Vortex-homogenized matrix solid-phase dispersion coupled with gas chromatography – electron-capture negative-ion mass spectrometry to determine halogenated phenolic compounds in seafood

Abstract

This study developed the efficacy of a method that was able to rapidly determine three halogenated phenolic compounds (HPCs, i.e., pentabromophenol (PeBP), tetrachlorobisphenol A (TCBPA) and tetrabromobisphenol A (TBBPA)) in marketed seafood samples. The method employed modified vortex-homogenized matrix solid-phase dispersion (VH-MSPD) prior to the detection of the halogenated phenolic compounds by injection-port silylation gas chromatography – electron-capture negative-ion mass spectrometry (GC-ECNI-MS), which was set to operate in the selected ion monitoring (SIM) mode. Parameters that impact the efficiency of extraction for VH-MSPD were optimized through the Box–Behnken design method. Under optimal conditions, 0.5 g of freeze-dried seafood tissue was mixed with 0.5 g of anhydrous sodium sulfate, and the treated sample was dispersed with 1.0 g of Florisil via a vortex. The blend was transferred to a glass column containing 1.0 g of octadecyl bonded silica (C18), used as clean-up co-sorbent. The target analytes were then eluted with 12 mL of acetonitrile. The concentrated extract was derivatized in the GC injection-port through reaction with silylating reagent (BSTFA), and the identity and quantity of the target analytes present in the samples were determined by the GC-ECNI-MS system. The limits of quantitation (LOQs) were 0.2 to 0.5 ng g⁻¹ (dry weight). A preliminary result revealed that trace levels of halogenated phenolic compounds were detected in various kinds of marketed seafood samples.

---

1. Introduction

Halogenated phenolic compounds (HPCs) are widely employed as flame retardants, and are often added to materials made out of plastic in order to suppress or impede combustion process. However, recent research suggests that exposure to such phenolic compounds can be toxic in organisms and humans: including neurological toxicity, teratogenicity, thyroid toxicity, embryo toxicity, and carcinogenicity are just a few of the examples.^1–6 Moreover, because of their physical and chemical properties, HPCs, like other hydrophobic and persistent organic pollutants, have been detected in aquatic systems worldwide, such as in surface water, sediment, sewage/sludge, and other aquatic biota, irrespective of whether the samples were collected in the vicinity of manufacturing plants areas, e-waste recycling areas or non-manufacturing remote areas.^7–14 Even more worrisome, however, is the detection of HPCs in humans, such as urine, plasma, serum, breast milk, and even in placental tissue.^15–21 These studies indicate that humans have been exposed to HPCs via inhalation and, perhaps, ingestion of food, especially seafood.²² Although in the environment, concentrations of these compounds are assumed to be low enough to be negligible in their toxicity, the potential adverse effects for human from long-term exposure to HPCs, via food ingestion, is enough to prompt us to develop a simple, yet effective method of detecting these compounds in marketed seafood. Tetrachlorobisphenol A (TCBPA), and tetrabromobisphenol A (TBBPA) are probably the most widely used organic flame retardants worldwide, and as a consequence, they are widely distributed and detected in the environment.⁹ Pentabromophenol (PeBP) causes the most concern because it is also a brominated flame retardant commonly detected in aquatic environments, compared to other brominated phenolic compounds.^8,13,15,16 Thus, these three HPCs were chosen as the model target analytes for the method development and validation reported in this study.

Many attempts have been made to determine the presence of HPCs in aquatic biota samples taken from the environment and aquaculture, and the techniques have been reviewed by Dirtu et al.²³ and Papachlimitzou et al.²⁴ The common steps taken to extract HPCs from aquatic biota samples are solid–liquid extraction, Soxhlet extraction, ultrasonic-assisted extraction, and pressurized liquid extraction (PLE) followed by additional clean-up steps, usually based on SPE cartridge and/or glass column purification. These methods are time consuming and tend to generate an unnecessary amount of organic waste. In comparison, matrix solid-phase dispersion (MSPD), first reported by Barker and his group,²⁵ requires less time and none special equipment. MSPD is SPE-based, and involves dispersing the sample over a solid adsorbent before the target analytes within the samples are eluted with comparatively small amounts of the solvent. The primary advantage of MSPD is that extraction, clean-up, and filtration are all performed in one step, so as to not only minimize the volume of solvent needed, but also simplify the analytical procedure. Moreover, MSPD produces lipid-free extract, which rids us of the need to obtain special equipment and perform extra steps to dispose of the extract.²⁶ MSPD has been successfully applied to determine various types of micropollutants in aquatic biota, animal tissue, and food samples, as reviewed by Capriotti et al.^26,27

A frequently employed technique for detecting HPCs is gas chromatography – mass spectrometry (GC-MS) or liquid chromatography – tandem mass spectrometry (LC-MS/MS). GC chromatographic separation is commonly used due to its high separation ability, superior distinguishing power, and ease of access.^{8,9,13–16,18} The method can be improved, however, through derivatization, which increases the volatility of hydroxylated analytes and improves sensitivity. However, off-line derivatization often requires a large amount of labor and time. Injection-port derivatization (IPD, or known as on-line derivatization) has successfully developed as an alternative technique to determine the presence of a wide range of hydroxylated organic pollutants in various matrices, as extensively reviewed by Wang et al.²⁸ IPD has been shown to reduce the amount of solvent wasted, simplify the preparation of samples, and avoid the necessity for particularly hazardous reagents.

MSPD may often be used as a way to pretreat biota samples, but the technique needs to be fine-tuned for the types of the matrix used and concentrations of analytes. To minimize the manually using of a mortar and pestle, and to avoid the extra time to clean-up the device between the each sample, this paper demonstrates a modified VH-MSPD method coupled with injection-port silylation GC-ECNI-MS that we developed to study the spread of three selected HPCs in our food; in particular, the presence of HPCs in samples obtained from marketed seafood. The parameters affecting VH-MSPD were systematically investigated to optimize conditions. Trueness and precision of the developed method were evaluated, and the suitability of the technique for detecting the trace levels of the target analytes in marketed seafood samples was also demonstrated.

2. Experimental

2.1. Chemicals and reagents

Standards: PeBP, TCBPA and TBBPA (purities of all compounds >97%) were purchased from Sigma-Aldrich (St. Louis, MO, USA). ¹³C₆-hexachlorobenzene (used as an internal standard; purity > 99%) was obtained from Cambridge Isotope Laboratories (Andover, MA, USA). Stock solutions of each analytes (1.0 mg mL⁻¹) and the appropriate mixtures of analytes in order to obtain working standards and sample fortification were all prepared in ethyl acetate. All stock solutions and mixtures were stored in the dark at 4 °C. Silylating reagent: N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA) was purchased from Sigma-Aldrich. Activated magnesium silicate (Florisil, <200 mesh) and Supelclean ENVI-18 (C18) were from Sigma-Aldrich. Other chemicals and solvents were purchased in high purity grade from Mallinckrodt Baker (Phillipsburg, NJ, USA), Sigma-Aldrich, and Merck (Darmstadt, Germany), and were used without further purification. Deionized water was further purified using a Millipore water purification device (Billerica, MA, USA).

2.2. Sample collection

Three freshly killed fish (striped bass, tilapia and milk fish), ten shrimps, and two squids were purchased from a local fish market. One imported frozen cod fillet was purchased from a supermarket in Chung-Li city, Taiwan. The seafood samples were all stored inside a thermos-insulator box, and quickly transported to the lab for further treatment and analysis. At the laboratory, the samples were washed several times with Milli-Q water, produced by Millipore Elix® 10 RO system. The skin of the cod fillet, the scales of the striped bass, tilapia, and milk fish, and shells of the shrimps were all removed. All muscle tissues were cut into smaller pieces with unstandardized dimensions, and then homogenized via a commercial blender. Four aliquots of about 2 g homogenized tissue from each kind of seafood sample was freeze-dried by a Panchum CT-8500D freeze-dryer (Panchum Sci. Corp., Taiwan) operated at −30 °C and eight mbar for 48 h. The freeze-dried tissue was then ground into a fine powder for VH-MSPD.

2.3. Vortex-homogenized matrix solid-phase dispersion

The MSPD procedure employed within this study has been described previously,²⁹ but was performed with some modifications. A portion of freeze-dried powdered seafood sample (0.5 g) was mixed with 0.5 g of anhydrous sodium sulfate and dispersed with 1.0 g of Florisil using a vortex for 2 minutes (in previous studies, the mixing was done using a mortar and a pestle). This blend was transferred to a polypropylene SPE cartridge containing 1.0 g of octadecyl bonded silica (C18), which was packed at the bottom of the sample and used as clean-up co-sorbent. A frit was used as cover over the dispersed sample before the sample was slightly compressed. The target analytes were then eluted by gravity flow with 12 mL of acetonitrile (optimized, see Section 3.1.), and at the end of the elution, a small vacuum was applied. The eluent was promptly gathered together, and evaporated until dry under a mild flow of nitrogen. The residue was re-dissolved in a solution of ethyl acetate (45 μL) that contains an internal standard of 10 ng mL⁻¹ plus silylating reagent BSTFA (5 μL), total 50 μL, and the entire mixture was subjected to injection-port silylation GC-ECNI-MS analysis.

2.4. GC-ECNI-MS analysis

The GC-ECNI-MS analysis was performed using a Finnigan Focus gas chromatograph coupled directly to a Focus DSQ quadrupole mass spectrometer (Waltham, MA, USA). Operation was done in the selected ion monitoring (SIM) mode, which was under electron-capture negative ionization (ECNI) for quantitation. The injection volume was 1 μL. The injection-port silylation conditions were as follows: a split ratio of 1 : 5, an injector temperature of 230 °C, and a glass inlet liner with a plug of silanized glass wool, as described elsewhere.³⁰ Derivatization of phenolic compounds using BSTFA is a very quick procedure (15 s is enough to complete derivatization).²⁸ Therefore, in a high injection-port temperature, BSTFA and target analytes were easily volatilized to ensure the complete derivatization during retained by glass wool, and then the silylated derivatives were channeled into the GC column. Every 25–30 injections, the inlet liners would be replaced. After each extract injection, 1 μL of ethyl acetate would be injected to check that none interfering species would be carried over to interfere with the next sample's analysis. A DB-5MS capillary column (15 m × 0.25 mm i.d., 0.25 μm film; Agilent, Santa Clara, CA, USA) was employed for the separation. The GC temperature program conditions were as follows: a 80 °C hold for 1 min, followed by a temperature ramp of 30 °C min⁻¹ up to 300 °C, and holding the temperature at 300 °C for 1.7 min (total 10 min). Methane (99.995% purity) was used as the reagent gas with a flow rate of 2.5 mL min⁻¹. Ion source and transfer line temperature were kept at 230 °C and 280 °C, respectively. The electron energy was 70 eV and the emission current was 100 μA. The dwell time was 75 ms per ion, and the solvent delay was 3 min. Under ECNI conditions, intense silylated molecular ion clusters with significant bromine and chlorine isotope patterns were observed for PeBP and TCBPA, respectively. For TBBPA, only the intense Br⁻ ions (m/z 79 and 81) with weak [M − Br]⁻ ion (m/z 609) were observed under ECNI conditions.³¹ These ions were employed as the quantitation ions to obtain maximum detection sensitivity and selectivity in the SIM mode, as listed in Table 1.

Table 1 Detection characteristic, retention time, linearity, LOD, LOQ, precision and trueness

Analytes	Retention time (min)	Quantitation ions (m/z)	Linear range (ng g^−1)	r^2	Mandel test F-test^a	LOD^b (ng g^−1)	LOQ^b (ng g^−1)	Intra-day (n = 5)	Inter-day (n = 20)
a Fcritical = 10.13 (95% confidence level).b ng g^−1 dry weight.c Average spiked recovery (final spiked concentration of 10 ng g^−1, dry weight).d Relative standard deviation (% RSD) of spiked recovery.
^13C-HCB	4.78	288, 290, 292
PeBP	7.08	406, 408, 410	2–100	0.9995	−0.508	0.06	0.2	89^c (8)^d	91 (4)
TCBPA	7.94	400, 402, 404	2–100	0.9995	0.264	0.15	0.5	83 (9)	88 (5)
TBBPA	8.89	79, 81, 609	2–100	0.9997	1.552	0.09	0.3	91 (5)	94 (5)

---

2.5. Lipid contents in aquatic biota

The amount of lipid contained in each of the aquatic biota tissue sample was determined via microwave-assisted extraction (MAE), as described in a previous study.³² The weight of the freeze-dried seafood tissue (400 mg) was measured, and then the tissue was extracted using a microwave system (CEM Mars Xpress, Matthews, NC, USA) in 10 mL of dichloromethane at 20 watts for 10 min. Then, the extract was dried by filtering it through purified cotton and evaporated under a flow of nitrogen (50 °C). The dry residue was weighed and the lipid contents expressed as the fraction of lipid (%) of the dry tissue, allowing concentrations to be converted to a lipid weight basis.

2.6. Method evaluation

The method developed above was evaluated with a spiked “tilapia” sample (free of target analytes), which was prepared with the sample pretreatment method described in Section 2.3. The tilapia sample was spiked with standard mixture to give the final concentrations of 2, 10 and 20 ng g⁻¹ (d.w.). Then, the sample was mixed in a blender for 12 hours, and subsequently lyophilized for 48 hours. The spiked sample was analyzed via VH-MSPD and injection-port silylation GC-ECNI-MS analytical procedures to evaluate the method developed. In order to eliminate matrix effects and achieve good quantitative results, the quantitation of target analytes was calculated from five-point matrix-matched calibration curves, as indicated by the response factors. The curve covers a range from 2 to 100 ng g⁻¹ (i.e., 2, 5, 25, 50 and 100 ng g⁻¹), and each divided by the fixed concentration of internal standard, which had a value of 10 ng g⁻¹. The limits of detection and quantitation (LODs and LOQs) were defined at a signal/noise (S/N) ratio of 3 and 10, respectively. The LODs and LOQs were calculated based on VH-MSPD coupled with injection-port silylation GC-ECNI-MS analyses of the samples. Precision and trueness were evaluated by intra- and inter-day analyses as well as low- and high-level of the spiked tilapia samples. The value of intra-day precision was found by analyzing five fish samples spiked with the standard mixture on the same day (n = 5), while the value of inter-day precision was found by five replicates on four consecutive days (n = 20). Trueness was calculated as the average recovery of these spiked samples.

3. Results and discussion

3.1. Optimizations of VH-MSPD procedure

Challenges for determining trace levels of organic pollutants in biota samples manifest in mainly two ways: the difficulty of extracting trace-level of target analytes from the biota matrix, and the removal of interfering specie from the extract. One of the most attractive advantages of MSPD is that target analytes can be extracted from the sample, lipids removed, and other interfering species discarded all in a single step.²⁶ Some parameters that may affect the extraction and clean-up efficiency of the VH-MSPD procedure are as follows: the types and the amounts of dispersant, clean-up co-sorbents, and elution solution. In this study, the preliminary experiments were performed using various combinations of Florisil, C18, or aluminum oxide as the dispersant and/or the clean-up co-sorbent, and acetonitrile (ACN), dichloromethane or methanol was used as elution solvents. From the preliminary results, it can be observed that the highest abundances (hereafter referred to as “extraction efficiencies”) were obtained by using 0.5 g of spiked fish sample mixed with 0.5 g of anhydrous sodium sulfate, and dispersed with 1 g of Florisil, then transferred to a SPE cartridge containing 1 g of C18 used as clean-up co-sorbents. The three HPCs were eluated from the cartridge by ACN 10 mL (Fig. 1). When dichloromethane was used as the elution solvent, abundances were low. On the other hand, when ACN was replaced with methanol, a light yellow color appeared, perhaps as a result of colored polar interfering species co-elution. The intense peaks also appeared in the GC-MS chromatograms, and interfered with peaks quantitation of the target analytes. These results are in agreement with our previous study: Florisil and C18 were employed as the dispersant and clean-up co-sorbent, respectively, and ACN was used as the elution solvent, when hydroxylated salicylate and benzophenone-type UV filters have been successfully extracted and cleaned up from freeze-dried marketed fish samples by the MSPD procedure.²⁹ The light yellowish elute persisted, however, even when the combination in reverse was performed.

Fig. 1 Effects of combinations of various dispersants and clean-up co-sorbents on peak abundance obtained using the VH-MSPD technique. The target analytes were eluated by 10 mL of acetonitrile. Three replicate experiments were performed; the error bars represent standard deviations. Flor: Florisil; Alum: aluminum oxide; C18: octadecyl bonded silica.

In order to increase the efficiency of extraction and to evaluate the effects of extraction on VH-MSPD, three parameters were further optimized using a Box–Behnken design (BBD): the amount of Florisil (as dispersant; 0.5, 1.0 and 1.5 g), C18 (as clean-up co-sorbent; 0.5, 1.0 and 1.5 g), and ACN (as elution solvent; 8, 12 and 16 mL). Stat-Ease Design-Expert 8.0.6 software (Stat-Ease, Inc., Minneapolis, MN, USA) was performed to evaluate the design of the experiment and to analyze the data. Table S1† shows the design matrix, and displays the 15 randomized experiment and the related experimental responses (evidenced by total recovery). The significance of the experimental parameters was evaluated by analyzing the variance (ANOVA), which is a method that is acceptable for interpreting the data in a statistically significant manner, and provides valid values.^29,33 Table S2† summarizes the ANOVA results for total peak abundance. F- and p-values were calculated in order to qualitatively determine whether or not a factorial effective exists.³³ For the optimal experimental conditions, the factors for the amount of Florisil and C18 present were statistically significant at the 95% confidence level. On that note, interactions between any of the two factors examined were determined to be insignificant, since none of the interactions were found at the 95% confidence level. After the ANOVA test, the optimal experimental factors were generated from the Design-Expert software, and determined to be as follows: 1.0 g of dispersant (Florisil), 1.0 g of clean-up co-sorbents (C18), and 12 mL of ACN. Fig. 2 displays response surface plots for the total recovery of the target analytes, estimated according to the BBD for each pair of independent variables. In accordance, Fig. 2(a) shows the response surface obtained by plotting the amount of C18 versus the amount of Florisil, with the volume of ACN fixed at 12 mL. Fig. 2(b), on the other hand, shows the response surface developed for the volume of ACN and the amount of Florisil, with a fixed amount (1 g) of C18. Lastly, Fig. 2(c) displays the response surface obtained as a function of the volume of ACN and the amount of C18, for 1.0 g of Florisil. According to these three plots, the medium amounts of Florisil (i.e., 1.0 g), C18 (i.e., 1.0 g), and acetonitrile (i.e., 12 mL) provided the highest total recoveries for target analytes. Maximum spiked recovery, obtained by estimating under the optimized conditions, ranges from 82 to 97% with an average of 90 ± 5%. Altogether, these results demonstrate that the optimal conditions for extracting the three desired HPCs from 0.5 g of freeze-dried spiked fish sample (mixed with 0.5 g of anhydrous sodium sulfate), using the VH-MSPD method, are as follows: dispersing with 1.0 g of Florisil, transferring the blend to a SPE cartridge containing 1.0 g of C18 as the clean-up co-sorbent, and eluting the target analytes with 12 mL of acetonitrile. The extract obtained through this method was dried via evaporation, and the residue was then re-dissolved in 45 μL of ethyl acetate containing 10 ng mL⁻¹ of internal standard plus silylating reagent BSTFA (5 μL). Finally, the mixture (1 μL) was subjected to injection-port silylation GC-ECNI-MS analysis.

Fig. 2 Response surface plots for the total recovery of three target analytes estimated from the BBD on each pair of independent variables: (a) amount of C18 vs. amount of Florisil (volume of ACN fixed at 12 mL); (b) volume of acetonitrile vs. amount of Florisil (amount of C18 fixed on 1.0 g); (c) volume of acetonitrile vs. amount of C18 (amount of Florisil fixed at 1.0 g).

Instead of a mortar and a pestle, this experiment utilized a vortex-homogenized (VH) procedure. The efficiency of the VH method over the mortar and the pestle were compared using their extraction recovery and precisions (n = 5) under the best MSPD conditions. The average extraction recoveries by using the vortex was 90 ± 6%, and 88 ± 5% for using a mortar and pestle. The precisions of these two methods were evaluated by means of a t-test (with confidence intervals at 95%). The t-calculated value (1.53) was lower than the t-tabulated value (t_(95%,df=15) = 2.131), indicating that the precision for the vortex-homogenized was not significantly different from what can be obtained using a mortar and pestle. Taking into account the amount of time needed for the two procedures, however, the VH method is faster, in total, less than 2 min of vortex time was required to complete blending more than six samples simultaneously. Manually using a mortar and pestle required more than 4 min for each samples, not to mention the extra time needed to clean-up the device between the examining of each sample.

3.2. Method performance

The ability of optimized VH-MSPD coupled with injection-port silylation GC-ECNI-MS method to perform analysis, evaluated by its linear response range, reproducibility, LODs and LOQs, were performed so that the efficiency and feasibility of this method's use in the analysis of seafood samples could be presented in a scientific manner. Table 1 shows that excellent linearities (coefficient of determination (r²) greater than 0.9995) were found for the target analytes in the five-level matrix-matched calibration standard curves, and range from 2 to 100 ng g⁻¹. Mandel's fitting test was also done to confirm the linearity.²⁹ For Mandel's fitting test, it was proven that linear regression, rather than quadratic regression, fit the experimental data better, as evidenced by the fact that since the calculated F-values (ranging from −0.508 to 1.552) were lower than the tabulated limit (F_tab = 10.13) at the 95% confidence level (Table 1). Moreover, the LODs, defined at S/N of 3, were 0.06 to 0.15 ng g⁻¹; and the LOQs, defined at S/N of 10, were 0.2 to 0.5 ng g⁻¹ in 0.5 g (d.w.) seafood samples (Table 1).

Other than showing the linearities, Table 1 also serves to demonstrate that intra-day precision (% RSD) for these three HPCs was 9% or less, whereas the inter-day precision was 5% or less. Both intra- and inter-day trueness (average recoveries) ranged from 83% to 94%. Moreover, Table 2 shows the precision and recovery of both the low- and the high-level of the spiked samples ranged from 4 to 9%, and 78 to 91%, respectively. The high precision and good recovery of these data indicate that VH-MSPD coupled with injection-port silylation GC-ECNI-MS is capable of achieving not only a high degree of reproducibility, but also excellent linearity and sensitivity for determining the presence of the three HPCs in the seafood samples tested.

Table 2 Recovery and precision of low- and high-level of the spiked samples

Analytes	Low level (2 ng g^−1)	High level (20 ng g^−1)
a Average spiked recovery (%) (n = 5).b Relative standard deviation (% RSD, n = 5) of spiked recovery.
PeBP	78^a (6)^b	82 (4)
TCBPA	81 (7)	85 (5)
TBBPA	89 (9)	91 (7)

---

3.3. Method applications

In order to determine how feasible the above-mentioned, developed method is for practical use, more seafood samples were obtained. The types of seafood samples obtained were commonly consumed and commercially available, and their muscle tissue samples were pretreated according to the method detailed in Section 2.2. Table 3 lists the lipid content (lipid (%)) of each seafood sample, the yields of the extracted spiked samples (final spiked concentration: 10 ng g⁻¹ for each analyte) and the concentrations of the three HPCs detected in these marketed seafood samples. Fig. S1† displays the SIM chromatograms of the injection-port silylation GC-ECNI-MS for (a) non-spiked and (b) spiked shrimp samples. The yields of the spiked extractions of the three HPCs ranged from 82 to 107% with a RSD of less than 7%. To summarize (Table 3), the analysis shows that according to preliminary results, three of the target HPCs were detected in some, but not all, of the six seafood samples. The concentrations of the target HPCs that were found ranged from 0.3 to 2.2 ng g⁻¹ (dry weight). The results we detected for TBBPA (ranging from n.d. to 2.2 ng g⁻¹, d.w.), the most commonly studied HPCs, were present in higher concentration when compared to the results from fish samples collected in the River Po from Italy, as reported by Luigi et al.,¹⁴ and the fish samples collected in various English lakes, as reported by Harrad et al.³⁵ The TBBPA concentrations from the fish samples in the River Po ranged from n.d. to 0.8 ng g⁻¹ (d.w.); and the fish samples from the English lakes ranged from n.d. to 0.2 ng g⁻¹ (d.w.). However, the results we detected for TBBPA are lower than the results from fish muscle samples collected in Lake Chaohu in China, as reported by Yang et al.,³⁶ which ranged in concentrations from 3.5 to 48.5 ng g⁻¹ (d.w.).

Table 3 Concentration (ng g⁻¹, d.w.) of selected halogenated flame retardants detected in marketed seafood samples and their recoveries from spiked samples^a

Sample	PeBP	Analyte TCBPA	TBBPA
a n.d., not detected at LOQ, as listed in Table 1.b The mean lipid contents and the standard deviation (n = 3).c Original concentration (ng g^−1, dry weight) of analytes found in seafood sample (n = 1).d Spiked mean recovery (%, n = 3) at final concentration 10 ng g^−1 (d.w.) for each analyte.e Relative standard deviation (% RSD) are given in parentheses (n = 3).f ttab value is ±t(95%,df=6) = ± 2.44 for evaluation of the matrix effect at 95% confidence interval.
Striped bass (lipid (%) = 8.2 ± 0.5%)^b
Background Conc. (ng g^−1)^c	n.d.	n.d.	2.2
Spike recovery (%)	105^d (6)^e	89 (7)	85 (1)
Tilapia (lipid (%) = 12.4 ± 0.6%)
Background Conc. (ng g^−1)	n.d.	n.d.	n.d.
Spike recovery (%)	107 (2)	82 (5)	92 (4)
Cod (lipid (%) = 24.8 ± 0.5%)
Background Conc. (ng g^−1)	0.3	n.d.	n.d.
Spiked recovery (%)	103 (1)	89 (7)	99 (6)
Milk fish (lipid (%) = 12.1 ± 0.7%)
Background Conc. (ng g^−1)	n.d.	n.d.	0.4
Spiked recovery (%)	104 (7)	102 (2)	93 (2)
Shrimp (lipid (%) = 3.1 ± 0.4%)
Background Conc. (ng g^−1)	0.3	n.d.	0.7
Spiked recovery (%)	104 (2)	88 (5)	101 (1)
Squid (lipid (%) = 4.1 ± 0.4%)
Background Conc. (ng g^−1)	n.d.	1.8	n.d.
Spiked recovery (%)	85 (9)	89 (5)	94 (6)
tcal-Values of matrix effect^f
For striped bass	0.91	0.84	1.97
For shrimps	0.73	0.88	0.91
For squid	1.22	0.79	1.05

---

The matrix effect was examined using the “recovery function,” and the procedure was as follows: the standard mixture was first added to the final extraction from the samples of stripe bass, shrimp and squid (as described elsewhere³⁴), then the recovery functions were calculated. The calculation was done at five concentrations, and done by plotting “found concentrations” versus results obtained from the calibration curves. Then, the calculated slope and intercept of the recovery functions, done for all three of the HPCs, were compared with unity and zero by a t-test (with confidence intervals at 95%). For this study, the t-calculated values, ranging from 0.84 to 1.97 for stripe bass, 0.73 to 0.91 for shrimp, and 0.79 to 1.22 for squid, were all in the confidence interval of the t-tabulated value (±t_(95%,df=6) = ± 2.447) (Table 3). The t-values indicate that by adding the standard mixture to the final extract of seafood samples, the results were not significantly different from those obtained using calibration curves. Therefore, we can come to the conclusion that accurate quantitative determination of the three HPCs in various marketed seafood samples was achieved due to the absence of significant matrix effect.

In comparison with studies done previously, as listed in Table 4, PLE might provide high throughput capability, but solvent-consuming column purification steps were needed to cleanup interferences from biota samples. The values of LOD for our developed method are higher than that of detection by UPLC-MS/MS, but similar with those found by other LC-MS/MS methods. The recoveries of the spiked samples were not significantly different. Notice that less than 12 mL of the organic solvent was consumed in the VH-MSPD method that we have developed, and that the operation time required for analysis was completed in less than 30 minutes.

Table 4 Comparison of the developed method with previously published methods for fish samples^a

Extraction method	Detection	Recovery	LOD (ng g^−1, d.w.)	Ref.
a PLE: pressurized liquid extraction.
VH-MSPD	Online silylation-GC-ECNI-MS-SIM	85–107%	PePB: 0.06	This study
			TCBPA: 0.15
			TBBPA: 0.09
PLE	UPLC-MS/MS	75–112%	TCBPA: 0.005	14
			TBBPA: 0.011
PLE + acidified silica column cleanup	LC-MS/MS	85–99%	TBBPA: 0.09	35
PLE + C18-SPE cleanup	LC-MS/MS	Mean 98%	TBBPA: 0.04	36

---

4. Conclusions

This paper shows a modified VH-MSPD method coupled with injection-port silylation GC-ECNI-MS that we developed to study the spread of the three selected HPCs in our food through analysis of their presences in samples from marketed seafood. The parameters affecting VH-MSPD were systematically investigated to optimize conditions. It was demonstrated that this method is suitable for detecting trace levels of target analytes in marketed seafood samples, through good levels and ranges for precision, linear range, and detection limits in the sub-ng g⁻¹ level. Matrix-matched calibration standards were used to lessen matrix effects. This method has also been proven to be effective in analyzing and validating different sample matrices such as striped bass, tilapia, cod, squid and shrimp. Overall, VH-MSPD appears to be a reliable and sufficient technique for determining the presence of hydroxylated semi-volatile micropollutants in biota samples. Not only that, but it is also simple, and effectively prepares samples for analysis. Furthermore, it does not require the presence and use of dedicated specialty instruments, and only small amounts of organic solvents are needed. All of these factors make this method a viable candidate for use by monitoring programs in routine analysis.

Acknowledgements

This study was supported by a grant from the Ministry of Science and Technology of Taiwan under contract No. MOST 103-2113-M-008-005-MY2. We thank the Instrumental Center of National Central University for instrumentation support. We also thank Erica Ding, who checked over the English and grammar usage within this manuscript.

References

1. S. Kitamura, N. Jinno, S. Ohta, H. Kuroki and N. Fujimoto, Biochem. Biophys. Res. Commun., 2002, 293, 554–559.
2. J. Y. Qu, H. H. Shi, Q. P. Liu and J. F. Shen, Acta Sci. Circumstantiae, 2008, 28, 1625–1630.
3. H. Sun, O. X. Shen, X. R. Wang, L. Zhou, S. Q. Zhen and X. D. Chen, Toxicol. In Vitro, 2009, 23, 950–954.
4. A. Nakajima, D. Saigusa, N. Tetsu, T. Yamakuni, Y. Tomioka and T. Hishinuma, Toxicol. Lett., 2009, 189, 78–83.
5. M. Montaño, A. C. Gutleb and A. T. J. Murk, Environ. Sci. Technol., 2013, 47, 6071–6081.
6. M. Y. Song, D. Liang, Y. Liang, M. J. Chen, F. B. Wang, H. L. Wang and G. B. Jiang, Chemosphere, 2014, 112, 275–281.
7. S. G. Chu, G. D. Haffner and R. J. Letcher, J. Chromatogr. A, 2005, 1097, 25–32.
8. M. Polo, M. Llompart, C. Garcia-Jares, G. Gomez-Noya, M. H. Bollain and R. Cela, J. Chromatogr. A, 2006, 1124, 11–21.
9. C. Sanchez-Brunete, E. Miguel and J. L. Tadeo, J. Chromatogr. A, 2009, 1216, 5497–5503.
10. Y. J. Yang, L. B. Lu, J. Zhang, Y. Yang, Y. N. Wu and B. Shao, J. Chromatogr. A, 2014, 1328, 26–34.
11. S. J. Song, M. Y. Song, L. Z. Zeng, T. Wang, R. Z. Liu, T. Ruan and G. B. Jiang, Environ. Pollut., 2014, 186, 14–19.
12. L. B. Lu, Y. J. Yang, J. Zhang and B. Shao, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci., 2014, 953–954, 80–85.
13. J. K. Xiong, T. C. An, C. S. Zhang and G. Y. Li, Environ. Geochem. Health, 2015, 37, 457–473.
14. V. Luigi, M. Giuseppe and R. Claudio, Environ. Sci. Pollut. Res., 2015, 22, 14050–14066.
15. C. Thomsen, K. Janák, E. Lundanes and G. Becherc, J. Chromatogr. B: Biomed. Sci. Appl., 2001, 750, 1–11.
16. C. Thomsen, K. Janák, E. Lundanes and G. Becherc, J. Anal. Toxicol., 2002, 26, 129–137.
17. S. G. Chu and R. J. Letcher, J. Chromatogr. A, 2013, 1320, 111–117.
18. Z. X. Shi, Y. Jiao, Y. Hu, Z. W. Sun, X. Q. Zhou, J. F. Feng, J. G. Li and Y. N. Wu, Sci. Total Environ., 2013, 452–453, 10–18.
19. F. Vela-Soria, O. Ballesteros, A. Zafra-Gómez, L. Ballesteros and A. Navalón, Anal. Bioanal. Chem., 2014, 406, 3773–3785.
20. Y. J. Yang, J. Guan, J. Yin, B. Shao and H. Li, Chemosphere, 2014, 112, 481–486.
21. F. Vela-Soria, O. Ballesteros, F. J. Camino-Sánchez, A. Zafra-Gómez, L. Ballesteros and A. Navalón, Microchem. J., 2015, 118, 32–39.
22. R. Cruz, S. C. Cunha and S. Casal, Environ. Int., 2015, 77, 116–131.
23. A. C. Dirtu, M. Abdallah and A. Covaci, Trends Anal. Chem., 2013, 43, 189–203.
24. A. Papachlimitzou, J. L. Barber, S. Losada, P. Bersuder and R. J. Law, J. Chromatogr. A, 2012, 1219, 15–28.
25. S. A. Barker, A. R. Long and C. R. Short, J. Chromatogr. A, 1989, 475, 353–361.
26. A. L. Capriotti, C. Cavaliere, P. Giansanti, R. Gubbiotti, R. Samperi and A. Laganà, J. Chromatogr. A, 2010, 1217, 2521–2532.
27. A. L. Capriotti, C. Cavaliere, P. Foglia, R. Samperi, S. Stampachiacchiere, S. Ventura and A. Laganà, Trends Anal. Chem., 2015, 71, 186–193.
28. Q. Wang, L. Ma, C. R. Yin and X. Li, J. Chromatogr. A, 2013, 1296, 25–35.
29. D. Y. Tsai, C. L. Chen and W. H. Ding, Food Chem., 2014, 154, 211–216.
30. J. A. Field, D. J. Miller, T. M. Field, S. B. Hawthorne and W. Giger, Anal. Chem., 1992, 64, 3161–3167.
31. A. Covaci, S. Voorspoels and J. de Boer, Environ. Int., 2003, 29, 735–756.
32. M. H. Dévier, P. Labadie, A. Togola and H. Budzinski, Anal. Chim. Acta, 2010, 657, 28–35.
33. C. Stalikas, Y. Fiamegos, V. Sakkas and T. Albanis, J. Chromatogr. A, 2009, 1216, 175–189.
34. C. L. Chen, D. Y. Tsai and W. H. Ding, Food Chem., 2014, 164, 286–292.
35. S. Harrad, M. A. Abdallah, N. L. Rose, S. D. Turner and T. A. Davidson, Environ. Sci. Technol., 2009, 43, 9077–9083.
36. S. W. Yang, S. R. Wang, F. C. Wu, Z. U. Yan and H. L. Liu, Environ. Sci. Pollut. Res., 2012, 19, 4090–4096.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra20680h

---