown that fluoroquinolones such as ofloxacin and norfloxacin form stable 1 : 1 complexes with Ca²⁺ and Mg²⁺ under physiological conditions,²⁸ which can boost ionization efficiency in ESI-MS by facilitating charge transfer and surface activity. When analyzing real samples, quantification using solvent-based standards may result in slight overestimation; therefore, compounds showing recoveries above 120% in this study were excluded from the target analytes.

Macrolide antibiotics, including leucomycin A5, spiramycin I, neospiramycin I, and tylosin A, showed extremely low trueness values (<15%) in bovine liver compared with other matrices. For all of these compounds, the MEs in liver were not significant, and high recoveries were obtained when the analytes were spiked into the extract, indicating that the losses likely occurred before extraction due to metabolic transformations in the liver. Mourier et al. reported that the aldehyde group in the macrolactone ring of spiramycin I can react with cysteine residues, forming a thiazolidine derivative.²⁹ Because leucomycin A5, neospiramycin I, and tylosin A also contain comparable aldehyde functional groups, similar transformations are plausible. In contrast, the trueness values of erythromycin A and tilmicosin, which do not possess, were not markedly decreased, further supporting the conversion of leucomycin A5, neospiramycin I, and tylosin A to thiazolidine derivatives. Moreover, the use of 0.1% formic acid in both the mobile phase and final test solution could exacerbate acid-induced degradation of macrolides, especially erythromycin A, during LC-MS analysis and sample storage in autosampler vials.³⁰ Similarly, sulfaquinoxaline also exhibited significantly low trueness values (∼10%) in bovine liver, whereas those in muscle, fat, and milk consistently exceeded 80%. Because the ME in bovine liver was within ±10%, the low recovery was not attributable to matrix effects. Furthermore, when the analyte was spiked into the extract, satisfactory recoveries were obtained, indicating that no loss occurred during the cleanup process. Therefore, the low trueness observed in liver samples was presumed to result from degradation prior to extraction. This assumption is supported by the findings of Hoff et al., who demonstrated that sulfaquinoxaline undergoes hydroxylation in the liver of cattle and other animals, leading to reduced recoveries.³¹ Consequently, the pronounced decrease observed in liver samples was most likely due to metabolic transformations, such as hydroxylation, occurring in the homogenized liver tissue.

The intra-day (repeatability) and inter-day (within-laboratory reproducibility) precision, expressed as relative standard deviations (%), satisfied the acceptance criteria (<25% for intra-day and <30% for inter-day) for all compounds that fulfilled the trueness criteria. These results demonstrate that the limits of quantification, defined as the lowest concentrations of the analytes validated with acceptable accuracy using the developed method, were determined for the 52 target compounds, with approximately 80% at 0.01 mg kg⁻¹. Compounds that did not meet the target criteria for trueness or precision were presumed to be affected by degradation or matrix effects and were therefore excluded from the scope of this analytical method.

Application to real samples

To evaluate the applicability of the developed method to real samples, three samples each of bovine muscle, liver, fat, and milk, commercially available in Japan, were analyzed. As a result, all target compounds were below the limits of quantification, and no interference peaks were observed. These results indicate that the developed method effectively removed interfering substances and is suitable for the determination of veterinary drug residues in real bovine matrices.

Conclusions

This study developed and validated an automated SPE method employing a dual-cartridge configuration composed of two C18 cartridges connected in series, with in-line solvent modulation. The first cartridge efficiently removed low-polarity matrix components, thereby preventing overload and maintaining the performance of the second cartridge. Additionally, the introduction of a water addition step between the cartridges optimized the solvent polarity for the second stage, likely enhancing the removal of matrix components responsible for matrix effects. The method demonstrated satisfactory analytical performance, with approximately 80% of analyte–matrix combinations meeting the criteria outlined in the Japanese method validation guideline. Notably, matrix effects were sufficiently minimized to enable accurate quantification using solvent-based calibration alone, eliminating the need for matrix-matched standards or isotope-labeled internal standards. Overall, the developed method provides a robust and reliable platform for the simultaneous determination of multi-residue veterinary drugs in a variety of bovine-derived food matrices. This method is particularly well suited for routine food safety monitoring due to its high cleanup efficiency and operational simplicity. However, some compounds, such as macrolide antibiotics and sulfaquinoxaline, exhibited low trueness in liver samples, likely due to degradation. Further studies are warranted to develop strategies to suppress degradation and improve recoveries for these analytes.

Author contributions

R. M.: investigation, methodology, validation, writing—original draft; S. S.: methodology, conceptualization, project administration, supervision, writing—review & editing; M. S.: investigation, writing—review & editing; T. T. (Takaaki Taguchi): funding acquisition, resources, writing—review & editing; T. T. (Tomoaki Tsutsumi): project administration, resources, supervision, writing—review & editing.

Conflicts of interest

There are no conflicts to declare.

Data availability

The data supporting our article have been included as part of the supplementary information (SI). Supplementary information is available. See DOI: https://doi.org/10.1039/d5ay01515d.

Acknowledgements

This work was supported by the Consumer Affairs Agency of Japan (grant number: JPCACAA22KA1009).

Notes and references

1. R. E. Baynes, K. Dedonder, L. Kissell, D. Mzyk, T. Marmulak, G. Smith, L. Tell, R. Gehring, J. Davis and J. E. Riviere, Food Chem. Toxicol., 2016, 88, 112–122.
2. Y. S. Jung, D.-B. Kim, T. G. Nam, D. Seo and M. Yoo, Food Chem., 2022, 382, 132313.
3. A. Desmarchelier, T. Bessaire, M.-C. Savoy, A. Tarres, C. Mujahid, A. Beck, P. Mottier and T. Delatour, J. AOAC Int., 2021, 104, 650–681.
4. S. Biselli, U. Schwalb, A. Meyer and L. Hartig, Food Addit. Contam.,:Part A, 2013, 30, 921–939.
5. M. L. G. Pérez, R. Romero-González, J. L. M. Vidal and A. G. Frenich, J. Sep. Sci., 2013, 36, 1223–1230.
6. G. Stubbings and T. Bigwood, Anal. Chim. Acta, 2009, 637, 68–78.
7. D. Shin, H.-S. Kang, J. Jeong, J. Kim, W. J. Choe, K. S. Lee and G.-S. Rhee, Food Anal. Methods, 2018, 11, 1815–1831.
8. J. Kim, J. H. Suh, H.-D. Cho, W. Kang, Y. S. Choi and S. B. Han, Food Addit. Contam.,:Part A, 2016, 33, 420–432.
9. M. Aguilera-Luiz, J. M. Vidal, R. Romero-González and A. G. Frenich, Food Chem., 2012, 132, 2171–2180.
10. Z. Hajrulai-Musliu, R. Uzunov, S. Jovanov, D. Jankuloski, V. Stojkovski, L. Pendovski and J. J. Sasanya, BMC Chem., 2021, 15, 1–21.
11. S. Lee, B. Kim and J. Kim, J. Chromatogr. A, 2013, 1277, 35–41.
12. A. Schürmann, C. Crüzer, V. Duss, E. Kämpf, T. Preiswerk and H.-J. Huebschmann, Anal. Bioanal. Chem., 2024, 416, 689–700.
13. A. Martínez, A. R. Aguilera and A. R. Fernández-Alba, Anal. Bioanal. Chem., 2023, 415, 6551–6560.
14. L. M. Sánchez, F. Jesús, C. Ferrer, M. M. Gómez-Ramos and A. Fernández-Alba, J. Chromatogr. A, 2023, 1694, 463906.
15. E. Hakme and M. E. Poulsen, J. Chromatogr. A, 2021, 1652, 462384.
16. A. Goon, R. Shinde, B. Ghosh and K. Banerjee, J. AOAC Int., 2020, 103, 40–45.
17. A. Khaled, E. Gionfriddo, V. Acquaro Jr, V. Singh and J. Pawliszyn, Anal. Chim. Acta, 2019, 1056, 34–46.
18. L. Xia, Y. He, X. Xiao and G. Li, Anal. Bioanal. Chem., 2019, 411, 4073–4084.
19. T. Wei, G. Li and Z. Zhang, J. Sep. Sci., 2019, 42, 888–896.
20. S. J. Lehotay, L. Han and Y. Sapozhnikova, Chromatographia, 2016, 79, 1113–1130.
21. B. D. Morris and R. B. Schriner, J. Agric. Food Chem., 2015, 63, 5107–5119.
22. Ministry of Health, Labour and Welfare, Director Notice: Guidelines for the Validation of Analytical Methods for Testing Agricultural Chemical Residues in Food, Syoku-An No. 1224-1, Japan, 2010, https://www.caa.go.jp/en/policy/standards_evaluation/residues_of_pesticides/assets/181130_21.pdf, accecced on July 31, 2025.
23. J. Wang, W. Zhao, Y. Li, R. Jiang, H. Li, S. Wang and Z. Li, Food Chem., 2022, 393, 133260.
24. W. Zhao, E. Jiang, W. Guo, C. Guo, S. Li, J. Wang, S. Wang and Y. Li, Bull. Environ. Contam. Toxicol., 2021, 107, 228–238.
25. D. Castilla-Fernández, D. Moreno-González, M. Beneuto-Cambra and A. Molina-Díaz, Anal. Bioanal. Chem., 2019, 411, 1433–1442.
26. Y. Zhang, X. Xue, S. Su, Z. Guo, J. Wang, L. Ding, Y. Liu and J. Zhu, Food Anal. Methods, 2018, 11, 2865–2884.
27. M. J. Schneider, S. J. Lehotay and A. R. Lightfield, Anal. Bioanal. Chem., 2015, 407, 4423–4435.
28. H. R. Park, G. Y. Jeong, H. C. Lee, J. G. Lee and G. M. Baek, Bull. Korean Chem. Soc., 2000, 21, 849–854.
29. P. Mourier and A. Brun, J. Chromatogr. B:Biomed. Sci. Appl., 1997, 704, 197–205.
30. Y.-H. Kim, T. M. Heinze, R. Beger, J. V. Pothuluri and C. E. Cerniglia, Int. J. Pharm., 2004, 271, 63–76.
31. R. B. Hoff, F. Barreto, J. Melo, L. Jank, M. d. C. R. Peralba and T. M. Pizzolato, Anal. Methods, 2012, 4, 2822–2830.

---