Standards and reagents

A collection of 25 multiclass pharmaceuticals was chosen for the current investigation; Table S1 provides information on the target compounds, including formula, specific physicochemical parameters, and corresponding isotopically labelled internal standards (ILISs). All the selected pharmaceuticals were high-purity pharmaceutical standards (purity > 98%). Citalopram, venlafaxine, and fluoxetine hydrochloride were obtained from TCI (Zwijndrecht, Belgium). O-Desmethyl venlafaxine, norfluoxetine hydrochloride, erythromicin, carbamazepine, N-acetyl sulfamethoxazole, sulfadiazine, sulfamethazine, sulfamethizole, sulfamethoxazole, sulfamethoxypyridazine, sertraline hydrochloride, sulfapyridine, sulfaquionoxaline, sulfathiazole, caffeine, paracetamol, trimethoprim, carbamazepine, carbamazepine-10,11-epoxide, phenazone, oxytetracycline and oxolinic acid were purchased from Sigma-Aldrich (Darmstadt, Germany) and N-desmethyl sertraline hydrochloride was purchased from Santa Cruz Biotechnology (Santa Cruz, USA). Labelled internal standards carbamazepine-d₁₀ and fluoxetine-d₅ were obtained from A2S (Saint Jean d’Illac, France). The stock standard solutions were prepared in methanol, at concentrations of approximately 2,000 mg L^–1 and 1,000 mg L^–1 and were stored at −20°C. By suitably diluting the stock standard solutions in methanol/water 10/90 (v/v), fresh individual and blended composite working solutions were prepared at various concentrations. A distinct mixture of ILISs was prepared in methanol at a concentration of 20 mg L⁻¹, and thereafter diluted accordingly. Fisher Scientific (Leicester, UK) provided the methanol (MeOH) and water used as LC-MS grade solvents, while formic acid (FA) was purchased from Carlo Erba Reagents (Chaussée du Vexin, French) and hydrochloric acid (HCl) fuming 37% for analysis from Merck (Darmstadt, Germany).

Sample collection and preservation

Without any pharmaceutical contamination, seawater from an aquafarm in the Ionian Sea close to Ioannina was used for method development. Samples were collected and then transferred to the lab in a transportable freezer. The samples were vacuum-filtered with Millipore (0.45 μm) before analysis to eliminate any suspended solid material, and they were then stored at –20°C until extraction.

Sample extraction (SPE)

Through the use of a standard offline SPE connected to a vacuum pump, selected analytes were isolated and pre-concentrated from water samples. In order to optimize the extraction method, five different types of cartridges were tested, Oasis HLB (200 mg, 6 mL), Oasis MCX (150 mg, 6 mL), C8-CNWBond (500 mg, 6 mL), C18-CNW Bond Elut (500 mg, 6 mL) and Telos Neo PRP (polar-modified reversed phase) (200 mg, 6 mL). Another parameter tested was the pH value, pH 3 and pH 7. Water LC-MS and methanol LC-MS + 2% acetic acid 20:80 (v/v) were examined for the washing step, whereas methanol and methanol + 2% acetic acid were tested as elution solvents. Finally, the highest recoveries were obtained using Oasis HLB cartridges (200 mg, 6 mL), loading 250 mL of sample, washing with water LC-MS and elution with methanol at pH 3 adjusted with HCl 37% 1 M.

Oasis HLB cartridges were preconditioned with 5 mL of methanol and 5 mL of water LC-MS in the optimized method before extraction. Water samples (250 mL) with 5 mL of 5% Na₂EDTA and a pH 3 value adjusted with HCl 1 M were passed through the SPE cartridges at a flow rate of roughly 6 mL/min using a vacuum manifold that maintained a constant pressure before the cartridge dried out. Following the percolation of the entire sample, 5 mL of water LC-MS was used to wash the cartridges. Following a 20-min vacuum drying period to eliminate any remaining water, the analytes were eluted using 2 × 5 mL of methanol, by gravity, at 1 mL/min flow rate. At 40°C, methanol extracts were dried off with a light stream of nitrogen, reconstituted in a mixture of 500 μL water/methanol 90:10 (v/v) and stored at −20°C until chromatographic analysis (Figure S1).

LC-LIT/Orbitrap MS analysis

Accela UHPLC system (Thermo Fisher Scientific, Bremen, Germany) including Accela quaternary gradient UHPLC pump (model 1.05.0900) and Accela autosampler (model 2.1.1) was used for the LC separation. First, we tested different proportions of the reconstitution solvent, 50-50, 70-30, 80-20, and 90-10 H₂O:MeOH (LC-MS). The optimal solvent was determined to be H₂O:MeOH (90:10 v/v).

Water (A) and methanol (B), both containing 0.1% FA, made up the mobile phase. A reversed-phase Hypersil Gold C18 analytical column (Thermo Fisher Scientific, Göteborg, Sweden) measuring 100 mm in length, 2.1 mm in diameter, and 1.9 μm in particle size was used for the separation. The column was kept at 35°C for the duration of the analysis. The mobile phase gradient started with 95% mobile phase A and was maintained for 1 min. In the following minute, the amount of mobile phase A was reduced to 30%, reaching 0% in 5 min, and then stabilizing for an additional 2 min. Afterwards, the mobile phase was restored to 95% A and maintained for more than 3 min for re-equilibration. 5 μL was the injection volume that was set, the autosampler tray temperature was stabilized at 20°C and the mobile phase flow rate was stable at 250 μL throughout the analysis.

Thermo Fisher Scientific (Bremen, Germany), provided the hybrid LIT Orbitrap XL Fourier transform mass spectrometer, and an Ion Max Electrospray Ionization (ESI) probe was installed in the hybrid MS system’s linear ion trap. The following ESI settings were used in the positive ionization mode: tube lens, 90 V; sheath and auxiliary gas flow rates, 10 arbitrary units (au); spray voltage, 4 kV; capillary temperature, 320°C; mass range, 120–1,000 m/z; and automatic gain control (AGC) target, 4 × 10⁵. The resolution was 60,000 in full-scan (FS) MS mode, while in the data-dependent MS² mode (ddMS²), it was 15,000 and the normalized collision energy (NCE) was 30%. Xcalibur 2.2 software (Thermo Electron, San Jose) was used for instrument control and qualitative and quantitative evaluations.

Method validation

In compliance with the Document No. SANTE/12682/2019, method validation was performed on the optimized method in accordance with the actual guidelines [76]. The method’s accuracy, linearity, precision, limits of detection (LOD) and qualification, matrix effect (ME), and process efficiency (PE) were all assessed.

By fortifying the seawater samples at the start of the extraction and comparing the concentrations obtained after the extraction with the initial spiking level, accuracy—expressed as method recoveries—was assessed at three different concentration levels of 25 ng L⁻¹, 100 ng L⁻¹, and 250 ng L⁻¹ (n = 5). Recoveries were computed using the provided equation (Equation 1):

where C_spiked is the concentration for spiked seawater samples; C_reference is the concentration obtained for seawater extracts spiked after the whole process; and C_blank is the concentration of a seawater extract without any spike to confirm the absence of analytes.

Using linear least squares fit for the areas under the peak versus concentrations, a nine-point method calibration curve was constructed for each compound, allowing for the determination of the linearity of the optimized technique in a range from the limits of quantification (LOQ) to almost 100 times the LOQ. The relative standard deviation (%RSD_r) of each of the five repetitions on the same day was used to calculate intra-day precision, also known as repeatability. Five days in a row, the same analysis (n = 5) was performed to determine the inter-day precision (reproducibility), which was reported as %RSD_R. In a quantitative approach, the mean recoveries range from 70% to 120% with an RSD ≤ 20%, according to SANTE guidelines. If the mean recovery is not less than 30% or more than 140%, with an RSD ≤ 20%, mean recovery values outside of the range of 70–120% may be acceptable in exceptional circumstances.

Using the signal-to-noise ratio (S/N) of 3:1 and 10:1, respectively, the LOD and LOQ were calculated from the extracted ion chromatograms (XICs) at the lowest concentration obtained from the recovery test. To check for variations in the analytical signal, a non-spiked blank sample was injected each time quantification was performed. The values obtained from this sample were then deducted from the values obtained from the spiked samples. A five-point matrix-matched calibration curve in methanol/water extract 10/90 (v/v) was then created using blank seawater sample extracts and injected at the start of each sequence. Every five samples, a solvent was supplied to prevent partial blockage of the ion transfer capillary and a drop in signal intensities.

In order to assess ME, two calibration curves were created: one using seawater extract and the other using a solvent mixture. Both calibration curves were prepared within the same concentration range (2.5–1,000 ng L⁻¹), and a blank extract was used to represent the signal of the non-spiked matrix extract. Calculations were based on the following equation (Equation 2):

According to Scordo et al. [77] (2020), a positive matrix indicating signal enhancement is present when ME > 0, while a negative ME and signal suppression are indicated by values of ME < 0. However, ME% values in the range of ± 20% to ± 50% are regarded as low, medium, and strong, respectively. To reduce the impact of the matrix on the analytes, suitable techniques must be used for values more than ± 20% [78–81]. The following equation (Equation 3) was used to calculate the %PE in order to correlate the ME with the recovered recoveries.

As the PE values are close to 100% indicate that the optimized method is accurate and effective, while recoveries are also close to 100% and ME values are low [81, 82].

Solid phase extraction optimization

First, for the optimization of SPE, five different types of cartridges, Oasis HLB (200 mg, 6 mL), Oasis MCX (150 mg, 6 mL), C8-CN WBond (500 mg, 6 mL), C18-CNW Bond Elut (500 mg, 6 mL), Telos Neo PRP (200 mg, 6 mL) and two different pH values, pH 3 (adjusted with HCl 1 M) and pH 7 (no pH adjustment needed), were tested. Prior to the extraction a conditioning step took place, 5 mL of methanol (LC-MS) followed by 5 mL of water (LC-MS) were added to the cartridges. Then, 250 mL of the sample was loaded into the cartridges at a flow rate of 6 mL/min and washed with 5 mL of water (LC-MS). Afterwards, the cartridges were dried under vacuum for approximately 20 min, and the elution step was carried out using 2 × 5 mL methanol (LC-MS) [19, 52]. The final extracts were evaporated under a gentle stream of nitrogen (N₂) until dryness and reconstituted in 500 μL 90/10 water/methanol (LC-MS). In all cases, 5 mL of Na₂EDTA 5% solution was added to the samples. Prior to extraction, Na₂EDTA is added in order to complex with metal ions and prevent metal interferences [19, 67, 83–85]. Preliminary experiments took place at pH 7 for all cartridges and the two cartridges that gave better recoveries were further investigated at pH 3. Results showed that Oasis HLB (200 mg, 6 mL) gave recoveries from 10% to 180%, Oasis MCX (150 mg, 6 mL) 23.4–99.2% for 13 pharmaceuticals, C8-CNWBond (500 mg, 6 mL) gave recoveries for only 9 pharmaceuticals and values ranged between 4.0–59.4%, C18-CNW Bond Elut (500 mg, 6 mL) recoveries were found at values 4.3–147.0% for 15 pharmaceuticals and Telos neo PRP (200 mg, 6 mL) 10.8–138.2% at pH 7 (Figure 1). In accordance with the results analyzed above, we further examined Oasis HLB and Telos neo PRP at pH 3. Finally, better recoveries were obtained for Oasis HLB cartridges at pH 3, 65.0–110.2%, while recoveries for Telos neo PRP ranged between 11.0% and 122.7% (Figure 2).

Display full size

Figure 1. 

Recoveries obtained with five different cartridges (Oasis HLB, Oasis MCX, C8-CNW, C18-CNW, Telos neo PRP) at pH 7. PRP: polar-modified reversed phase

Display full size

Figure 2. 

Recoveries obtained with cartridges Oasis HLB and Telos neo PRP cartridges at pH 3. PRP: polar-modified reversed phase

Orbitrap optimization

In the current study, all selected pharmaceuticals were detected and quantified with FS mode in positive ionization mode as [M + H]⁺ adduct ions, along with precise mass spectra obtained at 60,000 FWHM as profile data. The calculated accurate mass, which should match the theoretical accurate mass within a mass tolerance of ± 10 ppm, and the retention time, which needs to match the analyte in the sample within a tolerance of ± 30 sec in the same analytical batch, were used to identify them. MS fragmentation (at least one fragment ion) was used for a more precise identification. First, a mix of standard solutions (50 μg L⁻¹) of the pharmaceuticals analyzed was injected directly into the mass analyzer in FS mode to find the precursor ion, the most abundant fragments and the necessary collision energy (Table 1). A preliminary test to determine the most suitable column was carried out by testing Thermo Scientific, diphenyl (50 mm × 2.1 mm; 2.6 μm) and hypersil gold C18 (100 mm × 2.1 mm; 1.9 μm) from Thermo Scientific. Peak shape and sensitivity were good in both cases, but a slightly better separation of the standards was achieved on the hypersil gold C18 column due to more specific interactions possible between the stationary phase and the analytes. In the present study, two different mobile phases were tested to optimize the separation and detection conditions. It is known that the addition of FA or ammonium formate to the mobile phase is a common practice in positive mode to improve ionization: (1) water and methanol both acidified with 0.1% FA and (2) water and methanol with 0.1% FA and 5 mM ammonium formate. Τhe first combination was selected because of its applicability and effectiveness in multi-residue methods covering a broad spectrum of ECs giving better signal for most of the analytes. Another parameter in the MS is the flow rate. 400 μL min⁻¹ and 250 μL min⁻¹ were tested and the optimal value was found to be 250 μL min⁻¹. Column temperature was also evaluated at 27°C and 35°C, with the latter being the best option. In addition, different solvent compositions (water/methanol 50:50 v/v, methanol/water 10:90 v/v, methanol/water, 10:90 v/v, methanol/water acidified with 0.1% FA, water/methanol 70:30 v/v, and methanol/water 20:80 v/v) were tested as reconstitution solvents for the final extract. Better and sharper peaks were obtained with 10:90 (v/v) methanol/water acidified with 0.1% FA. The high resolution allowed for the separation of peaks that were not fully separated in the chromatographic dimension. Regarding MS conditions, AGC target, tube lens, sheath gas, auxiliary gas and spray voltage were studied as a function of the obtained analytical signal. For AGC target, among the two tested values (4 × 10⁵ and 10⁶), 4 × 10⁵ was selected as the best. Tube lens values of 90 V and 110 V were tested, and 90 V significantly improved the signal. Sheath gas tested values were 38 and 10, with the second one found to be the best, while the auxiliary gas flow studied values of 15, 10 and 10 were selected. Spray voltage values of 3.6 kV and 4 kV were tested and the last one was found to be the most appropriate. As shown in Table 1, at a resolution of 60,000, the mass error was always less than 0.5 ppm. The fact that accurate mass measurements are not influenced by the complex matrix components proves that LTQ Orbitrap MS contributes to the separation of compounds among potential interferences.

Μethod validation

The final optimized parameters were as follows: Oasis HLB (200 mg, 6 mL) cartridges, conditioning of sorbent with methanol and water, both LC-MS analysis, wash with water and as elution solvent was used methanol. Then, 250 mL of seawater was filtered, followed by pH adjustment with HCl 37% 1 M at pH 3 and 5 mL of Na₂EDTA 5% was added to the sample. Elution extracts were evaporated under a gentle stream of nitrogen until dry and reconstituted with 90:10 water/methanol. The analytical method was validated for its identification and quantification capabilities. All validation performance characteristics are shown in Table 2.

Three spiking levels (25, 100, and 250 ng L^–1) were selected to conduct the recovery experiments. A nine-point matrix-matched calibration curve was used to calculate concentrations in the spiked samples compared with the initial spiking level. Recoveries in all cases, at the intermediate spiking level ranged from 64.8% for norfluoxetine to 115.4% for caffeine, while recoveries for the higher level were found to be between 62.4% for sulfamethizole to 118.8% for caffeine and for the lower level 61.6% for sulfamethizole and 98.2% for sulfamethoxazole. Method precision was calculated as the standard deviation ranging from 0.5% for caffeine and paracetamol to 1.9% for carbamazepine-10,11-epoxide and from 0.9% for sulfapyridine to 4.8% for carbamazepine-10,11-epoxide for intra-day and inter-day, respectively.

Linearity was checked using spiked seawater samples in order to construct a nine-point calibration curve, in a range from LOQ to 1000 ng L^–1. The developed method exhibited excellent linearity (r² > 0.991) for all compounds. The linear range and correlation coefficients for each compound are listed in Table 2.

Method precision was assessed in terms of repeatability (intra-day) and reproducibility (inter-day) at three spiking levels of 25, 100, and 250 ng L^–1 (Table 2). This was measured as the standard deviation of the recovery percentages of the spiked samples obtained on the same day (n = 5) and five different days. Low relative standard deviations are crucial because of sensitivity variations in the equipment, even within the same sequence, which may result in variable results.

LOD and quantification were found in nanograms per gram as described previously. LODs ranged from 0.3 ng L^–1 for venlafaxine to 9.8 ng L^–1 for oxolinic acid, while LOQs from 1.2 ng L^–1 for carbamazepine to 26.4 ng L^–1 for oxolinic acid. It is noteworthy that in high-resolution MS detectors, the chromatogram may not provide information about the background noise, making the calculation of the S/N difficult. When background noise is not present in the chromatogram, the limits are estimated by injecting decreasing concentrations corresponding to a peak signal value around 10⁴, as recommended elsewhere [33, 86].

Matrix effects

The ME was studied as a validation parameter to assess the potential interferences of the matrix components in the ionization. The residual matrix components contained in complex solid matrices typically interfere in the ionization, affect the analytical signal measurements and hence cause errors, leading to inaccurate results. The use of matrix-matched calibration is highly recommended and isotope standards constitute an additional way to minimize ME. Matrix compounds can be eluted at the same retention time as the target ones. A matrix-matched calibration curve from pristine samples was prepared and the slope was compared with the slope of a curve in a solvent mixture. As a result, the ME was calculated using Equation 2. As shown in Table 2 and Figure 3A ME values found to be low, between –18.8% for sulfadiazine to 18.1% for citalopram, indicating a slight ME (lower than ± 20%) [81]. The overall method efficiency was expressed as the PE, correlating the MEs with the recoveries obtained. PE values were calculated using Equation 3. Results revealed that the PE was over 60% for the majority of the compounds analyzed. The highest values were calculated for norfluoxetine, while the lowest were found for caffeine and O-desmethyl venlafaxine (Table 2 and Figure 3B).

Display full size

Figure 3. 

A. Matrix effects calculated from the slopes of the solvent and matrix-matched calibration curves for the pharmaceuticals studied; B. calculated process efficiency (PE%) for the selected compounds

Application to real samples

After the development, optimization, and validation of the SPE method for the determination of pharmaceutical compounds, their application to real seawater samples from points within a fish farm located in the Epirus region (Ionian Sea) followed. First, the physicochemical characteristics of the water samples were measured as soon as they were received in the laboratory. The physicochemical characteristics of the samples are listed in Table 3. According to the results listed in Table 4, positive detections were found in the seawater samples for some of the selected pharmaceutical compounds, such as oxytetracycline, sulfadiazine, caffeine, paracetamol, and trimethoprim. In more detail, the table also shows the concentrations for each month of the annual sampling.

The pharmaceuticals detected in seawater samples were mainly antibiotics and they presented the highest concentrations during winter months. Specifically, oxytetracycline given the highest concentration (43,817 ng L^–1) in seawater samples in February 2021, sulfadiazine was detected in September 2020 (145.5 ng L^–1), February 2021 (1,092 ng L^–1), April and May 2021 (411.3 ng L^–1 and 339.3 ng L^–1, respectively), while trimethoprim was found in a low concentration of approximately 32.6 ng L^–1 and 34.2 ng L^–1 during February and May 2021, respectively. Caffeine and paracetamol were detected at levels below the LOQ values.

However, the distribution of antimicrobial resistance related to antibiotics, needs to be a major priority, in order to preserve the aquaculture environment and guarantee seafood safety.

Challenges and recommendations for future work

Monitoring of pollutants in water can reveal details about the destiny of pharmaceuticals in the environment, such as the processes of transformation, partitioning, and transport, in addition to the goals of the monitoring approach. Therefore, additional research on the behavior, destiny, and toxicity of the parent chemicals and TPs is necessary for such inclusion. With the advancement and complementarity of non-target analysis, the incorporation of metabolites and degradation products into monitoring strategies is crucial and feasible, it would allow for the evaluation of their possible combined ecotoxicological consequences. The identification of priority substances would allow for the connection between the occurrence of contaminants and associated ecological responses through integrated monitoring of the environmental matrices carried out concurrently with the measurement of biomarkers of exposure and effects. It would be possible to create environmental quality standards and/or guidelines for medicines in seawater, sediments, and biota if the data regarding the ecological effects of pharmaceuticals present in the examined environment were to expand. Among the short-term steps that have been suggested to reduce the amount of pharmaceuticals that humans, animals, and ecosystems are exposed to include close monitoring of the chemicals’ effects on living things and their environmental presence. Finally, a trade-off must be made between suggested scientific methods and practical and financial limitations.