Hydroxylated Metabolites of Polybrominated Diphenyl Ethers in Human Blood Samples from the United States

Background A previous study from our laboratory showed that polybrominated diphenyl ethers (PBDEs) were metabolized to hydroxylated PBDEs (HO-PBDEs) in mice and that para-HO-PBDEs were the most abundant and, potentially, the most toxic metabolites. Objective The goal of this study was to determine the concentrations of HO-PBDEs in blood from pregnant women, who had not been intentionally or occupationally exposed to these flame retardants, and from their newborn babies. Methods Twenty human blood samples were obtained from a hospital in Indianapolis, Indiana, and analyzed for both PBDEs and HO-PBDEs using electron-capture negative-ionization gas chromatographic mass spectrometry. Results The metabolite pattern of HO-PBDEs in human blood was quite different from that found in mice; 5-HO-BDE-47 and 6-HO-BDE-47 were the most abundant metabolites of BDE-47, and 5′-HO-BDE-99 and 6′-HO-BDE-99 were the most abundant metabolites of BDE-99. The relative concentrations between precursor and corresponding metabolites indicated that BDE-99 was more likely to be metabolized than BDE-47 and BDE-100. In addition, three bromophenols were also detected as products of the cleavage of the diphenyl ether bond. The ratio of total hydroxylated metabolites relative to their PBDE precursors ranged from 0.10 to 2.8, indicating that hydroxylated metabolites of PBDEs were accumulated in human blood. Conclusions The quite different PBDE metabolite pattern observed in humans versus mice indicates that different enzymes might be involved in the metabolic process. Although the levels of HO-PBDE metabolites found in human blood were low, these metabolites seemed to be accumulating.

to their hydroxylated metabolites, especially the hydroxylated PBDEs (HOPBDEs). For instance, levels of serum thyroxine (T 4 , a thy roid hormone and the precursor of active thy ronine, T 3 ) were significantly decreased when rats were exposed to PBDEs (Darnerud et al. 2007;Stoker et al. 2004;Zhou et al. 2002). The effect of PBDEs on T 4 levels may require metabolic activation because HOPBDEs, but not the PBDE congeners themselves, behave as ligands for human transthyretin (TTR; a major thyroid hormone transport protein) in vitro (Meerts et al. 2000). In addition, Hamers et al. (2008) reported that the trans thyretinbinding potencies of HOPBDEs were orders of magnitude higher than that of BDE47. Similarly, PBDE has mild estro genic effects in mice (MercadoFeliciano and Bigsby 2008a), and these effects are likely caused by HOPBDEs that act as ligands for the estrogen receptor (MercadoFeliciano and Bigsby 2008b). In addition, HOPBDEs were shown to inhibit estradiolsulfotransferase (Hamers et al. 2008) and placental aromatase (Cantón et al. 2008).
HOPBDEs have been identified in blood samples from rats and mice after exposure to PBDE mixtures (Malmberg et al. 2005;Qiu et al. 2007) and observed in blood samples from wild animals such as fishes, birds, and mammals (Marsh et al. 2004;Valters et al. 2005;Verreault et al. 2005). However, with one exception, there have been no reports about PBDE metabolites in human blood. The exception was a report about HOPBDEs in pooled human blood samples taken from children living or working at a municipal waste disposal site in Managua, Nicaragua (Athanasiadou et al. 2008). To study the metabolism of PBDEs in humans, we have identified and quantitated the hydroxylated metabolites of PBDEs, including HOPBDEs and bromophenols, in 20 individual human blood samples from pregnant women and newborn babies from the United States. HOPBDEs and bromophenols were both found to be important metabolites of PBDEs in plasma after mice were exposed to a com mercial PBDE mixture (Qiu et al. 2007).
Here we report these HOPBDE concentra tions and compare these levels to those meas ured in mouse blood after exposure to high levels of DE71, a commercial pentaBDE mixture (Qiu et al. 2007 All the phenolic compounds were methy lated with fresh diazomethane, which was prepared from Diazald (Sigma Chemical Co., St. Louis, MO) (Black 1983). All the organic solvents and water used for the extraction and cleanup procedures were residueanalysis grade.
Sample collection. Human studies were performed in accordance with the guidelines and approval of the Institutional Review Board at Indiana University School of Medicine. Pregnant women were enrolled in the study during . Random fetal blood samples (n = 16) were collected from the umbilical cord vein by syringe after delivery; these samples were not accompanied by any clinical data or other information about the pregnant women. Upon consent to enter the study, four women were administered a questionnaire to determine potential sources of contamination; none of the subjects had any identifiable source of occupational expo sure to PBDE. Maternal blood was obtained upon admission to the maternity ward; in this study, only one of the four maternal sam ples was matched with a fetal sample. Blood samples were collected in heparinized tubes, maintained at 4°C, and centrifuged at 800 × g for 15 min to allow collection of the plasma fraction. All the plasma samples were kept at -20°C until extraction.
Sample extraction and preparation. We slightly modified previous methods (Hovander et al. 2000;Qiu et al. 2007) for this study. Twenty samples (from 4.2 g to 13.8 g, with an average wet weight of 8.9 g) were analyzed. Before extraction, each sample was transferred to a clean centrifuge tube and spiked with a known amount of BDE77 and 4HO 13 C 12 PCB187 as recovery surrogate standards. Hydrochloric acid (1 mL, 6 M) and 2propanol (6 mL) were added; the sam ple was vortexed after each addition. After denaturizing the samples, they were extracted three times, each time with 6 mL of a hexane/ methyl tertbutyl ether mixture (1:1 by vol ume). The organic extracts were combined, and approximately 20% of the solution was removed for gravimetric determination of lipid mass. The rest of each sample was blown down to 2 mL with clean nitrogen, and 2 mL of potassium hydroxide (0.5 M in 50% etha nol) was added to ionize the phenolic ana lytes. After extraction with hexane three times to separate the PBDEs, the aqueous phase was acidified with hydrochloric acid (2.1 mL, 0.5 M), then the phenolic compounds were extracted three times with a hexane/methyl tertbutyl ether mixture (9:1 by volume).
The neutral fraction was treated with 5 mL of concentrated sulfuric acid twice to remove lipids, followed by alumina column chromatography (6 cm × 0.6 cm i.d., with 0.5 cm anhydrous sodium sulfate on the top). The column was eluted with 8 mL of hexane, followed by 8 mL of a hexane/dichlo romethane mixture (3:2 by volume). The PBDE congeners were in the second fraction. BDE71 was added as an internal standard, and the samples were blown down to approx imately100 µL before gas chromatographic mass spectrometry (GC/MS) analysis.
We concentrated the phenolic fraction by nitrogen blowdown, and the residual water was removed with an anhydrous sodium sul fate column (5 cm × 0.6 cm i.d.). To meth ylate the phenolic analytes, samples were treated with diazomethane at room tempera ture overnight. After methylation, the samples were treated with concentrated sulfuric acid three times to remove lipid, followed by alu mina column chromatography, which was the same as that used for the neutral fraction. Finally, BDE166 was added as the internal standard, and samples were blown down to approximately 50 µL for GC/MS analysis. To prevent potential photodegradation, during the whole process the centrifuge tubes were wrapped with aluminum foil, or amber vials were used.
Instrumental analysis. We analyzed both neutral and methylated phenolic fractions by GC/MS (Agilent 6890/5973) with an electroncapturenegative ionization (ECNI) ion source. We used selected ion moni toring of m/z 79 and 81 for quantitation. The GC injection port was held at 285°C, with an injection volume of 2 µL. A non polar Rxi5ms column (15 m length; 250 µm i.d.; 0.25 µm film thickness; Restek Corp., Bellefonte, PA) was used to separate both the neutral and methylated phenolic ana lytes. The GC oven temperature program was as follows: held at 60°C for 1 min; 10°C/ min to 240°C; 25°C/min to 325°C; held for 8 min. The same instrument, but with a polar SP2331 column (30 m length; 250 µm i.d.; 0.20 µm film thickness; Supelco Inc., Bellefonte, PA), was used for the confirma tion of the methylated phenolic analytes. In this case, the GC temperature program was as follows: held at 80°C for 1 min; 10°C/min to 260°C; held for 16 min. Method detection limits were 0.5-2 pg/g plasma.
Quality control. Several quality con trol criteria were used to ensure the correct identification and quantitation of the target compounds. First, the GC retention times matched those of the standard compounds within ± 0.1 min. Second, the signaltonoise ratio was greater than 5:1. Third, the isoto pic ratios for bromine ion pairs were within ± 15% of the theoretical values. In addition to the 20 blood samples, we also prepared 11 blank samples with pure water (~ 10 mL) as the blank matrix. Only 2,4DBP (dibromo phenol), 2,4,6TBP (tribromo phenol), BDE47, and BDE99 were detected in the blank samples, and the blank values were around 9%, 7%, 2%, and 9% of the average concentration values measured in the blood samples, respectively. The recoveries (mean ± SE) of the surrogate standards were 99 ± 3 % for BDE77, and 90 ± 6 % for 4HO 13 C 12 PCB187, respectively. In this article, the data were not blank or recovery corrected.

Results
We measured PBDE congeners 28, 47, 99, 100, 153, and 154, which were the most abundant PBDE congeners observed in human blood in most other studies (Hites 2004). The mean and median concentrations of these six congeners ranged from 2.3 to 70 ng/g and from 0.8 to 13 ng/g lipid, respec tively, in the fetal samples and from 0.5 to 17 ng/g and 0.3 to 15 ng/g lipid, respectively, in the maternal samples [ Table 1; the full data set is given in the Supplemental Material (available online at http://www.ehponline. org/members/2008/11660/suppl.pdf)]. Among these congeners, BDE28, 47, 99, and 100 were detected in all 20 samples, BDE153 was detected in 19 samples, and BDE154 was detected in 17 samples. No methoxylated PBDEs (MeOPBDEs) were detected in the neutral cleanup fraction.
We also measured the concentrations of 18 potential phenolic metabolites, includ ing 3 bromophenols and 15 hydroxylated PBDEs [see Supplemental Material (avail able online at http://www.ehponline.org/ members/2008/11660/suppl.pdf) for the structures of these HOPBDEs]. The 3 bromo phenols were detected in all samples, and 7 of the 15 HOPBDEs were identified and quantitated in all or some of the samples. The concentrations of these hydroxylated metabolites and their percent of the total phenolic compounds are shown in Table 1.

Discussion
PBDEs. Mazdai et al. (2003) reported that the concentrations of PBDEs in fetal blood did not differ from those in the corresponding mater nal blood, and, based on analysis of unpaired maternal and fetal samples, our study con firms this finding [see Supplemental Material (available online at http://www.ehponline. org/members/2008/ 11660/suppl.pdf) for a statistical analysis comparing the fetal and maternal concentrations]. In the present study there was only one set of paired maternalfetal samples; therefore, we could not directly determine if the HOPBDE blood levels corre lated between mother and baby. The fetal liver and adrenal glands express type I and type II enzymes capable of metabolizing xeno biotics that cross the placenta (Syme et al. 2004).
Furthermore, the placenta expresses type I enzymes, and these are inducible by xenobiot ics, such as those in cigarette smoke (Hakkola et al. 1998). Thus, the question of whether the fetus is at greater or lesser risk of exposure to PBDE metabolites requires further study.
In this study, the total concentrations of the 6 PBDE congeners in these 20 samples ranged from 4.7 to 800 ng/g lipid, with an average of 100 and a median of 31 ng/g lipid (Table 1). These values are close to previously reported PBDE concentrations in human blood from Indiana, where the concentra tions ranged from 14 to 580 ng/g lipid (n = 24; Mazdai et al. 2003), and they are close to concentrations measured in fetal blood from Baltimore, Maryland (from not detected to 310 ng/g lipid for BDE47, n = 297; Herbstman et al. 2007). All the concentrations reported here were much higher than those reported for human blood from Europe (e.g., 1.1-20 ng/g lipid for total PBDEs in 50 serum samples from Sweden and not detected to 6.1 ng/g lipid for BDE47 in 81 maternal and fetal serum samples from the Netherlands) (Meijer et al. 2008;Weiss et al. 2006), indicating that North Americans are exposed to higher levels of PBDEs than are Europeans.
The average profile of the PBDE con geners measured in this study is shown in Figure 1. For comparison, the congener profile of DE71, an important commercial penta BDE mixture and the presumptive source of PBDEs in human blood, is also shown. Figure 1 shows that the percentage of BDE99 in human blood was much lower than in DE71, whereas the percentage of BDE47 and 153 in human blood was higher than in DE71. Neglecting the potential for slightly different uptake efficiencies of these congeners, these data suggest that, in human blood, BDE99 may be the least persistent PBDE congener in DE71. In fact, although BDE99 is the most abundant congener of commercial DE71, BDE47 is usually the most abundant congener found in human blood samples (Hites 2004), and BDE153 is sometimes the most abundant congener in people from lowexposure regions (Meijer et al. 2008;Weiss et al. 2006).
Different metabolic rates among the congeners might cause the different conge ner profiles observed for DE71 and human blood. As found in this study, BDE99 was more likely to be degraded to HOPBDEs than were BDE47 and BDE100.
Hydroxylated metabolites. In this study, hydroxylated metabolites included both monohydroxylated PBDEs (HOPBDEs) and bromophenols. Of the 18 assessed hydroxylated metabolites, 7 HOPBDEs and 3 bromophenols were measured in almost all of the 20 samples. Typical chromatograms of methylated HOPBDEs on two different GC columns are shown in Figure 2.
One might assume that the hydroxylated metabolites, being more polar molecules, would be excreted more readily than the par ent PBDEs, and thus, the concentration of these metabolites would be much lower than those of PBDEs in blood. This is not what we observed. The average total concentration of HOPBDEs and bromophenols was 79 ng/g lipid (range, 2.0-900 ng/g lipid). The average concentration ratio of hydroxylated metabo lites to PBDEs was 0.85 (range, 0.10-2.8), indicating that concentration of these metabo lites was comparable to or even higher than that of PBDEs in these samples. This ratio was also much higher than that found in blood samples from young people from Managua, Nicaragua (Athanasiadou et al. 2008), perhaps because some abundant hydroxylated metabo lites (5HOBDE47, 6´HOBDE99, and bromophenols) were not measured in that study. The high concentrations of metabo lites and their relatively high ratio to PBDEs indicate that the hydroxylated metabolites of PBDEs may accumulate in human blood; and thus in this paper, we provide the concen trations of the hydroxylated metabolites on a lipid weight basis.
HO-PBDEs. As discussed above, BDE47 is an important congener in the commercial pentaBDE product and the most abundant PBDE congener found in human blood in this and in most other studies (Hites 2004). According to the proposed metabolic path way for mice (Qiu et al. 2007), there are six possible monohydroxylated PBDE metabo lites of BDE47, presumably produced by    Figure 3. Note that the last three metabolites require a bromine shift via an arene oxide during the hydroxylation process. In our previous study on mice, which were dosed with DE71 at 45 mg/kg, 4HOBDE42 was the most dominant metabolite of BDE 47 and accounted for 56% of the total HOtetraBDEs in mouse plasma, followed by 3HOBDE47 (16%) and 4´HOBDE49 (13%) (Qiu et al. 2007). The metabolite pro file is very different in the human blood sam ples studied here. Although 4´HOBDE49 and 4HOBDE42 were detected in some of the 20 human blood samples, two metabolites formed without a bromine shift (5HOBDE47 and 6HOBDE47) were more abundant, especially 5HOBDE47, which was not even detected in mice exposed to high doses of DE71 (Qiu et al. 2007). As shown in Table 1 and in Figure 3, 5HOBDE47 was the most abundant HOtetraBDE, followed by 6HOBDE47. These two metabolites were detected in all 20 samples and accounted for 90% of the total HOtetraBDEs. 2´HOBDE66 was not detected; 3HOBDE47, 4´HOBDE49 and 4HOBDE42 were detected in some of the samples but at much lower concentrations.

Metabolites of polybrominated diphenyl ethers in human blood
The difference in the metabolic profile between humans and mice may be the result of species differences in cytochrome P450 enzyme expression. The superfamily of P450 has many subfamilies based on amino acid sequence identities (Nelson et al. 1993), and each different subfamily of P450 has a differ ent selectivity in oxidation of the halogenated phenyl ring (Bogaards et al. 1995). Results of our study of PBDE metabolites in mice sug gest that oxidative debromination occurred, and this accounts for the production of several parahydroxylated metabolites from BDE47 (Qiu et al. 2007). The complex chemical reactions involved in oxidative dehalogena tion (Isin and Guengerich 2007) are likely to require specific CYP enzymes. Although the precise set of CYP enzymes involved in oxida tive dehalogenation in mammals is unknown, the lack of parahydroxylated metabolites in human serum is likely due to the differ ent subfamily profiles in mice and humans (Bogaards et al. 2000). The parahydroxylated metabolites are likely to behave as endocrine disruptors (Hamers et al. 2008;Meerts et al. 2000;MercadoFeliciano and Bigsby 2008a), and therefore the lack of these in humans may explain the lack of any correlation between PBDE and thyroid hormone concentrations in blood of mothers and their babies (Mazdai et al. 2003). However, the enzyme profiles of the placenta and fetal liver change through out the course of pregnancy (Hakkola et al. 1998), and there may be transient differences in the PBDE metabolite profiles as a result. Furthermore, placental CYP1A1 is highly inducible by cigarette smoke (Hakkola et al. 1998). The potential for transient differences    in PBDE metabolite profiles during the course of pregnancy and for a correlation between exposure to tobacco smoke and metabolism require further study. Some HOPBDEs are more toxic because of the specific position of hydroxyl group; for example, 4HOBDE42, 4´HOBDE49, and 3HOBDE47 were shown to have about four times stronger affinity to trans thyretin than thyroxin (Hamers et al. 2008).
Our data indicate that human P450 enzymes do not produce many of these toxic isomers. However, 5HOBDE47 has a three times stronger affinity to trans thyretin than thy roxin (Hamers et al. 2008). Given the con centration of 8.7 nM free thyroxin in serum during pregnancy reported by Sterling and Hegedus (1962) (compared with the aver age concentration of 5HOBDE47 of 0.1 nmol/kg plasma, with the highest con centration being 1.2 nmol/kg plasma in the present study), these metabolites, especially 5HOBDE47, might have substantial human effects because of their relatively high concentrations in blood.
BDE99 is the most abundant congener of DE71; however, as discussed above, con centrations of BDE99 were lower than that of BDE47 in human blood in this and in most other studies, perhaps because BDE99 was con verted to hydroxylated PBDEs. Like the metab olite pattern of BDE47, 5´HOBDE99 was the most abundant hydroxylated metabolite of BDE99, followed by 6´HOBDE99 (Table  1 and Figure 2). The ratio of 5´HOBDE99 + 6´HOBDE99 to 5HOBDE47 + 6HOBDE47 was 0.84 on average (range, 0.24-3.2), which was significantly higher than the ratio of BDE99 to BDE47 (0.39 on aver age; range, 0.17-0.69; p < 0.01, ttest). Given the symmetrical structure of BDE47, it has a higher probability to form 5HOBDE47 and 6HOBDE47 than BDE99 has to form 5´HOBDE99 and 6´HOBDE99. Thus, we conclude that BDE99 was more likely than BDE47 to be hydroxylated. This may also explain why the concentration of BDE99 was usually lower than that of BDE47 in human blood, although it was more abundant in the commercial pentaBDE mixtures.
As discussed above, in human blood BDE47 and BDE99 have similar meta bolic profiles, and hydroxylation mainly occurs on the phenyl ring with two bromines. This process resulted in 5HOBDE47 and 5´HOBDE99 as the two most abundant metabolites of BDE47 and BDE99, respec tively. If this process were true for BDE100, we might expect that 5´HOBDE100 should be the most important hydroxylated metabolite of BDE100; however, 5´HOBDE100 was not detected in this study. In fact, there were no other large GC peaks on either column in the retention time regions where hydroxylated pentaBDEs might be expected to elute ( Figure  2). The lack of detection of 5´HOBDE100 suggests that BDE100 is more resistant to hydroxylation than is BDE99, explaining the increasing ratio of BDE100 to BDE99 from commercial DE71 (0.21) to human blood (0.59) (Figure 1).
In the present study, we did not detect 5´HOBDE100 or several other HOpentaBDEs, including 4HOBDE90, 4´HOBDE101, and 4´HOBDE103. Theoretically, these paraHOpentaBDEs could be formed from BDE99 or BDE100 via hydroxylation with a bromine shift; appar ently however, this bromine shift was not caused by human P450 enzymes. We should note that 4HOBDE90 has the same reten tion time as 5´HOBDE99 on non polar GC columns (such as Rxi5 and DB5); however, on polar columns (such as SP2331) these two compounds can easily be distinguished. In another study, Athanasiadou et al. (2008) quantified this peak relative to 4HOBDE90 with a DB5 column.
Bromophenols. In our previous study of mice (Qiu et al. 2007), three bromo phenols were identified in blood samples after expo sure to DE71. These bromophenols were also detected in human blood samples. To our knowledge, 2,4,5TBP is not a commer cial product, thus the presence of 2,4,5TBP in human blood (at 6.4 ng/g lipid) indi cates that the diphenyl ether bond of some PBDE congeners (BDE99 for example) can be cleaved. For 2,4DBP and 2,4,6TBP, although they were also detected in blanks in this and in other studies (Thomsen et al. 2001), the concentrations of these two compounds in the blood samples (averag ing 16 and 4.6 ng/g lipid for 2,4DBP and 2,4,6TBP, respectively) were > 10fold higher than those meas ured in the blanks, indicating that these two bromophenols came from sources other than the laboratory blank. One source of these two contaminants might be exposure to these chemicals because both 2,4DBP and 2,4,6TBP have been used as flame retardants (World Health Organization 1997). Another source, such as 2,4,5TBP, might be cleavage of the diphenyl ether bond of PBDEs. Based on the proposed metabolic pathway for mice (Qiu et al. 2007), 2,4DBP could be a metabo lite of BDE47, and 2,4,6TBP could be a metabolite of BDE100 and BDE154.
The average total concentration of the three bromophenols was 27 ng/g lipid (range, 1.2-310 ng/g lipid). For comparison, the average total concentration of the mea sured HOPBDEs was 52 ng/g lipid (range, 0.8-590 ng/g lipid). Thus, although the bro mophenols might have other sources, the concentrations of bromophenols were com parable to those of HOPBDEs, suggesting that cleavage of the diphenyl ether bond was an important metabolic pathway for PBDEs in humans.

Summary
Unlike the metabolites measured in mice after dosing with PBDE, the HOPBDE metabolites of BDE47 in humans without a bromine shift were abundant. Two metabo lites, 5HOBDE47 and 6HOBDE47, accounted for 90% of the total HOtetra BDE concentration in our subjects. 5´HOBDE99 and 6´HOBDE99 were the most abundant metabolites of BDE99 in our subjects. The relative concentrations between precursor and the corresponding HOPBDE indicated that BDE99 was more likely to be hydroxylated than BDE47 and BDE100, and this observation may explain the different congener profiles noticed for PBDEs in human blood as opposed to the commercial pentaBDE mixture. In addition to HOPBDEs, three bromophenols were also detected in human blood, indicating the cleav age of the diphenyl ether bond of PBDEs. The total concentrations of the hydroxylated metabolites (HOPBDEs and bromophenols) were close to those of the PBDEs, suggesting that these hydroxylated metabolites may be accumulating in human blood.