Penicillin G and Cloxacillin in Costa Rican Dairy Products: Quantitative Analysis and Lactic Acid Bacteria Resistance Profiling

Abstract

Background/Objectives: Milk and dairy products are among the most relevant foods both nutritionally and commercially. Costa Rica stands out as one of the main producers and consumers of dairy products in Central America. However, in recent years, the use of antibiotics in the livestock industry has increased, with implications for public health and food security, generating a need to monitor residues of these drugs in food. The present research focuses on developing a liquid chromatography method for the simultaneous quantification of penicillin G (PEN) and cloxacillin (CLO) in raw and commercial bovine milk, as well as in various dairy products, including fresh cheese and liquid yogurt. Methods/Results: During the validation of the methodology, average sensitivities of (960 ± 8)·10¹ mg L⁻¹ and (1580 ± 9)·10¹ mg L⁻¹ were achieved for PEN and CLO, respectively. Determination coefficients of 0.9995 and 0.9998 were also achieved, respectively. The limits of detection (LOD) and quantification (LOQ) for PEN and CLO were (0.330 ± 0.025) mg L⁻¹ and (0.65 ± 0.12) mg L⁻¹, (1.10 ± 0.083) mg L⁻¹ and (2.2 ± 0.4) mg L⁻¹, respectively, on both accounts. Recoveries were 68–77%, 92–106%, and 78–112% for PEN and 57–79%, 99–114%, and 95–120% for CLO in milk, cheese, and yogurt, respectively, across all three concentration levels evaluated. The precision of the method, in terms of reproducibility, was assessed for liquid yogurt (3–12% RSD_R for PEN and 4–12% RSD_R for CLO) and in cheese (8–14% RSD_R for PEN and 4–12% RSD_R for CLO). Nineteen bovine milk samples from the Cartago area were evaluated, including commercial and milk samples. Additionally, cheese (n = 17) and yogurt samples (n = 8) were analyzed. Conclusions: None of the samples showed detectable signals of the antibiotics. In addition, antibiotic sensitivity testing was performed on n = 9 Lactic Acid Bacteria (LAB) strains isolated from the raw milk samples, revealing the presence of some resistant traits to several antibiotics, including β-lactams.

1. Introduction

Milk and dairy foods are important suppliers of a range of key nutrients, with some being particularly important at certain life stages [1]. These products are of great nutritional, social, and economic importance; they require special attention, given that they are supplied globally by remarkably diverse production systems and technologies [2]. Not only are dairy foods an essential part of the diet, but regular intake of dairy products has been related to a decrease in chronic disease [3,4,5].

In addition to their health benefits, dairy products represent a primary global industry currently undergoing expansion and consolidation. Intrinsic to the dairy industry is its significance and productivity in most countries. The global dairy industry is undergoing a period of expansion and consolidation [6]. Cow milk production has increased slightly over the past half-decade, hovering around 549 million metric tons per year. The European Union produced the most cow milk in 2023, with the United States in second place at just under 105 million metric tons. Consumption of dairy products is considerable, with an estimated 80% of the world’s population regularly consuming them [6]. In 2023, India consumed the most cow milk worldwide, drinking over 87 million metric tons [5]. The next-largest milk consumption was in the European Union, at 23.7 million metric tons. Over 100 countries include dairy in their Food-Based Dietary Guidelines [7]. Similarly, Costa Rica produces 3 million liters of milk daily. Additionally, the sector comprises 300 companies that export 20% of their products to other markets [8,9].

Antibiotic use encompasses the prevention, metaphylaxis, and therapy of common animal pathologies (e.g., mastitis, respiratory diseases, and foot diseases), as well as prophylactic use to control bacterial infections in dairy cattle [10]. In this regard, mastitis is the most common infectious disease in dairy animals, resulting in significant economic losses in the industry [10]. Moreover, the general application of antimicrobials in animals destined for food production has been estimated to account for 73% of global use (80% in the United States alone) [10].

Most dairy farms include β-lactams (penicillin, cefapirin, ceftiofur, amoxicillin, hetacillin, and cloxacillin), macrolides (erythromycin), coumarins (novobiocin), and lincosamides (pirlimycin) as antimicrobial treatments. In the United States alone, at least 16% of all lactating cows receive antibiotic therapy each year to treat mastitis [11]. Additionally, more than 75% of all animals receive prophylactic intramammary antibiotic doses following each lactation, especially β-lactams [12].

Given such veterinary drug use practices, relevant food products, such as dairy, warrant ongoing surveillance and food safety monitoring [13]. Hence, guidelines and maximum levels for veterinary drugs in foods have been established locally (Technical Regulation RTCR 401-2006) and worldwide (Codex Alimentarius CX/MRL 2-2023, 21 CFR Part 556 or Directive 2001/82/EC, Regulation 726/2004, and Commission Regulations [EU] No 37/2010 and 470/2009). The presence of veterinary antibiotic residues in animal products, such as milk, may pose a health issue, causing allergies in humans and facilitating the development of resistant pathogens [13,14]. Therefore, effective veterinary antibiotic monitoring programs require specific, sensitive, and reliable analytical methods capable of detecting all antibiotic drug residues below regulatory levels [13]. It is worth noting that MRLs vary depending on the type of antibiotic, the animal, and the matrix (Table 1, [14]).

Table 1.

Maximum residue limits (MRLs) for the two antibiotics of interest and sensitivity for different determination approaches in milk [15].

Pharmacologically Active Substance	MRLs in Milk, µg kg−1	MRLs in Food-Producing Species, µg kg−1
Penicillin	4	50
Cloxacillin	30	300
Pharmacologically active substance/Method	PEN	CLO
Microbial inhibition tests	1–7	10–68
Rapid test	1.8–9.6	5–60
LC-UV	>4	>10
LC-FLD	>0.005	Not applicable
LC-MS (ESI+-QqQ)	>0.23	>0.1

Most methodological approaches available for determining β-lactams can be grouped into four main categories: i. microbiological approaches based on bacterial growth inhibition; ii. Biosensors iii. immunochemical techniques iv. chromatographic methods. The first two approaches enable rapid screening of samples, but they exhibit limited specificity, detecting only structurally similar compounds. On the other hand, immunochemical techniques can be fast and exhibit higher specificity for single analytes, as well as a disposition for semi-quantitative screening. Then, the need for quantitative, specific results led to the development of methods based on liquid chromatography (LC), especially those that use mass spectrometry (MS) detection [15].

Herein, we describe (i) the validation based on AOAC requirements (AOAC SMPR^® 2018.010) of an HPLC-DAD (diode array detection) salt-less method for two β-lactams [i.e., PEN and CLO] in dairy products (i.e., fluidized milk, soft cheese, and yogurt) (ii) the application of the method to commercial dairy samples, and (iii) veterinary drug formulations. Furthermore, dairy samples were tested for LAB and their resistance profile for several antibiotic classes (including β-lactams).

2. Results

2.1. Method Validation

2.1.1. Chromatographic Performance Parameters

To assess analyte concentration, an eight-point calibration curve was prepared. An average linear range was achieved from (4.0 ± 0.4) to (133 ± 8) mg L⁻¹ and (3.5 ± 0.2) to (118 ± 5) mg L⁻¹ for PEN and CLO, respectively. Under our conditions, PEN and CLO exhibited retention times of (13.89 ± 0.27) min and (16.54 ± 0.26) min, respectively (Figure 1A); calibration curves were represented in Figure 2A,B.

Figure 1.

Chromatograms for spiked samples at 112 and 115 mg L⁻¹ PEN and CLO, respectively, for (A) cheese and (B) yogurt.

Figure 2.

Typical calibration curves and equations obtained for (A). PEN and (B). CLO in two different solvents (aqueous [pure H₂O] and organic [pure MeOH]). Error for slope and intercept represented by the standard deviation of n = 5 independent curves.

Please note that these results were obtained under our experimental conditions. However, PEN and CLO are both found to be solid standards soluble in both H₂O and MeOH; a steeper slope (in terms of linearity) is obtained for both antibiotics, but especially for CLO. In terms of sensitivity, this represents 1.31-fold more when MeOH is used. This suggests a higher extinction coefficient and resistance to methanolysis. Then, solvent selection speaks to method robustness.

Detection limits (LOD) for PEN and CLO were (0.330 ± 0.025) mg L⁻¹ and (0.65 ± 0.12) mg L⁻¹, respectively. For n = 4 independent replicates (injections) of solutions at (119.0 ± 0.2) mg L⁻¹ and (141.0 ± 0.2) mg L⁻¹ of PEN and CLO, respectively, intra and inter-day variation was recorded at 1.74, 1.46, 0.55, and 1.16%, respectively. Finally, in terms of accuracy, recovery tests were performed for both analytes; the results are summarized below (Table 2). It was observed that at low and medium milk concentrations, more favorable recovery values are obtained than at high concentration levels. However, when working with cheese and yogurt, the results were favorable at all three levels but with greater variability.

Table 2.

Accuracy data obtained for dairy products (milk, cheese, yogurt).

Spike Level/Parameter	PEN	CLO	PEN	CLO	PEN	CLO
	Low Level		Medium Level		High Level
Milk sample
	Concentration, mg L−1
Theoretical/Expected Value	15	16	59	61	84	92
Experimental value	11.3 ± 0.8	12.3 ± 0.8	45.2 ± 0.8	46 ± 1	57.5 ± 0.8	52 ± 1
Recovery, %	76.8	78.9	76.5	72.1	68.4	57.0
Cheese Sample
	Concentration mg L−1
Theoretical/Expected Value	21	22	53	54	89	88
Experimental value	19.6 ± 3.4	22.8 ± 2.6	56.3 ± 7.6	65.5 ± 5.3	90 ± 13	100 ± 13
Recovery, %	92 ± 18	99 ± 10	106 ± 15	119 ± 10	101 ± 15	114 ± 14
Yogurt Sample
	Concentration mg L−1
Theoretical/Expected Value	21	22	52	54	89	88
Experimental value	23.5 ± 0.9	22.0 ± 1.6	41.4 ± 5.7	51.3 ± 3.6	81.1 ± 9.7	111.0 ± 0.3
Recovery, %	112.0 ± 4.0	94.7 ± 9.9	78 ± 11	93.2 ± 7.0	91.0 ± 7.0	119.6 ± 0.3

2.1.2. Antibiotic Stability in Milk

Both antibiotics exhibit similar behavior when refrigerated and frozen (Figure 3A,B). However, CLO degradation is insignificant, while PEN is highly susceptible to hydrolysis (Figure 4A,B). With an exponential rate of degradation, PEN by day 192 has lost more than 30% and 15% of its initial concentration when stored at −20 °C and −70 °C, respectively (Figure 3A). By the end of the experiment (384 days), PEN storage at −70 °C seems to delay hydrolysis (Figure 3A). For pH modifications, PEN degradation was observed at pH values below 6 (Figure 4A), whereas CLO degradation occurred at basic pH values (Figure 4B).

Figure 3.

Antibiotic stability assay for (A). PEN and (B). CLO. Analytes were stored at two different freezing temperatures. Asterisk(s)-marked bars, * and **, denote statistical differences at p < 0.05 and p < 0.01, respectively.

Figure 4.

Stability in aqueous solutions of 200 mg L⁻¹ at different pH, and at 25 °C for (A) PEN and (B) CLO.

The persistence of the parent compounds at −70° implies that during sample extraction and standard solution storage, the samples and standards should be kept under these conditions to preserve the antibiotic as is. At the same time, samples and standard solutions should be analyzed <192 days of storage. Failing to comply with these conditions may result in a false negative for the presence of possible antibiotic residues in samples.

In the case of β-lactams, the chemical reactivity of their ring has long been attributed to its inner tension. There is structural tension within the β-lactam that stems from the angles of bonding between the atoms (90°) are clearly inferior to the corresponding angles of sp³ hybridization (109.47°) or sp² (120°) [16]; in this way, the characteristic resonance of the amide moiety is interrupted, resulting in the instability of the structure of these compounds. Hence, peaks for PEN present at pH < 6 (Figure 4A) and at 12.9 min for CLO (Figure 4B, at pH > 10) are caused by hydrolysis (see discussion). In general, β-lactams are susceptible to hydrolysis and aminolysis. As shown, penicillin G undergoes pH-dependent hydrolysis, with negligible spontaneous degradation or H₂O-catalysed degradation (Figure 5).

Figure 5.

Construction of (A) PEN and (B) CLO show structural rigidity and nitrogen delocalized electrons in the β-lactam ring. A cis arrangement of hydrogen is also required for biological activity. (C) Penicillin can react with H₂O under acidic conditions in a hydrolysis reaction. The reaction mechanism is a nucleophilic acyl substitution reaction that forms penicilloic acid, which lacks the desired antibiotic activity—the penicillin is rendered useless. (D) The carbonyl moiety within the β-lactam ring is the most electrophilic site. (E) The C=O bond in the acylamino sidechain is not susceptible to nucleophilic attack as the nitrogen atom can feed its lone pair into the carbonyl group, rendering it a weaker electrophile. (F) Similarly, the carboxylate ion is not susceptible to nucleophilic attack because the oxygen atom can feed its lone pair into the adjacent carbonyl group. (G) The acylamino sidechain can act as an internal nucleophile and attack the β-lactam carbonyl, opening the β-lactam ring, forming a very strained intermediate. To impede the involvement of the acylamino side chain, an electron-withdrawing substituent is added (like in CLO). This makes the amide carbonyl group a weaker nucleophile, which, in turn, is less likely to react with the β-lactam carbonyl. The wavy lines indicate that only a partial structure is shown.

2.2. Quantitative Analysis of Dairy Samples

A total of n = 44 dairy product samples were analyzed for residues of penicillin (PEN) and cloxacillin (CLO) (n = 19 milk, n = 17 cheese, n = 8 yogurt). The milk subgroup comprised n = 15 raw milk and n = 4 ultra-high temperature (UHT) milk specimens. The cheese samples were exclusively fresh, low-humidity (less than 55%) “Turrialba type” cheeses, characterized by a short maturation period. The yogurt samples were pasteurized liquids exhibiting a low pH (3–4), a condition under which PEN hydrolysis would be anticipated. All 44 tested samples yielded results below the Limit of Detection (LOD) for both target analytes. For method quality control, n = 7 matrix-spiked samples were simultaneously analyzed. These exhibited recovery rates of 58.7−102.7% for PEN and 61.5−116.9% for CLO. Furthermore, the purity of the analytical standard was confirmed by testing a commercial PEN injectable, which showed a bias of 3−5%.

2.3. Antibiotic Resistance Patterns

As shown in Figure 6 above, the MAR index ranged from 0.07 to 0.50. Interestingly, most isolates (n = 9) exhibited different MAR patterns. However, isolate number two exhibited sustained resistance toward β-lactams (red-colored box). The multiple antibiotic resistance (MAR) index is described as a cost-effective and valid method for tracking the source of bacteria. Indices are larger than 0.2 if an isolate originates from a source where antibiotics are used to a great degree and/or in large amounts. Especially worrisome is an intermediate resistance toward VAN (all isolates colored yellow). VAN is a tricyclic glycopeptide antibiotic used to treat severe Gram-positive bacterial infections.

Figure 6.

Antibiotic resistance profiling for n = 9 isolates of LAB recovered from milk samples. Key: PEN: Penicillin G; AMO: Amoxicillin; AMP: Ampicillin; CTX: Ceftriaxone; MEM: Meropenem; SXT: trimethoprim/sulfamethoxazole; AMK: amikacin; STR: Streptomycin; GEN: Gentamicin; CIP: Ciprofloxacin; TET: Tetracycline; ERY: Erythromycin; CHL: Chloramphenicol. Resistant, Intermediate, Sensitive, < 0.1 from 0.1 to 0.2 > 0.2.

Overall resistance was 23.8% (n = 30/126) across all strains/antimicrobials. Individually, most resistance was observed for STR (88.9% n = 8/9), followed by PEN and AMK (44.4%, n = 4/9), and AMO and AMP (33.3%, n = 3/9). Bacteria showed intermediate tolerance for CIP and VAN (66.6 and 88.9%, respectively) (Figure 6). Strain coded as number 2, obtained from a raw milk sample (T-006) from Turrialba, Cartago, exhibited the most complex resistance profile, with a MAR index of 0.50 (PEN + AMO + AMP + CTX + MEM + SXT + VAN) (Figure 6).

3. Discussion

3.1. Method Validation: Dairy Products

Because the milk matrix contains interferences such as proteins, lactose, and inorganic ions, the success of the extraction procedure depended on the effective deproteinization and washing steps [17]. In this regard, MeOH appears to be an adequate eluent following solid-phase extraction (SPE). Inclusion of mobile phase (phosphoric acid aqueous solution) or acetonitrile (ACN) in the elution mixture caused chromatographic interferences near the retention times of the analytes (approximately between 12 and 16 min, Figure 7B), probably by destabilizing the remaining protein subunit’s micellar structure [18,19,20].

Figure 7.

Chromatograms for (A) PEN and CLO standard solutions at ca. 10 mg L⁻¹. (B) Interference and matrix effects exhibited during elution with aqueous-based media, phosphoric acid solutions, or ACN and their mixtures. (C) Spiked milk sample coded C-007 with PEN and CLO standard solutions at ca. 125 mg L⁻¹. (D) Stability of antibiotics after extended storage (both at −20 °C, blue line, and −70 °C, dark red line), pH adjusted to 12. The transient signal next to PEN highlights the first evidence of partial hydrolysis. (E) Blank/Negative sample for antibiotic residues PEN and CLO coded SJ-008 (blue line) versus a spiked sample after cleanup (dark red line).

Hence, matrix effects were considerably reduced when MeOH was used as the elution solvent, as evidenced by our recovery experiments (Figure 7C). Further comparison between a blank milk sample and a spiked one demonstrates that these matrix effects do not interfere with the analysis of PEN and CLO (Figure 7E). Finally, antimicrobial instability at pH extremes is again suggested as an additional signal, as indicated in the chromatograms (Figure 7D).

Special attention has been given to chromatographic methods for quantifying β-lactam antibiotics (among the most frequently used antimicrobial agents fed to dairy livestock [21]), primarily in milk samples (Table 3). Most methods based on HPLC-DAD achieve sensitivities near the maximum residue limits set for these compounds (established at 4 and 20 µg kg⁻¹ milk for PEN and CLO, respectively, Table 3 [22,23]). All procedures for β-lactams use a reverse-phase approach, and gradient elution is unavoidable when separating multiple compounds simultaneously (Table 3). Despite some researchers opting for buffer systems, we decided to eliminate the possibility of salt precipitation within the chromatographic system, even though phosphate (the most commonly used ion) is kosmotropic [24].

Table 3.

Method comparison with other reported HPLC-based procedures for the detection of β-lactams.

Matrix	Sample Pretreatment	Mobile Phase	Detector/Wavelength, nm	Chromatographic Column	Sensitivity (LOD/LOQ)	Reference
Milk
Milk	Sulphuric acid 0.17 mol L−1 + Sodium tungstate + 10 mL NaCl 20 g/100 g, SPE C18	A: ACN B: 0.2 phosphate buffer (60:20), isocratic	UV/1,2,4-triazole-mercuric chloride, 325	Nova-Pak C18, 4 µm, 150 × 3.9 mm	3 µg L−1	[25]
Cows’ raw milk	30 mL milk + centrifugation + 20 mL H2O + 6 mL 0.17 mol L−1 H2SO4 + 5.6 mL sodium tungstate + SPE C18	A: Phosphate buffer (0.1 mol L−1 pH = 6.5) B: ACN, gradient, 1 mL min−1	UV/1,2,4-triazole-mercuric chloride, 325	Nova Pak C18, 4 µm, 150 × 3.9 mm.	1.3 µg L−1	[26]
Milk	10 mL milk + 2 mL of 0.1 mol L−1 Et4NCl + 0.01 mol L−1 pH = 6 buffer (5:1 KH2 PO4–Na2HPO4) + evaporation	5 mmol L−1 KH2PO4 –acetonitrile (72:28), gradient, 1 mL min−1	UV/215	Supelcosil LC-18-DB	Not reported	[27]
Milk	Ethanol + ultracentrifugation	ACN and 0.1 mol L−1 phosphate buffer (pH 6.5) (35:65)	DAD/1,2,4-triazole-mercury (II) chloride derivatization, 325	Mightysil® RP-4GP	0.004 μg mL−1	[28]
Milk	Liquid–liquid extraction, 5 mL milk + 5 mL ACN, centrifugation, N2 preconcentrating, SPE (Oasis® HLB)	A: H2O/Trifluoroacetic acid (TFA) B: ACN/TFA, at 14 µL min−1	DAD/220	Luna C18, 5 µm, 150 × 0.5 mm	0.80–1.40 g L−1	[29]
Ewe milk	5 mL milk + 5 mL ACN + cold centrifugation + SPE (Spe-ed C18)	A: 25 mmol L−1 phosphate at pH = 3.4, B: CAN	DAD/210	Supelcosil LC 18 24 DB, 5 μm, 150 × 4.6 mm	3.4–8.6 μg kg−1	[30]
Cow milk	10 g milk + HCl pH = 4.5 + acetone/CHCl3 + centrifugation + evaporation	A: 5 mmol L−1 phosphate-buffer solution at pH 4.7 B: ACN, (75:25), gradient, 2 mL min−1	DAD/215	Inertsil ODS-C18, 3.5 µm, 250 × 4.6 mm	7.67 μg kg−1	[31]
Milk	1 mL milk centrifuged + acetone/ACN + µSPE (mesoporous carbon) + NaCl 5 g/100 mL	A: 5 mmol L−1 phosphate buffer (pH 6.0) B: ACN (75:25) at 1.0 mL min−1.	UV/215	Zorbax SB-C18, 5 µm, 100 × 4.6 mm	2.0–3.3 µg L−1	[32]
Raw milk	Dispersive extraction by QuEChERS in MSPD format	A: Ammonium acetate B: ACN, gradient 0.9 mL min−1	DAD/240 & 265	Perfectsil ODS-2, 5 μm, 250 × 4 mm	2–50 μg kg−1	[21]
Milk	10 mL milk + 20 g sodium sulfate + 20 mL ACN + centrifugation + SPE C18	A: H2O/Formic acid (FA) B: H2O/ACN/FA, (35:65), isocratic	UV/280 nm	C8 Enable, 5 µm, 250 × 4.6 mm	9–700 µg kg−1	[33]
Raw cow milk	10 mL + 10 mL McIlvaine. Centrifugation. SPE (Supel Select HLB)	A: H2O/FA, B: oxalic acid/ ACN/MeOH (60:30:10), C: ACN/H2O/FA (24.9:75:0.1), D: ACN/MeOH (2:1, 0.1 mL/100 mL FA).	DAD/210	HP-ODS Hypersil C18, 5 μm, 125 × 4 mm	0.006/0.018 g L−1	[34]
Other matrices
Animal feed	50 g sample 75 MEOH, N2 cocnetration, H2O, SPE (Strata-X 200 mg)	A: H2O, B: TFA/triethylamine/ACN	DAD/230	Zorbax SB-C18, 5 µm, 250 × 4.6 mm	3.89/13.0 g ton−1	[35]
Human Plasma	Centrifugation + H2O + ACN	A: 10 mmol L−1 phosphoric acid (pH = 2) B: ACN, (starting at 7 to 49%) gradient, 2 mL min−1	DAD/210 nm	Atlantis T3, 5 µm, 200 × 4.6 mm	1–2 µg mL−1	[36]
Equine Urine and Plasma	pH adjusted to 7.0, 3 mL phosphate buffer (0.1 mol L−1) + filter	A: Phosphate buffer (0.1 mol L−1) + 16 mmol L−1 potassium thiosulfate (pH = 4.5) B: ACN (74:26), isocratic, at 0.350 mL min−1	UV/1,2,4-triazole-mercuric chloride, 325	Zorbax SB-C18, 150 × 2.1 mm	0.05–0.1 µg mL−1	[37]
Penicillium chrysogenum	ACN/FA formic acid + ultrasound + SPE (Oasis® HLB) n-hexane + ACN & MeOH elution	A: ACN B: 0.3 mL/100 mL FA, isocratic, 1 m min−1	DAD/215	Kromasil KR100-5 C18, 5.0 µm, 150 × 4.6 mm	0.006/0.02 mg kg−1	[38]
Pharmaceutical formulations	Dilution with mobile phase and filtration 0.45 µm	A: 30 mmol L−1 phosphate buffer, pH 4.0 B: ACN, gradient, 1 mL min−1	UV/225	Shodex C18, 5 µm, 250 × 4.6 mm	Not reported	[39]
Honey	0.1 mol L−1 EDTA-McIlvaine buffer (pH 4.0), centrifugation, SPE (Supel Select HLB), MeOH elution	KH2PO4/ACN (80:20), gradient, 1 mL min−1	UV/230	Zorbax Eclipse C18, 5 µm, 250 × 4.6 mm, 35 °C	1 mg kg−1	[40]

In the case of fresh cheeses and liquid yogurts, which undergo enzymatic and fermentation treatments during processing, the transformation and availability of the different proteins present results in fewer interferences in the chromatogram (Figure 1A,B), allowing better identification and quantification of the PEN and CLO signals. The challenges encountered in implementing this methodology for cheese were ensuring sample homogeneity. Variability between replicates was observed, with 8–14% RSD for PEN and 4–12% RSD for CLO, which were higher than those observed in milk but comparable to those reported in other studies in the same matrix [41]. For liquid yogurt, variability between replicates ranged from 3–12% RSD for PEN and 4–12% RSD for CLO, which was also comparable to other studies [21], but for this matrix, was observed that pH (4–5) caused the hydrolysis of penicillin, so for its quantification, the signals at 7 min and 13 min were summed (Figure 5B); despite these the recovery results were acceptable at all three enrichment levels.

From a methodological standpoint, although LC-MS/MS offers superior sensitivity, HPLC-DAD provides moderate detection limits that are nonetheless sufficient to meet established maximum residue limits (MRLs) for β-lactam antibiotics [30,42]. Both techniques comply with regulatory thresholds; however, LC-MS/MS is typically preferred for confirmatory analyses due to its ultra-trace capabilities [43]. From a practical perspective, HPLC-DAD is significantly more cost-effective and broadly accessible, particularly in regional laboratories where LC-MS/MS may be unavailable due to its high acquisition and maintenance costs, as well as the need for specialized infrastructure and expertise [44] HPLC-DAD is also easier to operate, requiring minimal technical training, and is well-suited for routine surveillance programs. Additionally, its maintenance demands are relatively low compared to the frequent calibration and servicing required by LC-MS/MS systems.

The method developed in this study demonstrated excellent linearity (R² ≥ 0.9995), acceptable recoveries across matrices, and LOD/LOQ values below the MRLs, ensuring both analytical reliability and regulatory compliance [42,45].

3.2. Antibiotic Stability

β-lactams may undergo structural breakdown (i.e., hydrolysis over time), especially in penicillins, as shown in Figure 5A,D. Non-enzymatic hydrolysis can occur due to the nucleophilic attack on the carbonyl carbon atom of the β-lactam ring (i.e., basic hydrolysis) or by the protonation of the nitrogen atom of the β-lactam ring (i.e., acid hydrolysis) [46,47,48]. When the structures of PEN and CLO are compared, PEN has a simpler β-lactam core, which can make it more susceptible to hydrolysis. CLO, for its part, has a substituted 6-aminopenicillanic group, which may give greater stability to the molecule. Notably, the hydrolysis rate and overall reactivity of β-lactams have also been related to metabolic rate and antibiotic efficacy [49]. Both penicilloic and penilloic acids are of relevance since they can remain in milk after degradation [50]. Furthermore, under similar conditions, these related compounds elute before their parent compounds [51].

3.3. Method Application to Real Dairy Samples

As no samples were positive for PEN or CLO it can be cautiously assumed that none of these antibiotics are used continuously within the milk production system (samples were selected from a national milk supplier representative of the ca. 65% of the production, [9]) or at least a proper maintenance of withdrawal periods after antibiotic treatment can be inferred [52], which minimizes the risk of residues in milk [45]. Nevertheless, other artificial methods to decrease these values below permitted levels cannot be ruled out [53].

3.4. Antimicrobial Susceptibility Testing of Bacterial Isolates from Milk

Surveillance of the antibiotic resistance profile of bacteria from milk has been proposed previously [54], as some groups can serve as reservoirs for antibiotic resistance determinants [54]. For example, lactic acid bacteria (LAB) genera such as Lactobacillus, Lactococcus, Leuconostoc, Pediococcus, Streptococcus, and Bifidobacterium spp. have been isolated and identified from raw milk, cheese, and yogurt [55]. In this regard, gene transfer can be facilitated, as bacteria of the genera Lactobacillus and Bifidobacterium are commonly used as probiotics [56]. Although no strain identification was performed during our work, the reported species are well-established in dairy foods; however, it is suggested to include identification and the analysis of the presence of plasmids as future work, so that resistance profiles can be linked to strain identity and potential transfer. Still, studies incorporating more isolates will be necessary for a proper assessment.

Surprisingly, most reports on antibiotic resistance traits in bacteria from milk products are associated with food-borne pathogens, such as Staphylococcus aureus, or LAB from fermented foods [57]. Few reports, such as those by Munsch-Alatossava and Alatossava [58], confirm that resistance traits can be found in up to 60% of psychrotrophs isolated from raw milk; this study reports higher resistance levels to β-lactams, but Streptomycin was not assessed. In addition, a different methodology (ATB^TM [Automatic Testing of Bacteriology] strip) was employed to determine antibiotic resistance, underscoring the need for further studies using standard methods.

In this study, most of the isolates (5/9) showed a MAR index greater than 2.0, indicating a high level of antibiotic resistance. These values suggest that bacteria in Costa Rican milk harbor genetic traits consistent with significant exposure to antibiotics. However, this study’s results do not indicate that antibiotic exposure is associated with contamination of milk or dairy products, and additional sources along the production chain should be considered. Previous studies have assessed the antibiotic contamination of animal feed and water in Costa Rica. For example, it has been reported that up to 74.6% of animal feed samples (non-medicated) presented antibiotic residues [59]. Likewise, hazard assessment of pharmaceuticals in Costa Rica established that some antibiotics detected in water such as ciprofloxacin, doxycycline and norfloxacin should be considered as priorities during the implementation of environmental policies in the country [60]. This suggests a lack of control of antibiotics residues in Costa Rica which may lead to an overexposure of bacteria in the environment and the subsequent development of resistance traits.

4. Materials and Methods

4.1. Reagents

Sodium tungstate dihydrate (Na₂WO₄·2H₂O, catalog number 28195.238, ≥99%, AnalaR NORMAPUR^® analytical reagent) was acquired from VWR BDH Chemicals (Radnor, PA, USA). Sodium chloride (NaCl, ACS reagent, ≥99.0%, catalog number S9888), phosphoric acid (suitable for HPLC, LiChropur™, 85%, catalog 49685), Penicillin G potassium salt (C₁₆H₁₇N₂O₄S·K, VETRANAL^®, analytical standard, catalog 46609), Cloxacillin sodium salt (C₁₉H₁₇ClN₃NaO₅S, ≥95.0%, catalog number 27555), sodium hydroxide (reagent grade, ≥98%, anhydrous pellets, catalog number S5881), acetonitrile (ACN, suitable for HPLC, gradient grade, ≥99.9%, catalog 34851) and methanol (MeOH, suitable for HPLC, ≥99.9%, catalog 34860) all acquired from Sigma-Aldrich (EMD Millipore, Burlington, MA, USA). Gram’s decolorizing, safranine, iodine, crystal violet, and hydrogen peroxide solutions (catalog numbers 1.10218, 1.09217, 90107, 1.09218, and 88597, respectively) were also from Sigma-Aldrich. Sodium salt of Penicillin G (used as secondary standard, 1,000,000 IU, sterile powder for intramuscular injection, Vitalis, Bogotá, Colombia). De Man, Rogosa, Sharpe (MRS), M17, and Mueller-Hinton agars were purchased from Thermo Scientific™ (Oxoid™, Basingstoke, Hampshire, UK, catalog numbers CM1153, CM0785B, and CM0337, respectively). Bacteriological Peptone was also bought from Oxoid™ (catalog number LP0072). Ultrapure water (type I, 0.055 μS cm⁻¹ at 25 °C, 5 μg L⁻¹ TOC) was obtained using an A10 Milli-Q Advantage system and an Elix 35 system (EMD Millipore, Burlington, MA, USA).

4.2. Samples

Milk samples were contained in polyethylene bottles (Nalgene^®, catalog number B7907, Sigma-Aldrich), carried immediately to the laboratory, and fractionated in portions of 100 mL each using sample bags (Whirl-Pak^®, 118 mL capacity, catalog WPB00679WA, Sigma-Aldrich). Commercial samples were purchased in 1L amounts.

All serum was extracted from the cheese sample, and the sample was then reduced to cubes measuring 3 cm on a side. All the cubes were blended until a pasty and homogeneous mass was formed. A portion of 100 g of each homogenized sample was taken and stored in 50 mL conical polypropylene centrifuge tubes (catalog number 430828, Corning^®, New York, NY, USA).

All cheese and milk subsamples were stored at −76 °C until analysis. The yogurt samples were stored in their original commercial packaging and kept frozen at −76 °C. For treatment, the samples were thawed at room temperature, then shaken to ensure homogeneity before collection.

4.3. Chromatographic Performance Parameters and Method Validation

4.3.1. Sample Pretreatment

To a sample mass (20 g of fluid milk, 1 g of cheese, and 1 g of yogurt) contained in a centrifuge tube (50 mL conical, polypropylene, catalog number 430828, Corning^®, New York, NY, USA), 10 mL of water was added as a diluent. Aliquots of 2 mL of a solution of 5 g Na₂WO₄·2H₂O/100 mL and phosphoric acid (170 mmol L⁻¹, pH = 1.54) each were then added to the mixture to achieve protein precipitation [25]. The blend was shaken for 30 s (Scientific Industries Vortex Genie^® 2 Vortexer, Bohemia, NY, USA), then centrifuged at 3260× g at 4 °C (Sorvall™, ST16R, catalog number 75007203, ThermoFisher Scientific, Waltham, MA, USA) for 10 min. Milk samples were gravity sieved using glass microfiber filters (Cytiva™, Whatman™, Grade GF/F, 0.7 µm 90 mm, catalog number 01-184-861, Fisher Scientific, Waltham, MA, USA), while cheese and yogurt samples were filtered using Whatman paper #41 (Whatman™, Quialitative P8, 12.5 cm, catalog number 09-795, Fisher Scientific, Waltham, MA, USA) To the resulting filtrate, 10 mL of a 20 g NaCl/100 mL solution was added. The resulting solution is passed through a previously conditioned SPE vacuum cartridge (Sep-Pak^®, C₁₈, 3 cc, 500 mg sorbent, 55–105 µm, Waters™, catalog number WAT020805, Milford, MA, USA). Cartridges were conditioned using aliquots of MeOH and an aqueous solution of 2 g NaCl/100 mL. Solvents and sample extracts were transferred through the columns at a maximum flow rate of 1 mL min⁻¹ with an SPE 12-port vacuum manifold (operating at 15 mmHg, 57.044, Visiprep™, Supelco Inc., Bellefonte, PA, USA). Washes are discarded, and each solution is added in 1.00 mL increments. After sample addition and washing with an additional 5 mL of the 2 g NaCl/100 mL solution, the analytes are recovered by elution with eight fractions of 250 µL each of MeOH. Further cleanup of the resulting solution was achieved through a syringe filter (catalog number 1751AQ, Minisart^®, Polytetrafluorethylene (PTFE), Pore Size 0.2 µm, Göttingen, Germany) and finally transferred to a HPLC vial (2 mL, Type 1 borosilicate amber glass, PTFE/silicone screw cap and septa, catalog number 5182-0716, Santa Clara, CA, USA) (Figure 8A).

Figure 8.

(A). Sample treatment for the determination of PEN and CLO in raw and commercial cow milk using HPLC-DAD. (B). Procedure used to perform bacterial isolation and Antibiotic sensitivity test on raw and commercial cow milk samples. Created with BioRender.com.

4.3.2. Chromatographic Conditions

Detection settings were adapted from the work by Verdier and coworkers [36]. Briefly, a Shimadzu system (Prominence LC-20A, Shimadzu, Kyoto, Japan) equipped with a photodiode array detector (SPD-M20AV), a column compartment (CTO-20A), an autosampler (SIL-20A HT), a degasser (DGU-20A5), and a quaternary pump (LC-20AT) was used. Chromatographic data were processed using the proGamma LabSolutions Lite software (version 5.82, Shimadzu Corporation). A Zorbax Eclipse C₁₈ column (250 mm × 4.6 mm, 5 μm particle size, catalog number 959990-902, Agilent Technologies, Santa Clara, CA, USA) was used to perform the separation. A gradient elution using phosphoric acid (A, 10 mmol L⁻¹ at pH = 2.6) and ACN (B) was set as follows: at 0 min 93% A, at 6 min 93% A, at 16 min 81% A, at 17 min 49% A, at 22 min 93% A. Flow rate during chromatography was set at 1.8 mL min⁻¹. Detection was performed at 210 nm. The injection volume was set to 10 µL unless stated otherwise.

4.3.3. Performance Parameters

Calibration curves were prepared using methanolic solutions of both PEN and CLO at approximately 100 mg L⁻¹, and then the injection volume was cycled from 1 to 10 µL. Detection/Quantitation limits were calculated based on the standard deviation of the response (S_y) and the slope of the calibration curve (S), then LOD = 3.3·(S_y/S). The recovery was evaluated at three concentration levels—low, medium, and high—using commercial milk, cheese, and yogurt samples fortified with PEN and CLO. Appropriate standard solutions were prepared for the enrichments. Intra- and inter-day reproducibility was assessed by measuring the signal variation in a solution containing approximately 100 mg L⁻¹ of both analytes over time. Four injections of 10 µL of the standard solution were administered at four-hour intervals. The coefficient of variation of the data obtained was calculated. Additionally, stability tests were performed to evaluate the antibiotic response when diluted in MeOH at three different pH values. Area variation was measured over time for a ca. 200 mg L⁻¹ solution that was prepared and divided into two portions, each stored at different temperatures, i.e., −20 and −76 °C. Additionally, these solutions were independently adjusted with a solution of 3 mol L⁻¹ NaOH or 1 mol L⁻¹ H₃PO₄ at pH levels of 2, 3, 4, 6, 8, 11, and 12.

4.3.4. Quality Control and Statistical Analysis

Unless stated otherwise, three independent replicates were performed for each analysis. Recovery of spiked samples at 58 and 62.5 mg L⁻¹ of PEN and CLO, respectively, was used daily or each time a sample was measured as quality control. Where applicable, results are expressed as the mean of the replicates and their standard deviation. Fapas^® FCVD18-DRY4QC (Bovine Milk, Veterinary Medicines, β-lactams, Sand Hutton, York, UK) was also used as a standard. The coefficient of determination (r) was used to corroborate the association between drug concentrations and detector response. A value of r ~ 0 was deemed as a lack of correlation. Calibration curves were constructed each time an analysis was to be performed. This data was evaluated using Sigmaplot 15.0 software (Systat Software Inc., San Jose, CA, USA). A one-sample t-test was performed to assess antibiotic degradation using the initial nominal concentration as the reference. For any statistical analysis, α < 0.05 was considered the threshold for significance; all analyses were performed with SPSS^® Statistics (IBM^®, version 29.0, Armonk, NY, USA).

4.4. Bacterial Isolation

4.4.1. Bacterial Priming: Plating and Isolation from Milk

To isolate bacteria from milk, serial dilutions (from 10⁻¹ to 10⁻⁷) were prepared by adding the first mL of the sample to 9 mL of sterile distilled water [61,62]. Later, 0.1 mL of the last serial dilution was spread evenly onto a previously dried MRS/M17 agar plate; these media support the isolation of Gram-positive bacilli and cocci, respectively. After inoculation, the plates were incubated under anaerobic conditions using an anaerobic candle jar at 37 °C for 48 h [61,62]. After successful bacterial growth on agar, morphologically distinct colonies were further isolated by streaking onto new agar plates and incubating at 37 °C for 24–48 h [63]. Biochemical tests used to characterize the colonies further included the Gram reaction and the catalase test. Only Gram-positive bacteria with catalase-negative responses were recovered (potential lactic acid bacteria)

4.4.2. Antimicrobial Susceptibility Testing

Antimicrobial susceptibility testing (AST) was performed using the Kirby-Bauer disk diffusion method, as outlined in the Clinical and Laboratory Standards Institute guidelines [64]. Briefly, two or three fresh colonies were suspended in 3 mL saline solution, and the turbidity of the suspension was standardized to match the 0.5 McFarland standard (approximately 1.5 × 10⁶ CFU mL⁻¹). The bacterial inoculum was evenly spread over the surface of the Mueller-Hinton agar plate, onto which the antimicrobial disks were manually placed within 15 min. Plates were incubated for 16–24 h at 35–37 °C before the determination of results (i.e., inhibition zones). The diameters of the inhibition zones surrounding the disks were measured manually and compared with the break points reported by FEEDAP [65]. The disk diffusion was performed against n = 16 antimicrobials under 8 groups including β-lactams [penicillin (PEN) 10 IU, amoxicillin (AMO) 30 µg, ampicillin (AMP) 10 µg, ceftriaxone (cephalosporin CTX) 10 µg, meropenem (carbapenem MEM) 10 µg], diaminopirimidines [Trimetoprim-Sulfametoxazol (SXT) 25 µg], aminoglycosides [amikacin (AMK) 30 µg, streptomycin (STR) 10 µg, gentamicin (GEN) 10 µg], fluoroquinolones [ciprofloxacin (CIP) 5 µg], tetracyclines [tetracycline (TET) 30 µg], macrolides [erythromycin (ERY) 15 µg], fenicols [chloramphenicol (CHL) 30 µg], glycopeptides [vancomycin (VAN) 30 µg] (Figure 8B). Isolates resistant to three or more antimicrobials were defined as multidrug-resistant isolates [66]. Intermediate resistance is considered its entity. Nevertheless, it should be considered that the acquisition and transition from susceptibility to resistance had already begun [67]. The Lactobacillus acidophilus ATCC™ 4356™ strain was used as a known positive control.

4.4.3. Multiple Antibiotic Resistance (MAR) Calculation

The MAR index is calculated as the ratio of the number of resistant antibiotics to which the organism is resistant to the total number of antibiotics to which the organism is exposed [68].

5. Conclusions

The extraction method presented herein highlights the peculiarities that dairy products offer during analysis, especially for compounds in the low mg/kg range, as is the case for most veterinary drug residues. Although the method appears suitable for its intended purpose and relatively easy to implement in most laboratories, achieving lower limits of detection with newer technologies is essential for food safety. Lactic acid bacteria appear to be a convenient target for testing antibiotic resistance in dairy products. Although no antibiotic residues were detected in the tested samples, evidence of PEN use (and analogs) is indicated by bacterial resistance. However, some bacteria are inherently resistant to antibiotics, and cross-resistance caused by other xenobiotics is also possible.

Author Contributions

Conceptualization, F.G.-C., M.R.-S. and A.F.-C.; methodology, F.G.-C., C.C.-H., M.R.-S. and A.F.-C.; software, F.G.-C. and M.C.S.-R.; validation, M.C.S.-R., M.Q.-S., F.G.-C., C.C.-H., M.R.-S., A.F.-C., M.C.S.-R. and M.Q.-S.; investigation, F.G.-C., M.R.-S. and A.F.-C.; resources, M.R.-S. and A.F.-C.; data curation, M.C.S.-R., M.Q.-S., F.G.-C. and A.F.-C.; writing—original draft preparation, F.G.-C.; writing—review and editing, M.C.S.-R., M.Q.-S., F.G.-C., C.C.-H., M.R.-S. and A.F.-C.; visualization, M.C.S.-R., M.Q.-S. and F.G.-C.; supervision, F.G.-C., C.C.-H., M.R.-S. and A.F.-C.; project administration, A.F.-C.; funding acquisition, M.R.-S. and A.F.-C. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable for studies not involving humans or animals.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data are contained within the article.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Funding Statement

This research was mostly funded by the Research Provost’s Office of the University of Costa Rica, grant number C3106. The Technological Support Program for Industry (PATI, project number 917-02) and the School of Chemistry partially funded this research. The APC was funded by the Research and Teaching Provosts’ Offices, the School of Chemistry, and the Faculty of Microbiology.