Journal of Chromatography B 1099 (2018) 64–72

Contents lists available at ScienceDirect

Journal of Chromatography B
journal homepage: www.elsevier.com/locate/jchromb

Determination of polybrominated diphenyl ethers and novel brominated T

flame retardants in human serum by gas chromatography-atmospheric
pressure chemical ionization-tandem mass spectrometry
⁎
Yuwei Wanga, Yanmin Suna, Tian Chena, Zhixiong Shia, , Xianqing Zhoua, Zhiwei Suna,
⁎
Lei Zhangb, Jingguang Lib,
a
School of Public Health and Beijing Key Laboratory of Environmental Toxicology, Capital Medical University, Beijing 100069, China
b
The Key Laboratory of Food Safety Risk Assessment, Ministry of Health, and China National Center for Food Safety Risk Assessment, Beijing 100021, China

A R T I C LE I N FO A B S T R A C T

Keywords: The accurate detection of brominated flame retardants (BFRs) in humans is an area of high scientific interest and
GC-MS regulatory need due to their potential toxicity. The instrumental analysis of BFRs was commonly performed on
Atmospheric pressure chemical ionization gas chromatography-mass spectrometry (GC-MS) operating in electron ionization (EI) or negative chemical io-
Brominated flame retardants nization (NCI) modes. However, soft ionization techniques, such as atmospheric pressure chemical ionization
Polybrominated diphenyl ethers
(APCI), may be more suitable for the analysis of BFRs because the BFRs show high fragmentation in EI and low
Decabromodiphenyl ethane
selectivity in NCI. Additionally, accurate quantifications of BFRs in complex matrices is challenging due to their
Serum
low concentrations and therefore, a highly sensitive technique is desperately needed. In this study, a new
methodology based on gas chromatography-atmospheric pressure chemical ionization-tandem mass spectro-
metry (GC-APCI-MS/MS) analysis was developed for the determination of thirteen BFRs (eight usually monitored
polybrominated diphenyl ethers (PBDEs) congeners and five additional novel BFRs) in human serum. The pri-
mary task was to evaluate the potential of the GC-APCI-MS/MS technique for the trace analysis of BFRs in human
serum. The results of the spiked recovery test using fetal bovine serum showed that mean recoveries of the
analytes ranged from 83.4% to 118% with reduced swing differential signaling (RSDs) of ≤21.1%. The meth-
odological limits of detection (mLOD) of the analytes ranged from 0.04 to 30 pg/mL, and these values were at
least one order of magnitude lower than those estimated by the authors in a previous study using GC-NCI-MS or
GC-EI-MS/MS, indicating that GC-APCI-MS/MS is more sensitive. Specially, compared to GC-NCI-MS and GC-EI-
MS/MS, when GC-APCI-MS/MS was used for the detection of highly brominated BFRs, such as BDE-209 and
decabromodiphenyl ethane (DBDPE), a notable improvement in sensitivity and reliability was obtained using a
deactivated capillary column connected to the analytical column as the transfer line and maintaining a high
temperature to improve the chromatographic behaviors. The developed methodology was successfully used for
the analysis of BFRs in human serum collected from residents living in a BFR production area and Beijing.

1. Introduction
Brominated flame retardants (BFRs) are widely used in a variety of
products including electronic equipment, textiles, foams and building
materials for fire prevention. Polybrominated diphenyl ethers (PBDEs)
are a class of the most important BFRs that have been produced and
used since 1970s. PBDEs are additive BFRs and can therefore leach or
volatilize from products and enter the environment [1,2]. There are
three types of commercial PBDE products: Penta-BDE (mainly com-
posed of BDE-47 and BDE-99 and with some tri- to hepta-BDEs), Octa-
BDE (mainly composed of hexa- to deca-BDEs), and Deca-BDEs
(92–97% BDE-209, plus tiny amounts of nona- and octa-BDEs) [3].
Penta-BDE and Octa-BDE were added to the list of persistent organic
pollutants (POPs) under the Stockholm Convention in 2009 and were
therefore globally phased out [4]. Deca-BDE was also listed as a POP in
2017. However, the production and application of Deca-BDE continues
in China to date. With the regulation of PBDEs, the production and
application of a series of novel BFRs (NBFRs), including deca-
bromodiphenyl ethane (DBDPE), 1,2-bis(2,4,6-tribromophenoxy)-
ethane (BTBPE), pentabromoethylbenzene (PBEB), hexabromobenzene
(HBB) and pentabromotoluene (PBT), have risen sharply [5,6]. Deca-
bromodiphenyl ethane (DBDPE) has been produced only since 2005 in

⁎
Corresponding authors.
E-mail addresses: szx0127@ccmu.edu.cn (Z. Shi), lijg@cfsa.net.cn (J. Li).

https://doi.org/10.1016/j.jchromb.2018.09.015
Received 15 May 2018; Received in revised form 7 September 2018; Accepted 11 September 2018
Available online 12 September 2018
1570-0232/ © 2018 Elsevier B.V. All rights reserved.
Y. Wang et al.
China and is regarded as a deca-BDE replacement. Its production in-
creases 80% per year [5], and it has become the most popular NBFR in
China [6]. BTBPE is an octa-BDE replacement, and the worldwide
production/usage of BTBPE was estimated to be 16,710 tons in 2001
[7]. Pentabromotoluene (PBT), pentabromoethylbenzene (PBEB) and
hexabromobenzene (HBB) belong to polybromobenzenes that have one
phenyl ring with several bromine atom substituents. These three NBFRs
are all used as additive FRs. PBT is normally added to unsaturated
polyesters, and PBEB is mainly used in thermoset polyester resins. HBB
is primarily added to paper and wood and is mainly used in Japan and
China. Although PBT, PBEB and HBB are presently produced and used
at lower volumes, they have higher vapor pressures than do PBDEs and
DBDPE. Therefore, they are more likely to evaporate into the sur-
rounding environment [5,8]. Since these several NBFRs are all currently
used in China, they are included in the present study.
BFRs are ubiquitous in various matrices in China [2,6], which in-
dicates that development of an accurate and sensitive analytical tech-
nique for BFR detection is not only of scientific interest but is also a
regulatory need. Nevertheless, accurate quantification of BFRs in
complex matrices (serum, breast milk, etc.) is quite difficult due to their
low concentrations [5,9]. For sample treatment, interfering substances
(lipids, proteins, etc.) present in complex matrices require extensive
clean-up procedures. In some studies, multistage liquid-liquid extrac-
tion and subsequent sophisticated purification step were used [10,11].
In our previous studies, a QuEChERS (quick, easy, cheap, effective,
rugged and safe) approach was introduced for BFRs analysis and proved
to be a simple, fast and efficient sample treatment technique [12,13].
For instrumental analysis, the most frequently used detection techni-
ques of NBFRs and PBDEs were developed on GC-NCI-MS or GC-EI-MS/
MS [11,14]. However, EI-MS/MS has proven to be unsuitable for the
analysis of highly brominated BFRs because of its low sensitivity [13].
NCI-MS is more sensitive for bromine than is EI. Nevertheless, tandem
mass spectrometry shows much better selectivity, which is especially
important when analyzing complex matrices. Atmospheric pressure
chemical ionization (APCI) is an alternative soft ionization technique
which was designed to compensate for the sensitivity limitation of the
EI source. In APGC, minimal fragmentation of the molecular ion can be
obtained and therefore results in higher signal intensity compared to EI
ionization [15]. More importantly, a combination of APCI and tandem
mass spectrometry shows higher selectivity than NCI-MS. Gas chro-
matography-atmospheric pressure chemical ionization-tandem mass
spectrometry (GC-APCI-MS/MS, sometimes written as APGC-MS/MS)
has been successfully used to detect POPs, such as dioxins, furans,
pesticides and polyaromatic hydrocarbons, in various matrices [16–18].
In the present study, a QuEChERS approach coupled to GC-APCI-MS/
MS was developed and optimized for the simultaneous pretreatment
and detection of eight PBDE congeners and five currently used NBFRs in
human serum. Additionally, this study is part of our research project
entitled, “The association between BFR exposure and female cancer”,
and the developed methodology was then applied to the analysis of two
groups of female serum samples.
2. Materials and methods
2.1. Chemicals and reagents
HPLC-grade organic solvent (n-hexane, acetone, etc.) was purchased
from Dikma (Lake Forest, CA, USA) or Merck (Darmstadt, Germany).
Formic acid (96%) was purchased from Dikma. Anhydrous magnesium
sulfate (MgSO4) and sodium chloride (NaCl) were purchased from
Beijing Chemical Factory (Beijing, China). Primary secondary amine
(PSA) and octadecyl-modified silica (C18) were obtained from Agilent
Technologies (Palo Alto, CA, USA). The individual PBDE standards,
BDE 28, 47, 99, 100, 153, 154, 183, and 209, and the individual NBFRs
standards, 1,2-bis(2,4,6-tribromophenoxy)-ethane (BTBPE), pentabro-
moethylbenzene (PBEB), hexabromobenzene (HBB), PBT and DBDPE,
65
Journal of Chromatography B 1099 (2018) 64–72
were all provided by AccuStandard Inc. (New Haven, CT, USA). The
13
C-labeled internal standards, including 13C12-BDE-28, 47, 99, 153,
154, 183, and 209, and 13C6-BTBPE and 13C14-DBDPE, were all pro-
vided by Wellington Laboratories (Guelph, Ontario, Canada).
The standard reference material (organic contaminants in fortified
human serum) SRM 1958 was purchased from the National Institute of
Standards and Technology (Gaithersburg, MD, USA). Fetal bovine
serum was purchased from Tianhang Biotechnology (Hangzhou, China)
for method development and validation. Two groups of human serum
samples were collected. One group is from nonoccupational volunteers
(30 female donors) who lived in Weifang City, Shandong Province of
China. Weifang city is the largest BFR production base in China at
present. For comparison, another 30 female donors were recruited from
Beijing, the capital of China. Serum sample was collected when the
donors participated in routine physical examinations at a local hospital.
After fasting for at least 12 h, approximately 10 mL of blood was col-
lected with a BD Vacutainer serum tube (Franklin Lakes, NJ, US) by
medical professionals and then processed within 2 h to isolate serum by
centrifugation at 3000 rpm for 15 min. The serum was then kept at
−80 °C until analysis. This study was launched with the authorization
of the Ethics Committee of Capital Medical University, and all partici-
pants were clearly informed of the objective and gave their informed
consent before sample collection.
2.2. Sample treatment
Internal standards, including 5 ng each of 13C12-BDE-209 and 13C14-
DBDPE, and 0.5 ng each of 13C12-BDE-28, 47, 99, 154, 153, 183 and
13
C6-BTBPE, were introduced into a 5 mL polypropylene (PP) tube. For
the matrix spiking test, fetal bovine serum was used as the matrix and
was fortified with an appropriate volume of standard solution. Solvent
of the standard solution (n-hexane) was completely dried under ni-
trogen. Subsequently, 0.5 mL each of serum and water were added and
mixed by vortexing. After that, 2 mL of acetone/hexane (1:1, v/v),
400 mg of anhydrous MgSO4 and 100 mg of NaCl were loaded, and the
PP tube was shaken vigorously for 1 min, followed by centrifugation at
4000 rpm for 5 min. The upper layer was transferred into a new PP tube
containing 50 mg of MgSO4 and 100 mg of C18 and then immediately
shaken vigorously for 1 min for the removal of lipid and other inter-
fering substances. The tube was then centrifuged at 10,000 rpm (4 °C)
for 5 min. The upper layer was then transferred, concentrated to dry-
ness, and reconstituted in 100 μL of n-hexane for GC-APCI-MS/MS
analysis.
2.3. GC-APCI-MS/MS analysis
Instrumental analysis was performed on a triple-quadrupole mass
spectrometer equipped with an APCI source (Xevo TQ-S, Waters
Corporation, Milford, MA, USA) and coupled to a gas chromatograph
(7890B, Agilent Technologies, Santa Clara, CA, USA). A DB-5MS ca-
pillary column (15 m × 0.25 mm, 0.10 μm film thickness, J&W
Scientific, Folsom, CA, USA) was used for the chromatographic se-
paration with helium as a carrier gas (3 mL/min). An Agilent UM tubing
methyl deactivated capillary column (approximately 0.4 m, 0.25 mm id,
Agilent Technologies, Santa Clara, CA), which was installed in the
transfer line between the GC and the ion source, was connected to the
end of the DB-5MS column by a straight two-way valve, and the transfer
line temperature was maintained at 390 °C. The oven temperature of
the GC was set as: initial 100 °C hold for 1 min, 30 °C/min to 310 °C and
held at 310 °C for 10 min. The GC injector was performed in pulsed
spitless mode and the pulsed pressure was maintained at 50 psi for
1 min. The injection volume was 1 μL with the injector temperature was
280 °C.
On the Xevo TQS MS, the ion source was carried out in APCI+ mode
and run under dry conditions to promote charge transfer ionization.
Nitrogen was used as an auxiliary gas and cone gas and maintained at
Y. Wang et al. Journal of Chromatography B 1099 (2018) 64–72

Table 1 (MTBE), n-hexane, acetone/n-hexane (1:1,v/v), n-hexane/MTBE

Instrumental parameters of the optimized GC-APCI-MS/MS for PBDE congeners (1:1,v/v) and ethyl acetate, were evaluated for their efficiency of ex-
and NBFRs under charge transfer mode. traction. However, when MTBE, ethyl acetate and n-hexane/MTBE
Analyte tR (min) Precursor ion Product ion Collision Loss were used as the extractor, we observed that the organic solvent could
(m/z) (m/z) energy (eV) not be separated from the serum during serum pretreatment, even if
high speed centrifugation was conducted for the mixture (10,000 rpm
BDE-28 5.21 406 246a 30 -Br2
for 10 min). When n-hexane/acetone, n-hexane and acetonitrile were
406 139b 30 -CBr3O
BDE-47 5.90 486 326 30 -Br2 used for extraction, the organic solvents and the aqueous phase could
486 219 30 -CBr3O easily be separated. Nevertheless, low recoveries (< 60%) of the ana-
BDE-100 6.41 563 403 30 -Br2 lytes were obtained when n-hexane was used. In a further study,
563 297 30 -CBr3O
comparison by visual observation between acetonitrile and n-hexane/
BDE-99 6.55 563 403 30 -Br2
563 297 30 -CBr3O
acetone showed significant differences of these two extractors. The use
BDE-154 6.94 643 483 30 -Br2 of acetonitrile resulted in more colored and turbid extracts than that of
643 375 30 -CBr3O n-hexane/acetone; nonetheless, satisfactory recoveries (80% to 120%)
BDE-153 7.13 643 483 30 -Br2 were obtained for most analytes by using both acetonitrile and n-
643 375 30 -CBr3O
hexane/acetone. Finally, n-hexane/acetone was chosen as an extractor
BDE-183 7.67 719 452 50 -CBr3O
719 402 50 -Br4 because (I) the transfer of BFRs from serum into n-hexane/acetone
BDE-209 10.49 959 799 30 -Br2 (1:1,v/v) is a fairly simple task because BFRs are hydrophobic, and (II)
959 639 30 -Br4 n-hexane:acetone is easier to evaporate by nitrogen in the final step.
PBT 5.24 485 406 25 -Br
The second step of the QuEChERS approach was to remove the
487 406 25 -Br
HBB 5.75 549 470 30 -Br
coextracted interfering substances by a proper sorbent. PSA and C18
551 472 30 -Br were the two most frequently used sorbents and were compared in our
PBEB 5.37 500 485 25 -CH3 study. PSA has a WAX function and can remove polar compounds such
501 420 25 -Br as sugars and fatty acids from a nonpolar matrix [20]. However, PSA is
BTBPE 7.84 687 358 10 -C6H2Br3O
not efficient for lipid removal, and therefore C18 was normally used to
687 279 10 -C6H2OBr4
DBDPE 11.57 971 468 30 -C8H4Br5 improve the clean-up efficiency in samples with ≥2% fat [21]. The
971 485 30 -C7H2Br5 efficiency of C18, PSA and their mixture was evaluated by a recovery
test, which was conducted by using blank samples (fetal bovine serum)
a
Quantification ion. spiked at 0.1 ng/mL (1 ng/mL for BDE-209 and DBDPE). 100 mg each of
b
Qualifier ion. PSA and C18, and a mixture of PSA/C18 (50 mg each) were used and
compared. The results of the recovery test are shown in Fig. 1. The
250 L/h and 150 L/h, respectively. Argon was applied as the collision mean recoveries of most of the analytes were nearly 100% when C18
gas with a constant flow of 0.25 mL/min. The APCI corona pin current was used. However, the mean recoveries of PBT and DBDPE were
was operated in the constant current mode at 3 μA. Cone voltage was both < 65% with RSD > 15% when PSA was used. By contrast, the
set at 30 V for all the analytes. The APCI source temperature was mean recoveries of BDE-47, 100, 99, 153, 183, 209 and PBEB were
maintained at 150 °C under “dry” conditions. The MS was operated in higher than 120% with RSD > 10% when using PSA. For the PSA and
multiple reaction monitoring (MRM) mode and two MRM transitions C18 mixture, the mean recoveries of BDE-153 and 183 were higher than
were set for each compound. One transition served as the quantification 120%. Thus, we found that C18 was the optimal choice because sa-
transition, and the other was the qualifier transition. Target compounds tisfactory recoveries could be obtained for all the analytes. Ad-
were quantified using the corresponding 13C-labeled IS. Since we did ditionally, more lipophilic compounds would be coextracted when
not purchase 13C-labeled BDE-100, 13C12-BDE-99 was used as IS for using n-hexane/acetone as the extractor, and C18 was more suitable for
BDE-100. Additionally, 13C12-BDE-28 was used as IS for PBT and PBEB, removing lipophilic compounds. Subsequently, the amount of C18 was
and 13C12-BDE-47 was used as IS for HBB. Specific instrumental para- optimized also by the recovery test, and mean recoveries (n = 3) of the
meters, including retention time, collision energies and monitored analytes are shown in Fig. 2. When 100 mg to 500 mg of C18 were used,
MRM transitions are listed in Table 1. satisfactory recoveries were observed for most of the analytes except for
BDE-183 and DBDPE. For BDE-183, the mean recovery increasing from
2.4. Method validation 106% to 147% with the amount of C18 increasing from 100 mg to
500 mg. However, the opposite tendency was observed for DBDPE, as
After optimization of sample treatment and instrumental perfor- the mean recovery decreased from 103% to 57% with the amount of
mance, recovery experiments with three spiking levels were conducted C18 increasing from 100 mg to 500 mg. Generally, satisfactory re-
to validate the final methodology. Fetal bovine serum was used as the coveries were obtained when 100 mg of C18 was used. Therefore, this
spiked matrix with spiking levels of 0.01, 0.1 and 1 ng/mL, except for amount was used in the cleanup step for better recovery.
BDE-209 and DBDPE. Because the two highly brominated BFRs, BDE-
209 and DBDPE, had relatively low sensitivity due to thermal de- 3.2. Ionization optimization
gradation, the three spiking levels of BDE-209 and DBDPE were set to
0.1, 1 and 10 ng/mL, respectively. Former studies have proved that BFRs are favored over ionizing in
positive mode (APCI+) [22,23]. Therefore, the MRM optimization was
3. Results and discussion performed by injecting high concentrations of individual BFR standard
solution in APCI+ mode. The injected concentration for BDE-209 and
3.1. Optimization of the QuEChERS approach DBDPE is 1 μg/mL, whereas the injected concentration for other BFRs is
100 ng/mL. First, two ionization modes, charge transfer and proton
The first step of a QuEChERS approach is to extract target com- transfer, were compared to evaluate the impact of modifiers and to
pounds out of the matrix and transfer them into the organic phase with achieve the most abundance of precursor ions. In charge transfer mode,
an extraction solvent and salt combination. Acetonitrile was the most no modifier is added to the ion source chamber and is normally called
commonly used extractor in early use [19]. In the present study, a series the “dry” condition, whereas in proton transfer mode, modifiers (water,
of organic solvents, including acetonitrile, methyl tert-butyl ether methanol, formic acid, etc.) were placed in an uncapped vial located

66
Y. Wang et al.
160
140
120
Mean Recovery (%)
100
80
60
40
20
Fig. 1. Comparison of PSA and C18 as d-SPE sorbents in the clean-up step.
within a specially designed holder in the source door. Previous studies
showed that the presence of the modifier in the ion source chamber
dramatically enhanced the generation of protonated molecular ion
([M + H]+) [23–25]. In the present study, molecular ion clusters (M+%
or [M + n]+%) were observed in the spectrum of PBDE congeners as
base precursor ions under “dry” conditions. Under the proton transfer
mode, when water was used as the modifier, the spectra of PBDE
congeners showed that the water increased the proportion of
[M + H]+, and the presence of water led to a remarkable decrease of
the M+% intensity. On the other hand, for the NBFRs, we found that the
intensities of M+% under “dry” conditions were much higher than those
of the [M + H]+ under proton transfer conditions. Fig. 3 shows the
APCI spectrum of DBDPE (C14H4Br10) under both proton transfer mode
(top) and charge transfer mode (bottom), and clusters corresponding to
isotopic clusters of M+% can be observed as base peaks of the spectrum
under both modes. Additionally, the abundance of the molecular ion
under the charge transfer mode was stronger than that under the proton
transfer mode. Although a previous study found that the base peak of
DBDPE was a fragment ion of C8H4Br5+% under both charge transfer
mode and proton transfer mode and neither M+% nor [M + H]+ was
observed, the molecular ion M+% was more abundant than any other
fragment ion in this study [23]. Finally, molecular ions under “dry”
conditions were selected as precursor ions.
After the molecular ions were selected as precursor ions, product
ions fragmented by collision-induced dissociation were investigated
and optimized at different collision energies in the range of 5–60 eV,
180
160
Mean Recovery (%)
140
120
100
80
60
40
20
0
Fig. 2. Effect of different C18 amounts on lipid removal.
67
Journal of Chromatography B 1099 (2018) 64–72
C18
C18+PSA
PSA
and a compromise between intensity and selectivity was emphatically
considered. The optimal results of the collision energies, precursor and
product ions are summarized in Table 1. For PBDEs, the losses of Br2
and CBr3O were commonly observed, whereas for some highly bromi-
nated congeners, including BDE-183 and BDE-209, the loss of Br4 was
observed. For BTBPE, losses of -C6H2Br3O or -C6H2Br4O from the mo-
lecular ions (M+%) were observed. For DBDPE, losses of -C8H4Br5 and
-C7H2Br5 from the molecular ions (M+%) were observed. PBT and HBB,
the structure of which was based on benzene ring substituting by a
bromine atom, showed the loss of one bromine atom. PBEB lost methyl
first because of the branched carbon chain, and the loss of one bromine
atom was also observed.
A series of parameters, including transfer line temperature, corona
pin current and collision gas flow were optimized. We found that
transfer line temperature is a predominant influence in the signal in-
tensity of the analytes. For most of the analytes, the signal intensity
increased with increasing transfer line temperature (Fig. 4), and
therefore a high transfer line temperature (390 °C) was set for better
sensitivity. In particular, the chromatographic peak, retention time and
signal intensity of two highly brominated BFRs, BDE-209 and DBDPE,
were all found to be intensely affected by the transfer line temperature.
The instrumental analysis of BDE-209 and DBDPE will be discussed in
detail in Section 3.3. For the corona pin current, the signal intensity
also increased with increasing corona pin current (Supplementary Fig.
S1), and a relatively high corona pin current of 3 μA was used. For the
collision gas flow, the signal intensities of most analytes increased when
100 mg
200 mg
300 mg
400 mg
500 mg
Y. Wang et al. Journal of Chromatography B 1099 (2018) 64–72

Fig. 3. APCI spectra of DBDPE under proton transfer mode (top) and charge transfer mode (bottom).

360 Br− ion was selected for the analysis of DBDPE to achieve relatively
5 370 sensitive quantification under NCI mode, and 13C12-BDE-209 was used
380
Signal intensity (log transform)

390
as the IS for DBDPE [12,26,27]. However, our previous study found that
4 the recoveries and RSDs of DBDPE were both unstable and un-
satisfactory under NCI mode, regardless of sample treatment techni-
3 ques. This proved that 13C12-BDE-209 is not a suitable IS for DBDPE due
to their different response factors [26], and therefore accurate quanti-
2 fication of DBDPE is not easy using NCI [12]. Although NCI is more
sensitive to bromine than is EI, EI or APCI shows much better selectivity
1 than does NCI if they are coupled to tandem mass spectrometry, which
is especially important when analyzing complex matrices with plenty of
0
0 4 3 3 9 possible interferences [11]. More importantly, 13C-labeled IS of the
28 47 10 99 15 15 18 20 EB T BB
E PE
E- E- E- E- E- E- E- E- PB BP
PB H
BT BD lower brominated BFRs can be used for quantification under EI or APCI
BD BD BD BD BD B D B D B D D
mode, and is far more reliable than nonlabeled IS. Thus, we conclude
Fig. 4. Signal intensity of a standard solution under different transfer line that EI mode and APCI mode are more reliable than NCI mode because
13
temperatures (DBDPE and BDE-209100 pg/μL, other analytes 10 pg/μL). C-labeled IS can be used for the target compounds.
For the working modes (MS or MS/MS), the sensitivity of MS/MS
is much better than MS when they are coupled to the same ion
the collision gas flow increased from 0.1 mL/min to 0.25 mL/min. source. However, the comparison between EI-MS/MS and NCI-MS in
However, when the gas flow increased to 0.5 mL/min, lower signal our previous study showed that the sensitivity of EI-MS/MS was
intensities were observed, and therefore the collision gas flow was found to be comparable to that of NCI-MS for tri- to hepta-BDE,
maintained at 0.25 mL/min (Supplementary Fig. S2). whereas the signal intensities of both parent ion ([M-2Br]+) and
product ion ([M-4Br]+) of BDE-209 in EI-MS/MS were far lower than
3.3. Analysis of BDE-209 and DBDPE that of the [M-C6Br5]− ion in NCI-MS, resulting in a very high LOD
(> 1 ng/mL) in EI-MS/MS [13]. Similarly, for DBDPE, the signal
In previous studies, the accurate detection of BDE-209 and DBDPE intensities of both parent ion and product ion in EI-MS/MS were also
was found to be quite difficult because these two compounds are both found to be extremely low. That is, sensitive determination of BDE-
thermally unstable and have high boiling points. To date, GC-NCI-MS or 209 and DBDPE in EI-MS or EI-MS/MS is emphatically impossible.
GC-EI-MS/MS are the most commonly used techniques for the instru- When MS/MS was coupled to APCI, because APCI is a soft (low-en-
mental analysis of BDE-209 and DBDPE [12,13]. In this study, the de- ergy) ionization technique, more abundant molecular or quasi-mo-
veloped GC-APCI-MS/MS was compared with GC-NCI-MS and GC-EI- lecular ions of BDE-209 and DBDPE can be formed, resulting in
MS/MS. subsequent highly sensitive and selective detection, and more im-
For MS detection, because BDE-209 and DBDPE are both thermally portantly, the ability to use 13C-DBDPE in APCI-MS/MS for accurate
unstable, degradation of these two compounds during GC–MS analysis quantification of DBDPE. Compared to NCI-MS and EI-MS/MS, much
was inevitable. Therefore, 13C12- BDE-209 and 13C14-DBDPE were used better sensitivity was obtained when using APCI-MS/MS. In this
as IS to compensate for the degradation. For the ionization technique, study, LODs of BDE-209 and DBDPE were 1.08 and 30 pg/mL, re-
the ion [C6Br5O]− obtained in the fragmentation of BDE-209 was spectively, which were much lower than those obtained using NCI-
generally used under NCI mode. Nevertheless, fragment ions containing MS (LODs: 50 pg/mL for BDE-209 and 470 pg/mL for DBDPE) [12],
carbon atom of DBDPE or 13C14-DBDPE showed extremely low signal let alone EI-MS/MS (LOD > 1 ng/mL) [13].
intensities compared to the Br− ion under NCI mode, which prevented For GC analysis, a short column (15 m at most) and relatively high
us from using 13C14-DBDPE as an appropriate IS for DBDPE. Thus, the velocity of carrier gas should be used for the heat-labile compounds

68
Y. Wang et al.
Fig. 5. The chromatographams of BDE-209 and DBDPE. The injection of 100 pg/μL each of BDE-209 and DBDPE standard solution with transfer line temperature
increased from 310 °C to 390 °C.
including BDE-209, DBDPE and corresponding 13C-labeled ISs for re-
ducing retention time and avoiding on-column degradation as far as
possible. In this study, a 15 m DB-5MS column was used. However,
chromatographic peak tailing, retention time delay and reduced in-
tensity for BDE-209 and DBDPE were also observed, and we found that
the uneven transfer line temperature was the primary cause. The tem-
perature on the transfer line is not homogeneous, the temperature at the
end of the transfer line was much lower than that at the middle, thus,
the BDE-209 and the DBDPE are apt to cool and condense when they
arrive at the end of the transfer line. The condensation of BDE-209 and
DBDPE at the end of the transfer line could lead to unstable chroma-
tographic behavior. In this study, we found that the high transfer line
temperature may compensate for this influencing factor. Because the
temperature limit of the DB-5MS column is only 340 °C, and the quartz
capillary column is easily damaged at high temperatures, a deactivated
quartz capillary column with a maximum operating temperature of
400 °C was installed in the transfer line and connected to the end of the
DB-5MS column (Supplementary Fig. S3). As shown in Figs. 4 and 5,
with the transfer line temperature increased from 310 °C to 390 °C,
chromatographic peak tailing, retention time delay and signal intensity
loss of BDE-209 and DBDPE were found to be evidently improved. Fi-
nally, the transfer line temperature was maintained at 390 °C for op-
timal sensitivity. However, at such a high temperature, the capillary
column installed in the transfer line was damaged over time. According
to our observations, after a new deactivated capillary column was in-
stalled into the transfer line, the reappearance of peak tailing, retention
time delay and reduced intensity were observed after approximately 50
injections. Although the reduced signal intensity can be compensated
by the 13C-labeled ISs, we still suggest that the “sacrificed” column in
the transfer line should be displaced by a new one for another dozens of
injections for optimal sensitivity. In summary, we conclude that GC-
APCI-MS/MS is the best technique for accurate, sensitive and selective
detection of BDE-209 and DBDPE.
69
Journal of Chromatography B 1099 (2018) 64–72
3.4. Method validation
Typical chromatograms of a spiked fetal bovine serum sample and a
human serum sample are shown in Figs. 6 and 7. Five-level calibration
curves were drawn for quantification. Concentrations of standard so-
lutions ranged from 10 to 1000 pg/μL for BDE-209 and DBDPE, and 1 to
100 pg/μL for other compounds, respectively. Concentrations of 13C-
labeled BDE-209 and DBDPE were both 100 pg/μL, whereas those of
other 13C-labeled ISs were all 10 pg/μL. High correlation coefficients
(r > 0.99) were obtained (Table 2), indicating good linearity
throughout the validated range. The accuracy and precision of the de-
veloped methodology were validated by a matrix spiking test, wherein
spiked fetal bovine serum with three spiked levels was analyzed, and
each spiked level was tested with five replicates (Table 2). Unspiked
fetal bovine serum was analyzed beforehand and no target compounds
were found above the LODs. The recoveries and RSDs are also listed in
Table 2. Mean recoveries of spiked samples ranged from 83.4% to 118%
with RSDs of ≤21.1%. Additionally, before the detection of real serum
samples, the standard reference material SRM 1958 was tested as a
quality control sample, and our result showed that the average values of
most of the selected PBDE congeners agreed with the corresponding
certified values in SRM 1958 (Supplementary Table S1). In summary,
the recovery test and the SRM 1958 test both indicated that the sample
treatment and instrumental technique developed in this study are re-
liable and stable for the analysis of BFRs in serum. In particular, by
using 13C-DBDPE as IS, good recoveries of DBDPE were obtained, with a
range of 80.6% to 117% (RSDs ≤16.3%). In our previous study, serum
DBDPE was analyzed by GC-NCI-MS. However, both recoveries and
RSDs were unstable and unsatisfactory, and this was clearly the result
of the different response factors between DBDPE and the IS 13C-BDE-
209 in NCI-MS [12].
The S/N at the lowest spiked level was calculated to roughly esti-
mate the methodological limits of detection (mLOD) (S/N = 3) by
Y. Wang et al.
Fig. 6. Chromatogram of a spiked fetal bovine serum sample (1 ng/g serum for BDE-209 and DBDPE, 0.1 ng/g serum for other analytes).
extrapolation. The LODs were 0.04 to 1.08 pg/mL except for DBDPE
(Table 2). In our previous studies, serum BFRs were tested using NCI-
MS and EI-MS/MS [12,13]. LODs in the present study were at least one
order of magnitude lower than those estimated using NCI-MS or EI-MS/
MS, indicating that APCI-MS/MS is more sensitive.
3.5. Method application to real samples
The developed methodology was applied to test human serum
samples collected from 30 donors living in a production area located in
Weifang City, North China. For comparison, another 30 serum samples
were collected from Beijing and tested. The Weifang donors were re-
cruited workers in a local food processing plant, and there was no BFR
manufacturing plant in the vicinity of the factory. All 30 donors were
female, aged between 24 and 40 with an average age of 33.5 years and
all lived in Weifang for over one year. The Beijing donors were re-
cruited company employees. All donors were also female, aged between
25 and 39 with an average age of 31.5 years and all lived in Beijing for
over one year. The questionnaires showed no occupational exposure in
donors from both the cities. Serum BFR concentrations of the two
groups are listed in Table 3, and statistical analysis was conducted by
SPSS 22.0 (IBM). For Σ3–7PBDEs (tri- to hepta-BDEs), all 7 congeners
were detected above the LOD in all the serum samples except for BDE-
154, which was not detected in 8 Beijing samples and one Weifang
sample. Because China stopped producing and using Penta- and Octa-
BDE, we inferred that these lower brominated congeners derived from
some ‘older materials’ which contained Penta- and Octa-BDE. Ad-
ditionally, byproducts produced during deca-BDE production and de-
gradation may also be sources of the lower brominated congeners [28].
Spearman rank correlation analysis indicated that serum concentrations
70
Journal of Chromatography B 1099 (2018) 64–72
of some lower brominated congeners (BDE-28, 99, 154, 153 and 183)
were significantly correlated with those of BDE-209 in the Weifang
group (Supplementary Table S2), indicating that they have a similar
source of exposure. In 2016, we collected and tested 18 workshop air
samples from a deca-BDE manufacturing plant located in Weifang
(Supplementary Table S3), and BDE-209 was definitely the pre-
dominant congener in the air (median level of 45.8 μg/m3), whereas tri-
to hepta-BDEs were also detected in these air samples, though at low
levels (median level of 6.89 ng/m3), which indicated that tri- to hepta-
BDEs simultaneously came into existence with the production of deca-
BDE. However, in the Beijing group, no correlation was observed be-
tween BDE-209 and other lower-brominated congeners.
BDE-209 was the predominant congener in ΣPBDE in both the
groups, constituting at least 90% of the total median PBDEs. BDE-209 is
the primary component in deca-BDE. After the Penta-BDE and Octa-
BDE were phased out in China, deca-BDE became and presently is the
only produced and used commercial PBDE. Thus, the broad and heavy
industrial usage of deca-BDE may be the primary source for the pre-
dominance of BDE-209. Additionally, due to the short half-life of BDE-
209 in human blood (only 15 days) [29], the occurrence of BDE-209 in
human blood has been regarded as an indicator of recent exposure to
deca-BDE. Therefore, the predominance of BDE-209 may indicate that
local inhabitants of both Weifang and Beijing are continuously exposed
to elevated levels of deca-BDE. When the two groups were compared,
the median level of Σ3–7PBDEs in the Weifang City group was 29.1 pg/
mL, which was two times higher than that in the Beijing group. A
nonparametric test (Mann–Whitney U test) showed that levels of
Σ3–7PBDEs in the Weifang group were significantly higher than those in
Beijing group (p = 0.006). Higher BDE-209 levels were also found in
the Weifang group (Mann–Whitney U test, p = 0.034). In summary, the
Y. Wang et al. Journal of Chromatography B 1099 (2018) 64–72

Fig. 7. Chromatogram of a human serum sample collected from a BFRs product area.

serum PBDE levels of the Weifang group were significantly higher than
those of Beijing group, indicating that the production of BFRs in Wei-
fang City has hardly affected the surrounding environment, resulting in
elevated BFRs exposure of local inhabitants.
For the NBFRs, PBT, PBEB, HBB and BTBPE could be detected above
the LOD in both groups but at relatively low levels. The median levels of
these NBFRs were all lower than those of Σ3–7PBDEs, indicating that
these emerging contaminants are widespread but their usage is limited
in China. In particular, although DBDPE was not found in most of the
Beijing samples, high levels of DBDPE were detected in Weifang group.
The median level of DBDPE in the Weifang group was 226 pg/mL,
which was slightly lower than that of BDE-209 but much high than
other compounds. DBDPE, as a replacement for deca-BDE, has become
one of the most popular BFR in China because of the regulation of
PBDEs in recent years [30]. Weifang is a main DBDPE production area
in China, and the high levels of DBDPE in serum from Weifang was
likely a result of its production.

Table 2
Correlation coefficient, LOD, recovery and precision of the optimized method.
Analyte r2 LOD (pg/mL) Recovery (%)a RSD (%) Recovery (%)a RSD (%) Recovery (%)a RSD (%)

Spiked 0.01 ng/mL Spiked 0.1 ng/mL Spiked 1 ng/mL

BDE-28 0.9997 0.04 106 8.76 98.2 5.52 103 4.91
BDE-47 0.9984 0.11 99.5 8.52 113 8.28 118 6.97
BDE-100 0.9966 0.06 108 7.94 100 12.4 103 6.29
BDE-99 0.9978 0.06 107 7.16 10 6.53 114 6.72
BDE-154 0.9972 0.18 98.7 11.2 102 5.49 112 7.11
BDE-153 0.9984 0.19 113 2.61 116 7.87 118 9.84
BDE-183 0.9951 0.22 106 9.85 92.1 13.9 104 6.64
PBT 0.9996 0.11 92.3 7.58 98.8 8.59 103 5.51
HBB 0.9971 0.13 89.7 12.6 83.4 21.1 86.7 19.5
PBEB 0.9931 0.06 117 8.95 102 4.49 103 10.2
BTBPE 0.9978 0.07 112 9.98 96.8 4.59 99.1 11.9

Spiked 0.1 ng/mL Spiked 1 ng/mL Spiked 10 ng/mL

BDE-209 0.9962 1.08 105 9.35 103 8.51 11 8.73
DBDPE 0.9991 30 115 16.3 100 10.2 103 10.2

a
Mean recovery, n = 5.

71
Y. Wang et al. Journal of Chromatography B 1099 (2018) 64–72

Table 3
BFR levels in real serum samples (pg/mL).
Analyte Donors from Weifang (n = 30) Donors from Beijing (n = 30)

Mean SD Median Min Max Mean SD Median Min Max

BDE-28 1.24 1.74 2.05 0.56 8.19 0.76 0.64 0.53 0.1 2.96
BDE-47 2.15 1.51 2 0.57 6.74 1.35 0.73 1.15 0.6 4.22
BDE-100 1.06 1.77 0.61 0.21 10.1 1.19 2.05 0.52 0.15 10.7
BDE-99 1.07 0.69 0.69 0.16 9.09 3.84 4.61 2.07 0.71 21.4
BDE-154 1.33 2.05 0.92 nd 11.5 0.95 2.63 0.25 nd 13.6
BDE-153 15.4 10.1 12.3 2.38 38.8 17.2 40.7 6.04 2.67 204
BDE-183 9.49 6.9 7.01 1.29 25.6 13 39.2 2.6 1.23 198
∑3-7PBDEs 33 21.2 29.1 7.13 93.1 38.2 79 14.2 7.52 335
BDE-209 290 174 244 71.1 710 226 217 147 36.8 906
∑PBDEs 323 187 266 87.1 741 264 236 160 51.9 933
PBEB 1.73 4.15 0.43 nd 19.5 0.12 0.31 nd nd 1.71
PBT 16.4 9.73 13.8 nd 39.1 1.49 1.46 1.06 0.57 7.28
HBB 4.24 2.81 3.24 0.53 11.8 0.77 0.60 0.54 0.22 2.98
BTBPE 21 32.3 9.02 nd 130 2.59 4.64 1.1 0.11 18.54
DBDPE 380 531 226 32.6 2640 nd nd nd nd 277

nd: not detected.

4. Conclusion
The main task of this study was to develop a reliable and sensitive
analytical method to monitor BFRs in human serum. New methodolo-
gies based on a QuEChERS approach and GC-APCI-MS/MS were de-
veloped and validated for the simultaneous extraction, cleanup and
analysis of eight usually monitored PBDE congeners and five additional
NBFRs. The QuEChERS approach, as a simple and fast sample treatment
procedure, was optimized and proved to be efficient and straightfor-
ward with satisfactory recovery. The use of APCI-MS/MS was evaluated
as an alternative technique for GC-MS analysis of BFRs, and a notable
improvement in sensitivity and reliability was obtained compared to EI-
MS/MS and NCI-MS. GC-APCI-MS/MS proved to be the best technique
for accurate, sensitive and selective detection of two highly brominated
BFRs, BDE-209 and DBDPE.
Conflicts of interest
The authors have no conflict of interest to declare.
Acknowledgements
The authors thank all the donors, Shandong Center for Disease
Control and Prevention, and Beijing Shunyi Hospital for assisting this
study. This work was supported by the National Natural Science
Foundation of China (21477083, 21777107, 21537001, 21507018,
81703198); the National Key Research and Development Program of
China (2017YFC1600500); the Chinese Academy of Sciences “Strategic
Leading Science and Technology Projects” (XDB14010100).
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://
doi.org/10.1016/j.jchromb.2018.09.015.
References
[1] T.J. McGrath, A.S. Ball, B.O. Clarke, Environ. Pollut. 230 (2017) 741–757.
[2] G. Yu, Q. Bu, Z. Cao, X. Du, J. Xia, M. Wu, J. Huang, Chemosphere 150 (2016)
479–490.
[3] M.A. Abdallah, S. Harrad, A. Covaci, Anal. Chem. 81 (2009) 7460–7467.
[4] UNEP, http://chm.pops.int/Implementation/NIPs/Guidance/
GuidancefortheinventoryofPBDEs/tabid/3171/Default.aspx, (2015).
[5] A. Covaci, S. Harrad, M.A. Abdallah, N. Ali, R.J. Law, D. Herzke, C.A. de Wit,
Environ. Int. 37 (2011) 532–556.
[6] Z. Shi, L. Zhang, J. Li, Y. Wu, Chemosphere 198 (2018) 522–536.
[7] M. Ren, H. Zeng, P.A. Peng, H.R. Li, C.M. Tang, J.F. Hu, Environ. Pollut. 226 (2017)
394–403.
[8] P. Li, H. Wu, Q. Li, J. Jin, Y. Wang, Environ. Monit. Assess. 187 (2015) 719.
[9] S. Krol, B. Zabiegala, J. Namiesnik, Talanta 93 (2012) 1–17.
[10] Z. Shi, Y. Wang, P. Niu, J. Wang, Z. Sun, S. Zhang, Y. Wu, J. Sep. Sci. 36 (2013)
3402–3410.
[11] I. Fulara, M. Czaplicka, J. Sep. Sci. 35 (2012) 2075–2087.
[12] L. Gao, J. Li, Y. Wu, M. Yu, T. Chen, Z. Shi, X. Zhou, Z. Sun, Anal. Bioanal. Chem.
408 (2016) 7835–7844.
[13] J. Li, T. Chen, Y. Wang, Z. Shi, X. Zhou, Z. Sun, D. Wang, Y. Wu, J. Sep. Sci. 40
(2017) 709–716.
[14] A. Covaci, S. Voorspoels, J. de Boer, Environ. Int. 29 (2003) 735–756.
[15] J. Rivera-Austrui, K. Martinez, M. Abalos, C. Sales, T. Portoles, J. Beltran, J. Saulo,
B.H. Aristizabal, E. Abad, J. Chromatogr. A 1513 (2017) 245–249.
[16] K.L. Organtini, A.L. Myers, K.J. Jobst, E.J. Reiner, B. Ross, A. Ladak, L. Mullin,
D. Stevens, F.L. Dorman, Anal. Chem. 87 (2015) 10368–10377.
[17] Y. Zhang, R. Li, J. Fang, C. Wang, Z. Cai, Chemosphere 198 (2018) 303–310.
[18] Z. Cheng, F. Dong, J. Xu, X. Liu, X. Wu, Z. Chen, X. Pan, J. Gan, Y. Zheng, Food
Chem. 231 (2017) 365–373.
[19] C. Diez, W.A. Traag, P. Zommer, P. Marinero, J. Atienza, J. Chromatogr. A 1131
(2006) 11–23.
[20] C. Cabrera Lda, S.S. Caldas, O.D. Prestes, E.G. Primel, R. Zanella, J. Sep. Sci. 39
(2016) 1945–1954.
[21] S.J. Lehotay, K. Mastovska, S.J. Yun, J. AOAC Int. 88 (2005) 630–638.
[22] S. Lv, Y. Niu, J. Zhang, B. Shao, Z. Du, Sci. Rep. 7 (2017) 43998.
[23] T. Portoles, C. Sales, B. Gomara, J.V. Sancho, J. Beltran, L. Herrero, M.J. Gonzalez,
F. Hernandez, Anal. Chem. 87 (2015) 9892–9899.
[24] T. Portoles, L. Cherta, J. Beltran, F. Hernandez, J. Chromatogr. A 1260 (2012)
183–192.
[25] M. Raro, T. Portoles, J.V. Sancho, E. Pitarch, F. Hernandez, J. Marcos, R. Ventura,
C. Gomez, J. Segura, O.J. Pozo, J. Mass Spectrom. 49 (2014) 509–521.
[26] A. Konstantinov, G. Arsenault, B. Chittim, T. Kolic, K. MacPherson, A. McAlees,
R. McCrindle, D. Potter, E.J. Reiner, C. Tashiro, B. Yeo, Chemosphere 64 (2006)
245–249.
[27] P. Lopez, S.A. Brandsma, P.E. Leonards, J. de Boer, Anal. Bioanal. Chem. 400 (2011)
871–883.
[28] H.M. Stapleton, N.G. Dodder, Environ. Toxicol. Chem. 27 (2008) 306–312.
[29] S.C. Hammel, K. Hoffman, A.M. Lorenzo, A. Chen, A.L. Phillips, C.M. Butt, J.A. Sosa,
T.F. Webster, H.M. Stapleton, Environ. Int. 107 (2017) 181–189.
[30] S. Zhang, X. Gu, Chin. J. Plast. Addit. 6 (2013) 12–15.

72