1. IntroductionFlame retardants are chemicals added to different consumer products to inhibit possible combustion in the event of a fire, consequently reducing product flammability and improving product safety. To date, more than 175 flame retardants have been commercially produced, among which brominated flame retardants (BFR) are the most widely used due to being very effective and relatively cheap [1]. They are incorporated into electronics, textiles, plastics, and into building materials. Some of the products contain even up to 30% (w/w) of BFR compounds [2]. Mostly present as additives, they lack covalent bonding to the material and easily leach into the environment during usage, recycling, and disposal [3]. Their lipophilic character results in high affinity to adipose tissue; therefore, they bioaccumulate and biomagnify along the food chain and have been detected in numerous environmental and biological samples. They are also very persistent in the environment [1,2,3,4,5].The use of brominated flame retardants started in the 1970s [1]. In the past, the most widely used flame retardants were hexabromocyclododecane (HBCDD) and mixtures of polybrominated diphenyl ethers (PBDEs), commercially known as penta-, octa- and decaBDE mixture, named after the degree of bromination of the main ingredient [6]. PBDEs are diphenyl ethers with different degrees of bromination, ranging from 1 to 10, which results in 209 possible congeners. HBCDD also exists in different isomers, with alfa, beta, and gamma-HBCDD being the predominant diastereomers in the commercial technical mixture HBCDD [6]. However, research has shown that PBDEs and HBCDD cause adverse effects on humans and ecosystems, with endocrine disrupting potential, neurodevelopmental effects, and toxicity to reproductive systems being of the highest concern [7,8]. All these facts led to their inclusion as persistent organic pollutants (POPs) in the annexes of the Stockholm Convention from 2009 to 2017 [9]. To date, 185 countries, European Union (EU) countries included, have signed the agreement to restrict and eliminate the production and use of PBDEs and HBCDD [10].In general, the main exposure route to PBDEs and HBCDD is dietary intake [6]. Nevertheless, no regulations for the control of PBDEs and HBCDD in food and feed, such as MRLs, have yet been established by the EU authorities. However, a Commission Recommendation (2014/118/EU) on monitoring of traces of BFRs in food and feed was issued in 2014, specifying congeners of PBDEs of interest (BDE-28, 47, 49, 99, 100, 138, 153, 154, 183, and 209), recommending diastereomer specific determination of HBCDDs (α-, β-, and γ-) and setting different criteria on analytical methodology, including LOQ, which should be lower or equal to 0.01 µg/kg wet weight [11].Considerable research has already been performed in this area resulting in several published methods. The separation technique of choice for the analysis of PBDEs is gas chromatography (GC), which allows for the lowest detection limits. Mass spectrometry (MS), with its superior selectivity, is the most frequent detection technique in the determination of PBDEs. Authors report coupling GC with different types of mass analysers and utilising both electronic (EI) and chemical (CI) ionisation modes [12,13,14,15,16,17,18]. Nevertheless, some authors have demonstrated fitness for the purpose of high-performance liquid chromatography (LC) for analysis of PBDEs, using mass spectrometry with atmospheric pressure chemical ionisation (APCI) [19] or atmospheric pressure photoionisation (APPI) [20,21,22], while electrospray ionisation (ESI) has proven to be ineffective for analysis of PBDEs, owing to poor ionisation of nonpolar and aromatic compounds [19,20]. Thermal interconversions of HBCDDs occur at temperatures above 160 °C; therefore, for speciated determination of HBCDD isomers, which is a requisite stated in the 2014/118/EU, the only appropriate chromatographic technique is liquid chromatography [23]. Coupled with tandem mass spectrometry (MS/MS), it provides an instrument for the analysis of HBCDDs at sub-ppb levels, which is already well documented [6,15,18,24,25,26,27,28].Despite the substantial number of publications describing the methodology for the analysis of PBDEs in food and feed, only a handful of methods deal with the simultaneous analysis of HBCDDs and PBDEs [6,12,15,16,18,20,25,29,30,31,32,33,34]. Furthermore, only a few include BDE-209 [6,12,15,16,18,20,32,34], the main component of decaBDE commercial mixture, in the method for analysis of PBDEs. BDE-209 was the last PBDE put under regulation by the Stockholm Convention in 2017, and owing to its high boiling point and thermal instability, achieving the LOQ recommended by the 2014/118/EU is particularly challenging [2,6,29]. Authors reporting such low LOQs suggest using GC coupled to high-resolution mass spectrometry (HRMS) [12,13,14,16,18]. This sophisticated technique is undoubtedly fit for the purpose. However, the high cost of the instrument and the requirements of well-trained personal makes this instrument less available, especially in developing countries. An interesting alternative is tandem mass spectrometry (MS/MS), which is also highly selective and sensitive, while its price is substantially lower than HRMS and is less operationally demanding. There have been a few recent papers utilising atmospheric pressure chemical ionisation (APCI) GC-MS/MS for the analysis of PBDEs [35,36]. This soft ionisation technique makes quantitation at an ultra-trace level less challenging also for the PBDEs with higher bromination degrees, achieving very low quantification levels. However, this instrument is rarely found in laboratories, while GC-MS/MS with electron ionisation (EI) mode is much more common, owing to its robustness and versatility.To address the issue of a simple method using commonly available instrumentation, we have developed a method for simultaneous determination of all nine relevant PBDE congeners (BDE-28, 47, 49, 99, 100, 153, 154, 183, and 209) and HBCDD isomers (α-, β-, and γ-) in different types of food and feed, using single sample preparation method and dual detection, GC-MS/MS(EI) for PBDEs and LC-MS/MS(ESI) for HBCDDs. To the best of our knowledge, only one method with single sample preparation for both HBCDDs and PBDEs, including BDE-209, and using the alternative tandem mass spectrometry in EI mode for PBDE detection has been published, but the detection level for BDE-209 was not sufficient to meet the 2014/118/EU recommendation [6,11]. The performance of the developed method was evaluated with fortified samples of uncontaminated infant milk formula and further assessed using archived proficiency testing samples. 2. Materials and Methods 2.1. Reagents and SorbentsAll solvents, reagents, and chromatographic sorbents used were analytical grades or above and were of suitable purity for residual analysis. Dichloromethane, n-hexane, and toluene were purchased from Honeywell (Seelze, Germany). Methanol, acetone, acetonitrile, and anhydrous sodium sulfate (Na2SO4) were supplied by J.T. Baker (Deventer, the Netherlands). Potassium hydroxide (86%) was obtained from Fisher Chemicals (Pittsburgh, PA, USA). Silica gel (60 Å, 35–60 mesh) and Carbopack C (60/80 mesh) were purchased from Supelco (Bellefonte, PA, USA), sulfuric acid (96%) from Carlo Erba (Rodano, Milano, Italy) and BioBeads S-X3 (200–400 mesh) from Bio-rad (Munich, Germany). All chromatographic sorbents used (acidic silica, potassium silicate, silica and Na2SO4) were prepared as stated elsewhere [37]. 2.2. Standards

All standards used were 97.0% purity or higher.

Reference standard solutions of BDE-209 (50 µg/mL) and BDE-49 (50 µg/mL) were purchased from Cambridge Isotope Laboratories (Andover, MA, USA). Mixture of 13C12-labelled α-, β- and γ-HBCDD at 50 µg/mL and commercial mixtures of PBDEs at 1 ng/mL (10 ng/mL for BDE-209), containing 13C12-labelled (“Method 1614 labelled surrogate stock solution”) and native BDE-28, 47, 99, 100, 153, 154, 183, and BDE-209 (“Method 1614 native par stock solution”) as well as calibration solutions containing native (at 1, 5, 50, 500 and 2500 ng/mL) and 13C12-labelled analytes (at 100 ng/mL) (“Method 1614 calibration solutions CS1-CS5”) and standard mixture, containing 28 native PBDEs at 5–25 ng/mL and 12 13C12-labelled PBDEs at 100–500 ng/mL (“ROHS PBDE calibration solution CS2”), used for evaluating selectivity of the developed method, were also obtained from the same supplier. Analytical standards α-, β- and γ-HBCDD were supplied by Sigma-Aldrich (Steinheim, Germany).

Calibration solutions “Method 1614 calibration solutions CS1-CS5”, containing native and labeled standards (1–2500 ng/mL) were used for quantifying BDE-28, 47, 99, 100, 153, 154 and BDE-183. Calibration solutions for BDE-209 and BDE-49 were prepared separately at the same concentration levels by combining and diluting certified standards of BDE-209 or BDE-49 and Method 1614 labelled surrogate stock solution so that same concentration levels were achieved.

Individual stock solutions of α-, β- and γ-HBCDD at 1 g/L were prepared in methanol. Stock solutions and labelled HBCDDs stock solutions were combined and diluted in methanol to prepare standard calibration solutions at 0.2, 0.5, 1, 2, 5, 10, 20 and 80 ng/mL of native HBCDDs and 60 ng/mL of labelled HBCDDs. The vials containing calibration standards were carefully concentrated under a gentle stream of nitrogen to dryness and re-dissolved in methanol/water (75:25, v/v).

For fortification experiments, standard solutions at 10 and 100 ng/mL, containing native PBDEs and HBCDDs, were prepared by combining and diluting native PBDEs stock solution and BDE-49 reference standard and by combining and diluting individual stock solutions of α-, β- and γ-HBCDD. For fortification experiments of BDE-209 at 0.01 µg/kg, separate standard solution at 10 ng/mL was prepared by diluting reference standard BDE-209.

For instrumental method development, individual standard solutions of α-, β- and γ-HBCDD at 50 ng/mL and standard solution of BDE-209 at 50, 100, 500, and 5000 ng/mL were prepared by making appropriate dilutions of the stock solution.

All standards were stored in amber containers and kept at 4 °C.

2.3. Test Samples

During screening of different foodstuffs and feed, a sample of infant formula (ecological cow milk) with very low levels of contamination (<0.001 µg/kg) was traced. This sample enabled spiking at the target LOQ and was therefore used in our fortification experiments. For further assessment of the trueness of the proposed method and its applicability to different food and feed matrices, test materials from previous proficiency testing (PT) exercises organised by the European Union Reference Laboratory for halogenated persistent pollutants (EURL-POPs) in feed and food, were used. Performance of the method was assessed on three different food and feed matrices: baby food, composed of pork, pumpkin, and vegetable oil (BF-2101-BF), fish fillet (2001-Fl), and dried citrus pulp (2105-DCP).

2.4. Sample TreatmentThe extracts were prepared following the modified EPA 1614 standard method [37].

10 g of homogenised sample was weighted in a 50 mL PP centrifuge tube (or 100 mL Erlenmeyer flask in case of wet samples) and spiked with 1 ng of labelled PBDEs (10 ng of labelled BDE-209) and 12 ng of labelled HBCDDs. After 30 min equilibration time, a sufficient amount of anhydrous Na2SO4 was added to form a free-flowing powder. After 1 h drying time, 30 mL of dichloromethane:hexane (1:1, v/v) was added to the mixture. The mixture was vortexed for 1 min and ultrasonically extracted (UAE) for 1 h at 40 °C in an ultrasonic bath (Sonis 4, Iskra PIO, Šentjernej, Slovenia). After centrifugation at room temperature for 3 min at 2500 rpm, the supernatant was carefully decanted into a distillation bulb, and the extraction process was repeated.

The combined crude extract was purified using gel permeation and adsorption chromatography. First purification step was performed by using a multi-layer silica column packed with (bottom to top): Na2SO4 (2 g), neutral silica (1 g), potassium silicate (1.5 g), neutral silica (1 g), acidic silica (35% w/w, 25 g), neutral silica (1 g) and Na2SO4 (2 g). The combined extract, without prior concentrating step, was transferred to the top of the prepared column, and the analytes were eluted with additional 50 mL of dichloromethane:hexane (1:1, v/v). The eluate was concentrated using rotary evaporator to about 100 µL, quantitatively transferred to a vial and further concentrated to incipient dryness under a gentle nitrogen stream. The residue was re-dissolved in 300 µL dichloromethane:hexane (1:1, v/v) and put on an autosampler of an HPLC (1100/1200 Series, Agilent Technologies, Santa Clara, CA, USA) consisting of quaternary HPLC pump, degasser, fraction collector and two valves. The extract was injected into a GPC column (Bio Beads S-X3, 70 g) with a flow of dichloromethane:hexane (1:1, v/v) set at 2.5 mL/min. The compounds in the collection fraction of 64–100 mL were further isolated by fluxing through active carbon (Carbopack C, 200 mg). The analytes were eluted from carbon with additional 45 mL of hexane.

The purified extract was concentrated to about 100 µL, transferred to a vial equipped with an insert and further concentrated to dryness under a gentle stream of nitrogen, and the dry residue was dissolved in 20 µL of toluene.

The extracts were analysed for PBDEs with GC-MS/MS (EI). After the analysis, they were carefully concentrated to dryness, and the dry residue was re-dissolved in 200 µL of methanol:water (75:25, v/v) and analysed with LC-MS/MS (ESI) for HBCDD on the same day.

During validation of the method, blank infant milk formula was spiked at two different levels (at 0.01 µg/kg and 0.3 µg/kg) for PBDE 28–183 and HBCDDs and at three different levels for BDE-209 (at 0.01 µg/kg, 0.1 µg/kg, and 0.3 µg/kg). The fortification experiments on each level were performed in sextuplicate using 50 mL PP centrifuge tubes. To assess within-laboratory reproducibility, all experiments were repeated at least a week later by a different technician. The methods’ performance was further evaluated by analysing archived PT samples of fish fillet, dried citrus pulp, and composite baby food.

To optimise the extraction method, the sample of fish fillet was analysed in two modifications. First in six parallels, using PP centrifuge tubes and 20 g of anhydrous Na2SO4, and second by employing 100 mL Erlenmeyer flask, facilitating bigger amount of the desiccant (40 g).

2.5. Instrumental Analysis 2.5.1. GC-MS/MS (EI) Method

Separation and detection of PBDEs were carried out by a 7890B gas chromatograph (Agilent Technologies), equipped with a 7693 autosampler (Agilent Technologies) and coupled to a 7010B triple-quadrupole mass spectrometer (Agilent Technologies). In the optimised method, 5 µL of standards and extracts were injected using a multi-mode inlet (MMI) set on solvent vent mode. Vent flow and pressure were set at 100 mL/min and 5 psi, vent time was 0.25 min, and inlet was switched to purge mode at 2.62 min (with a purge flow of 60 mL/min). Inlet initial temperature was set at 70 °C (0.25 min), followed by a quick ramp of 600 °C/min to 325 °C. PBDEs were separated on a 20 m × 0.18 mm I.D. ZB-Semivolatiles (Phenomenex, Macclesfield, Cheshire, UK) analytical column with 0.18 µm film thickness and helium as the carrier gas at constant flow of 1.4 mL/min. GC oven temperature programme was as follows: 70 °C (1.25 min), 20 °C/min to 240 °C and 50 °C/min to 320 °C (23 min).

Mass spectrometer was operated in electron ionisation mode with ionisation energy set at 70 eV. Ion source, transfer line and both quadrupoles were maintained at 280 °C, 300 °C, and 150 °C, respectively. Helium (2.5 mL/min) was used as quench gas, and nitrogen (1.5 mL/min) as collision gas.

Quantitation of PBDEs was performed using five-point calibration curves (four-point for BDE-49) in multiple reaction monitoring (MRM) mode, whereas selected ion monitoring (SIM) was used for analysis of BDE-209. Two specific transitions from the same cluster were monitored for each congener BDE-28 to BDE-183 and its corresponding internal standard (IS)—one quantitative and the other confirmative, shown in Table 1. For BDE-209 and 13C12-BDE-209 three and two m/z were monitored, respectively.

Resolution of the quadrupoles was set to unit (0.7 Da). Masshunter (Agilent Technologies) was used to control the system and process the data.

2.5.2. LC-MS/MS(ESI)

Analyses of HBCDDs were performed on a Chronet Symbiosys Plus HPLC system (Axel Semrau, Sprockhövel, Germany) coupled to Sciex 6500+ triple quadrupole mass spectrometer (AB Sciex, Darmstadt, Germany). The mass spectrometer was operated in negative electrospray ionisation mode.

During method development, two analytical columns were tested: Phenomenex Luna PFP (150 × 2 mm, 3 µm particles) and Luna C18 (150 × 2 mm, 3 µm particles), held at 30 and 40 °C. Mobile phase flow was set at 0.2 mL/min. Analytes were separated using binary gradient system consisting of methanol:water (75:25, v/v) (A) and acetonitrile (B). Initial composition 80% A was held for 1.00 min, followed by linear decrease in the next 1.00 min to 45% A, held at 45% A for 9.0 min and returned to initial conditions in 0.1 min. Injection volume was 70 µL.

In the optimised method, temperature of the ion source was kept at 300 °C, ion spray voltage was −4500 V, whereas optimised curtain, nebulising, drying and collision gas were 35.0, 35.0, 50.0 and 10.0 psig N2, respectively.

For native HBCDDs, two transitions per parent m/z were monitored, while only one transition was monitored for their labelled analogues. Selected MRM transitions and optimised parameters are presented in Table 2. 2.6. Quality Control

Due to UV lability of PBDEs, all standards and extracts were kept in amber glass or wrapped in aluminium foil. To improve reliability of the analyses, isotopically labelled analogues were used for all the analytes except BDE-49.

To assure reliable identification, two transitions (three ions for BDE-209) were monitored for every analyte, and the deviation of the ratios between both transitions (ions for BDE-209) were within 20% of the value obtained in the reference standard, analysed in the same sequence.

To control possible contamination, instrumental and procedural blanks were performed in every batch. The procedural blank was processed through the entire analytical procedure along with the samples. Batch was accepted when procedural blanks were less than 20% of the LOQ or the contribution to the result was less than 5%. The results were not corrected for blanks.

In every batch, the instrument was either calibrated or the calibration was checked before, and in both cases after the injection of extracts, by at least one reference standard. The acceptance criteria were set at 20%.