Levels and Concentration Ratios of Polychlorinated Biphenyls and Polybrominated Diphenyl Ethers in Serum and Breast Milk in Japanese Mothers

Target populations

The present study was approved by the Ethics Committee of the Kyoto University Institutional Review Board, and appropriate written informed consent was obtained from all the participants before sample collection.

After obtaining formal informed consent, we collected blood and breast milk samples from mothers who had delivered and were lactating in maternity hospitals in four regions: Sendai city (population, 1 million) in Miyagi Prefecture, Takarazuka city (population, 250,000) in Hyogo Prefecture, Takayama city (population, 200,000) in Gifu Prefecture, and Shizunai-cho (population, 23,000) in Hokkaido Prefecture.

Collection of serum samples and breast milk samples

Milk samples were self-collected manually into breast pumps with glass containers at the individual hospitals and transferred to 50-mL polypropylene conical tubes (milk tube) that had been thoroughly rinsed with methanol and acetone before use; samples were kept frozen at –20°C. The target volume was > 20 mL from each mother per sample. Blood samples (10 mL) were collected into two 5-mL vacuum blood collection polypropylene tubes (Venoject II; TERUMO Inc., Tokyo, Japan) (blood tube) from cubital vein by physicians or nurses. The blood and milk samples were shipped within 48 hr to Kyoto University. The serum samples were separated by centrifugation at 3,000g for 15 min, transferred to new blood tubes, and stored at –20°C in the Department of Health and Environmental Sciences, Kyoto University Graduate School of Medicine, until analysis.

When the milk samples were collected, we asked the mothers to fill out questionnaires that contained necessary items for milk surveillance (LaKind et al. 2004) and sources of exposure to PBDEs (Ohta et al. 2002; Sakai et al. 2001; Schecter et al. 2005; Wilford et al. 2003), including the duration of lactation, parity, residential history within the previous 5 years, lifestyle and habits, and indoor environment (Supplemental Table 1; available online at http://www.ehponline.org/docs/2006/9032/suppl.pdf).

We prepared eight field blanks per site, each consisting of an empty milk tube and an empty blood tube. In addition, we prepared eight milk tube/blood tube pairs filled with 5 mL of distilled water at the sampling site as field operational blanks. All the blank samples were sent to Kyoto University and run through complete extraction, cleanup, and analysis procedures.

Serum extraction

The internal standard from mono- to deca-¹³C₁₂-PBDE and mono to deca-¹³C₁₂-PCB was spiked in the serum (3 g) and extracted by liquid–liquid extraction following the method of Takasuga et al. (2004, 2006). Briefly, in the serum spiked with internal standard, 3 mL ammonium sulfate, 1 mL ethanol, and 2 mL hexane were mixed and extracted twice. The final extract was washed with hexane-washed water, dehydrated with sodium sulfate, and concentrated to 5 mL for further cleanup.

Milk extraction

The internal standard from mono- through deca-¹³C₁₂-PBDE and mono- to deca-¹³C₁₂-PCB was spiked in the milk (3 g) and extracted by liquid–liquid extraction. Briefly, in the milk spiked with internal standard, 1 mL saturated potassium oxalate, 2 mL ethanol, 2 mL diethyl ether, and 1 mL hexane were mixed and extracted twice. The final extract was washed with 1 mL of 5% sodium chloride and then dehydrated with sodium sulfate and concentrated to 5 mL for further cleanup.

Cleanup of serum and milk

The 5-mL extract from serum or milk was subjected to multilayer Florisil silica gel column cleanup (Takasuga et al. 2004, 2006). The multilayer cleaned samples were further concentrated to the injection volume by nitrogen purge.

Identification and quantification of PBDEs and PCBs

We used high-resolution gas chromatography (HRGC; HP6890, Agilent)/high-resolution mass spectrometry (HRMS; Autospec Ultima; Micromass, Cary, NC, USA) for analysis of PBDEs and PCBs. Details on the HRGC/HRMS program are reported elsewhere (Takasuga et al. 2004, 2006). Briefly, for PBDE analysis we used either a BP-1 [15 m × 0.25 mm i.d. (0.1 μm); SGE Analytical Science Pty. Ltd., Austin, TX, USA] column or a ENV-5MS [15 m × 0.25 mm i.d. (0.1 μm)] column. The column was used with a temperature program of 120°C (1 min), increased 20°C/min to 160°C (0 min), 10°C/min to 260°C (0 min), and 20°C/min to 300°C (8 min). For analysis of PCBs, we used an HT-8 PCB column (60 m × 0.25 mm i.d.; SGE Analytical), which was used with an initial temperature of 150°C (0 min), increased 20°C/min to 200°C (0 min), 5°C/min to 260°C (0 min), and 10°C/min to 300°C (11.5 min). We used an on-column injection program with a 2-μL sample injection volume and with a resolution of M/ΔM > 10,000 (10% valley). We determined the individual and total concentrations of 13 PBDE congeners [ΣPBDE₁₃; International Union of Pure and Applied Chemistry (IUPAC) congeners 15, 28, 47, 99, 100, 153, 154, 183, 196, 197, 206, 207, and 209] and 15 PCB congeners (ΣPCB₁₅; IUPAC congeners 74, 99, 118, 138, 146, 153, 156, 163/164, 170, 180, 182/187, 194, 199, 206, and 209).

The limit of detection (LOD) for each PCB congener was 1 pg/g in both serum and breast milk. The LOD of each PBDE congener in serum and milk was between 0.2 and 2 pg/g for di-BDE to hepta-BDE and between 0.3 and 2 pg/g for octa-BDE to deca-BDE. The serum and milk concentrations of PCBs and PBDEs were expressed as nanograms per gram lipid. The lipid content in the serum samples was estimated from the total cholesterol and triglyceride concentrations (Phillips et al. 1989). The lipid content of the milk samples was determined from 2 mL crude extracts by gravimetric method.

Quality assurance and quality control

PBDE and PCB (native as well as ¹³C₁₂−labeled) standard solutions that contained the major congeners of mono-BDE or mono-CB to deca-BDE or deca-CB (> 95% pure) were purchased from Wellington Laboratories (Guelph, Ontario, Canada). The average recovery of individual PBDE congeners was 54–84% in serum (n = 100) and 54–103% in milk (n = 100), and the average recovery of PCB congeners was 61–79% in serum (n = 100) and 68–115% in milk (n = 100). The coefficient of variation for each determination was within 15% for both PBDEs and PCBs.

For all field blanks and field operational blanks, all PBDE and PCB congeners were < LOD. Operational blank tubes filled with 5 mL distilled water in an analytical laboratory (Shimadzu Techno-Research Inc., Kyoto, Japan) were also prepared for each eight-sample batch. These operational blanks were < LOD for all PBDE and PCB congeners in both the serum and milk batches. Thus, we did not correct the results for background levels.

Structure–activity relationship

For the QSAR analysis, we chose congeners that were detected in > 50% of both the serum and milk samples. Theoretical molecular descriptors for the compounds, which included constitutional descriptors, atom-centered fragments, and molecular properties, such as hydrophilicity, molar refractivity, polar surface area, and octanol/water partition coefficient (K_ow), were calculated using Dragon software (version 5.0; Milano Chemo Metrics and QSAR Research Group, Milan, Italy) and ADMET Predictor 1.2.3 (Simulations Plus, Lancaster, CA, USA). The K_ow calculated by Hansch’s method (CLogP) and the molar refractivity calculated by Hansch’s method (CMR) were calculated using Web applications provided by Daylight Chemical Information Systems (Aliso Viejo, CA, USA). Descriptors that had a bivariate correlation > 0.70 were removed.

We performed a stepwise multiple linear regression analysis using the SAS statistical package (version 8.2; SAS Institute Inc., Cary, NC, USA). All independent variables in the regressions had a significance of at least 95%, based on Student’s t-score.

Statistical analysis

Statistical analyses were conducted after logarithmic transformation of the concentrations of the PBDEs and PCBs. We tested differences between means by analysis of variance (ANOVA) or Student’s t-test when appropriate. A stepwise multiple regression analysis was used to explore determinants for the serum and milk levels of contaminants using a forward–backward stepwise regression model (F-statistic to enter and stay in the model with a p-value of < 0.25). We evaluated the determinants for PBDEs and PCBs in serum and breast milk using a conservative approach based on multiple comparisons of the questionnaire items. Thus, a p-value of < 0.01 was considered significant in the multiple regression analysis for the questionnaire items. For the other analyses, a p-value of < 0.05 was considered significant. All statistical analyses were carried out with SAS software.

Demographic features of the participants

On the whole, there were 20 participants from Hokkaido, 40 from Miyagi, 20 from Gifu, and 9 from Hyogo. The ages of the participants ranged from 20 to 43 years (mean ± SD, 30.1 ± 4.6 years). The results of the questionnaires are summarized in Table 1.

Determination of PBDEs and PCBs in serum and milk

The concentrations of some congeners in the human samples were < LOD. We treated these samples as 0 pg/g lipid when we calculated the total amount.

The distributions of ΣPBDE₁₃ in serum and milk followed log-normal distributions (Kolmogorov-Smirnov-Lilliefors test, p > 0.05). The geometric mean (GM) values for the total amounts of ΣPBDE₁₃ in the milk and serum samples were 1.56 and 2.89 ng/g lipid, respectively (Table 2). The PBDE congener levels and detection rates for milk and serum are available online (Supplemental Tables 2 and 3, respectively; http://www.ehponline.org/docs/2006/9032/suppl.pdf). BDE-209 was the predominant congener in serum and accounted for 38% of the total PBDEs but was a minor congener in milk and accounted for 8% of the ΣPBDE₁₃ (Figure 1A). In milk, BDE-47 and BDE-153 were the major congeners and accounted for 28 and 23% of the total PBDEs, respectively.

The distributions of the ΣPCB₁₅ in serum and milk also followed log-normal distributions (Kolmogorov-Smirnov-Lilliefors test, p > 0.05). The GM values for ΣPCB₁₅ in the milk and serum samples were 63.9 and 37.5 ng/g lipid, respectively (Table 2). The PCB congener levels and detection rates for milk and serum are available online (Supplemental Tables 4 and 5, respectively; http://www.ehponline.org/docs/2006/9032/suppl.pdf). CB-153, CB-138, and CB-180 were the major congeners in both milk and serum (30, 17, and 13% of the total for milk and 28, 16, and 15% of the total for serum, respectively) (Figure 1B).

It should be noted that approximately the same concentrations of the lighter PBDEs (e.g., BDE-47) are present in serum and milk, but BDE-209 is found at 10 times lower concentrations in milk than in serum (Supplemental Tables 2 and 3; available online at http://www.ehponline.org/docs/2006/9032/suppl.pdf). Likewise, almost double the serum concentration of CB-153 is found in milk, whereas more than double the milk concentration of CB-209 is found in serum (Supplemental Tables 4 and 5; available online at http://www.ehponline.org/docs/2006/9032/suppl.pdf).

Determinants for PCBs and PBDEs in serum and milk

We found significant correlations between ΣPCB₁₅ and ΣPBDE₁₃ levels in both milk and serum (r² = 0.194, p < 0.0001 for milk; r² = 0.1808, p < 0.0001 for serum). There were also significant geographic differences in ΣPBDE₁₃ concentrations in milk and serum (ANOVA, p = 0.00095 and p = 0.00030, respectively; Table 2). The GM for ΣPBDE₁₃ in the milk samples was higher for Hokkaido than for the other areas [Tukey’s honest significant difference (HSD) test, p < 0.05], whereas the GM for ΣPBDE₁₃ in serum samples was higher in Miyagi than in Gifu (Tukey’s HSD test, p < 0.05). The PCB levels also exhibited geographic differences (ANOVA, p = 0.0029 for milk and p < 0.0001 for serum; Table 2). The GMs for ΣPBDE₁₃ in both milk and serum samples were higher in Miyagi and Hyogo than in Gifu (Tukey’s HSD test, p < 0.05).

Multiple regression analysis revealed that the geographic factor was the primary determinant for the PBDE levels in both milk and serum (data not shown). In contrast, nursing duration was the significant determinant for PCB levels in both serum and milk (data not shown). To investigate the possible association between hospitals and nursing durations, we tested whether nursing duration was a determinant for PBDE or PCB levels within a single hospital. The nursing duration was correlated with the ΣPBDE₁₃ in serum in Miyagi (n = 38, Kendall’s τ= −0.266, p = 0.0187) and the ΣPCB₁₅ in both serum and milk in Miyagi (n = 38, Kendall’s τ= –0.426, p = 0.0002, and Kendall’s τ= –0.312, p = 0.0059, respectively; data not shown).

QSAR analysis

BDE-154, BDE-183, BDE-196, and BDE-206 were eliminated from the analysis because of their low detection rates in serum and/or milk (< 50%). In the first step, we calculated the mean ratios of milk concentrations (nanograms per gram lipid) to serum concentrations (nanograms per gram lipid) for individual congeners from milk and serum as surrogates for their partition coefficients (Supplemental Table 6; available online at http://www.ehponline.org/docs/2006/9032/suppl.pdf). Using these mean ratios, we then applied a multiple linear regression analysis using various descriptors for individual PCB and PBDE congeners. The descriptors that have been used for QSAR analysis include hydrophobicity [log K_ow, CLogP, (octanol/water partition coefficient calculated by Hansch’s method), and MLogP (octanol/water partition coefficient calculated by Moriguchi’s method)], size [MW (molecular weight) and MgVol (molar volume calculated by McGowan’s method)], polarizability [CMR, (molar refractivity calculated by Hansch’s method), AMR (calculated by Ghose and Crippen’s method), and PolarizG (polarizability calculated by Glen’s method)], and constitutional descriptors [TPSA (topologic polar surface area), HBA (number of hydrogen-bond acceptors), nCL (number of chlorines), and nBR (number of bromines)] (Table 3) (Abraham and McGowan 1987; Ghose and Crippen 1987; Glen 1994; Leo et al. 1971; Moriguchi et al. 1994).

Table 3 summarizes the correlation coefficients between pairs of the descriptors, together with regression coefficients for each descriptor. Regarding PCB and PBDE congeners, the descriptors for hydrophobicity (log K_ow, CLogP, and MLogP), molecular size (MW and MgVol), and polarizability (CMR, AMR, and PolarizG) were collinear, and each correlated well with the milk/serum partition coefficient (log P). We explored the combination of the descriptors that exhibited the highest multiple regression coefficient (r) and obtained the following equation:

Because partition coefficients have been reported to be dependent on the nursing period (LaKind et al. 2004), we tested the relationship between the predicted and observed mean partition coefficients for three nursing durations (Figure 2). For nursing durations ≤10 weeks, the partition coefficients predicted by the QSAR analysis agreed with the observed values. However, the coefficient of x was smaller for nursing durations > 10 weeks, suggesting that the prediction became weaker for longer nursing periods.