Research Articles
Predictor PCBs in Human Milk
Research Articles
Individual PCBs as Predictors for Concentrations of Non and Mono-Ortho PCBs in Human Milk':" 1Martin van den Berg, 1Theo L. Sinnige, 2Mats Tysklind, 1A.T.C. (Bart) Bosveld, 3Marcel Huisman, 4Corinne KoopmansEssenboom, Sjanna G. Koppe 1Research Institute of Toxicology, University Utrecht, P.O. Box 80.176, 3508 TD Utrecht, The Netherlands 2Institute of Environmental Chemistry, University of UmeCi, S-90187 Umefi, Sweden 3Academic Medical Hospital, Department of Obstetrics and Gynecology, P.O. Box 30001, 9700 RG Groningen, The Netherlands 4Sophia's Children's Hospital and Erasmus University, Department of Pediatrics, P.O. Box 70029, 3000 LL Rotterdam, The Netherlands 5Academic Medical Center, Department of Neonatology, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands
Corresponding author: Dr. Martin van den Berg
Abstract 32 Dutch human milk samples were analyzed for PCBs with either HRGC-ECD or HRGC-LRMS in the NCI mode. Samples were collected from three different locations in The Netherlands: Amsterdam, Rotterdam and Groningen. Quantitatively, no differences could be observed between the three localities, while in addition the congener specific pattern showed a striking similarity for all individual samples. Only principal component analysis revealed slight individual differences. Based on similarities in the PCB profiles, linear relationships were calculated between 2,3'4,4',5-PnCB (#118) or 2,2'4,4'5,5'HxCB (#153) and the most relevant non and mono-ortho PCBs exhibiting dioxinlike activity. These PCBs included 2,3,3',4,4'-PnCB (#105), 3,3',4,4'5-PnCB (#126) 2,3,3',4,4',5-HxCB (#156), 2,3,3',4,4',5';HxCB (#157), 2,3',4,4',5,5"-H• (#167) and 3,3',4,4',5'5-HxCB (#169). Good linear relationships were observed between individual PCBs. Based on the results of this study, PCB #118 can be used to predict concentrations of the PCBs #105 and #126. PCB #153 can be used as a predictor for the PCBs #156, #157, #167 and #169, but also for the total toxic equivalencies (TEQs) of non and mono-ortho PCBs
present in human milk. This method using certain PCBs as predictors for other toxicological relevant congeners, can be useful and cost effective, e.g. for epidemiological studies. However, before applied a number of conditions should be met. These are: 1) A stable composition of the PCB matrix should be established. 2) A possible time dependent change in composition of the matrix should first be excluded when used over different time periods.
1
Introduction
Chlorinated biphenyls (PCBs), dibenzo-p-dioxins (PCDDs) and dibenzofurans (PCDFs) are widespread contaminants in the global environment. PCBs have been produced commercially as late as the 1970's in many industrialized countries [1,2]. PCDDs and PCDFs have been formed as unwanted byproducts in industrial and combustion processes [3,4].
*) This research project was supported by The Dutch Ministry of Housing, Physical Planning and the Environment. Grant no. 361359.
ESPR-Environ. Sci. & Pollut. Res. 2 (2) 73-82 (1995) 9 ecomed publishers, D-86899 Landsberg, Germany
Due to their chlorine substitution and hydrophobic properties, these groups of compounds have a high persistency in the environment. These extreme physico-chemical properties lead to slow degradation in the environment resulting in biomagnification through the foodchain. As a result PCDDs, PCDFs and PCBs can be found at significant levels in species higher in the foodchain, humans included [ 4 - 11]. Although a significant number of PCDD, PCDF and PCB congeners has been released in the environment, uptake by living organisms and further transport through the food chain results in a significant change of the complex mixtures due to factors like molecular size and stereospecific requirements for metabolism [12,13]. Although 210 PCDD and PCDF congeners can be formed in different abiotic processes only a very limited number (17), those with a 2,3,7,8-substituted pattern, are usually found in the biotic samples. This biopersistency of 2,3,7,8-substituted PCDDs and PCDFs is a direct result of the presence of chlorine atoms on the four lateral positions, which block the most sensitive positions for metabolic breakdown towards more polar compounds [13]. For PCBs the structural requirements for biopersistency are more complex and governed both by the number and position of the chlorine atoms in the biphenyl molecule. In general, the absence of two vicinal hydrogen atoms and the presence of two adjacent chlorine atoms on the meta and para positions increase the resistance against metabolic breakdown in many species [12]. Due to these more complex structural requirements for facilitated metabolism and the high number of possible congeners (209), a more complex mixture accumulates in environmental biota. In general, 50 to 75 PCB congeners can be found in tissue samples from e.g. toppredators like fish-eating birds and mammals [5,6]. The mechanism of action of PCDDs, PCDFs and PCBs has been well studied during the last decades. It is generally assumed, that most of the biologic and toxic effects are mediated by a single receptor protein, the Ah-receptor, present in almost every species studied so far [14]. The structural requirements necessary for low capacity and high af-
73
Predictor PCBs in Human Milk
finity binding to this Ah-receptor are those of congeners with chlorines on the lateral positions. For PCDDs and PCDFs these include congeners with four chlorine substituents on the 2,3,7,8 positions. For PCBs those congeners with at least two adjacent chlorine atoms on the meta and para positions (3,4 or 5) act as Ah-receptor agonists. An increasing ortho substitution pattern decreases the Ah-receptor binding affinity and therefore reduces the "dioxin-like" activity of PCBs [15]. Based on this single receptor mediated mechanism, it has been postulated that for these compounds additivity of mixtures is a direct result. For PCDDs and PCDFs numerous in vivo and in vitro experiments have shown additivity for biologic and toxic effects [ 1 5 - 1 8 ] . For combinations of PCBs with PCDDs or PCDFs, antagonistic as well as synergistic effects have been reported [13 - 15]. However, the magnitude of the effect is minor compared to other uncertainties, e.g. species to species extrapolation, already present in the risk assessment. As a result of this prevalent additive effect, the toxic equivalency concept (TEQ) has been developed for risk assessment purposes. In this method each congener with Ah-receptor mediated activity is given a weighing factor relative to 2,3,7,8-TCDD, the biological most active compound. The congener specific weighing factors are called Toxic Equivalency Factors (TEFs) and can be used to calculate the total TEQs by simply adding up the products of congener specific concentrations and their TEF values [15]. During the last five years several studies concluded that non and mono-ortho PCBs with adjacent meta and para substitution patterns contribute significantly to the total TEQs of the biotic samples, including human milk. The PCBs contributing mostly to the total amount of dioxin- equivalents were 3,3',4,4'5-PnCB (#126), 2,3',4,4',5-PnCB (#118) and 2,3,3',4,4'-PnCB (#105) [19- 22]. Due to this significant contribution of these PCB congeners to the total TEQs of the mixture, the requirement for adequate chemical analysis of these PCBs became obvious. As a result several methods have been developed for sample clean-up and chemical analysis of especially the non-ortho PCBs [23 - 30]. Similar to PCDD and PCDF analysis, these methods proved to be costly and time consuming using advanced analytical methods like high resolution gas chromatography (HRGC) with or without mass-spectrometry (MS). GC or G C / M S analysis showed that complex mixtures of PCBs, PCDDs and PCDFs were remarkably stable in qualitative composition within one species in the same time period and area [20,22,25,31]. This stable mixture composition raises the question whether or not for PCBs some common and easy to measure congeners, e.g. 2,2'4,4'5,5'-HxCB (#153), can be used as marker compounds to predict the concentration of certain non and mono-ortho PCBs. Such a method could be cost-effective in e.g. epidemiological studies with populations in which the composition of the PCB mixture has been well established. Several groups have made an attempt for this approach but results are not conclusive [20,25]. In this study, we report on the PCB composition in Dutch human milk and the possibility to use certain congeners to predict either individual non and mono-ortho PCB concentrations or total PCB-TEQs.
74
Research Articles
2
Materials and M e t h o d s
Reagents Carbosphere activated carbon, 8 0 - 100 mesh with surface area 1000 m2/g was obtained from Alltech (Deerfield, USA) and pretreated as described earlier [24]. Basic alumina and Florisil were obtained from E.Merck (Darmstadt, FRG). Bio-beads SX-3 was purchased from Bio-Rad laboratories (Richmond, USA). 13C-labeled 3,3',4,4'- TCB (#77), 3,3',4,4'5PnCB (#126), 3,3',4,4',5,5'- HxCB (#169) and 12C-2,2'4,4',6,6'HxCB (#155) were obtained from Cambridge Isotope Laboratories (Woburn, M_A, USA) and used as internal standards for GC-ECD or GC/MS analysis. A reference m~nare of the PCBs (C.N. Schmidt BV, Amsterdam, The Netherlands) listed in Table 1 was used for congener specific identification on GC- ECD. All PCBs used had a purity > 99 %. All solvents and other reagents were of analytical grade.
Samples Individual human milk samples were collected from three different locations in The Netherlands. Samples were collected at The Academic Medical Center in Amsterdam (8), Academic Medical Hospital in Groningen (12) and Sophia's Children Hospital in Rotterdam (12). All samples were stored at -20 ~ prior to analysis.
Sample extraction Extraction of the human milk samples was done as described earlier by LIEM and coworkers [24]. Before extraction of 50 to 150 ml milk, 1 ml hexane containing 250 ng PCB #155 and 1 ml toluene containing 1.2 ng 13C-PCB #77, #126 and #169 were added to the sample. The percentage of lipid was determined gravimetrically. Clean-up of PCBs (except PCBs #126 and 169) 0.2 gram of milk fat was used for the column clean-up preceding the PCB analysis on GC-ECD. The clean-up procedure used for this PCB analysis was done as described earlier [6, 32] with the following slight modifications. A column (length 30 cm, i.d. 1 cm) was filled with 10 grams of 1.2 % deactivated Florisil, topped with 1 gram Na2SO4, was prewetted with 15 ml hexane. The fat sample was dissolved in approximately 1 ml hexane in a small beaker and transferred to the top of the column. The beaker was rinsed with 3 x I ml hexane also added to the top of the column. PCBs were eluated from the Florisil column with 46 ml hexane. The eluate was concentrated to approximately 1 ml under a gently flow of nitrogen and transferred to a GC vial. 50/al of n-nonane was added to the vial and the 9volume was further concentrated to less than 50/al, which was directly used for GC-ECD analysis. HRGC-ECD analysis The following standard PCBs, found to be major components in human milk, were detected and quantitated by GC-ECD analysis: 2,4,4'-TrCB (#28), 2,2',5,5'-TCB (#52), 2,2',4,5,5'PnCB (#101), 2,2',3,4,4',5'-HxCB (#138), 2,2',4,4',5,5'-
E S P R - E n v i r o n . Sci. & Pollut. Res. 2 (2) 1995
Research Articles
HxCB (#153), 2,2',3,4,4'5,5'-HpCB (#180). From the penta- and hexachlorinated mono-ortho PCBs, the following congeners were detected in the milk samples: 2,3,3',4,4'PnCB (#105), 2,3'4,4',5-PnCB (#118), 2,3,3',4,4',5-HxCB (#156), 2,3,3',4,4',5'-HxCB(#157), 2,3',4,4',5,5'-HxCB (#167). The presence of 2',3,4,4'5-PnCB (#123) could not fully be excluded as under these GC conditions this congener might coelute with PCB #118. PCB analyses were done on a Carlo Erba 5360 Mega Series Gaschromatograph fitted with an electron capture detector using a 2 m retention gap and a 60 m J & W DB5 fused silica column (i.d. 0.25 mm; stationary phase 0.1 /am). 2/al of a sample was injected splitless at a temperature of 300 ~ Helium gas was used as a carrier at 200 Kpa, while the makeup gas was a mixture of argon : methane = 90 : 10 at 150 Kpa. The column temperature program w a s : T 1 = 130 ~ 2 min; Rate1 = 5 ~ to 180 ~ Rate2 - 1 ~ to 250 ~ Rate 3 = 10 ~ to 275 ~ 10 minutes at 275 ~ The detector temperature was set at 365 ~ Detection limit was approximately 4 pg/injection, equivalent with approximately 500 p g / g lipid, depending on the congener. Analytical recoveries for the above PCB congeners ranged from 94 to 104 + 4 % relative to the internal standard PCB #155. PCB #155 could not be detected in milk samples without addition of this internal standard.
Clean-up of the non-ortho PCBs #77, #126 and #169 Basically the Carbosphere carbon clean-up procedure was used to separate the non-ortho PCBs as described earlier [24,32]. However, a number of modifications were adapted to reduce interfering residue components during H R G C / M S analysis. The remaining milk fat sample was dissolved in 10 ml dichloromethane and added on top of the Carbosphere column. The beaker was rinsed with an additional 2 x 10 ml dichloromethane, also added to the column. The Carbosphere column was refluxed for 1 hour with 25 ml dichloromethane and this fraction was discarded. Directly afterwards, the column was rinsed with 20 ml toluene and followed by a 2.5 hour reflux with 25 ml toluene. This fraction contained the non-ortho PCBs. The toluene eluate was carefully evaporated to dryness under a gentle flow of nitrogen. The residue was again dissolved in 5 ml hexane and added on top of a column (length 30 cm, i.d. 1 cm) containing 5 gram basic alumina and 1 gram anhydrous Na2SO 4. The beaker was rinsed with 2 x 2.5 ml hexane also added to the Alumina column. The non-ortho PCBs were eluted from the column with 50 ml hexane/dichloromethane (1:1 v/v). This eluate was slowly concentrated under nitrogen to approximately 0.5 ml. Initially this fraction was used directly for G C / M S analysis but it was found that still too many interfering components (either using El + or NCI mode) were present. Consequently, an additional clean-up step, involving gel permeation chromatography with Bio-beads SX-3 was introduced. 3 grams of Bio-beads SX-3 were swollen in an excess of a hexane/dichloromethane (1:1 v/v) mixture. The Bio-beads were slurried in a glass column (length 30 cm, i.d. 1 cm) equipped with a stopcock and plugged with glasswool. The Bio-beads were given at least an hour to sink and settle after which the Carbosphere fraction (concentrated to approximately 0.5 ml)
ESPR-Environ. Sci. & Pollut. Res. 2 (2) 1995
Predictor PCBs in Human Milk
was added on top of the Bio-beads column. The beaker was again rinsed with 2 x 0.5 ml hexane/dichloro- methane (1:I v/v). The column was then eluted with 9 ml hexane/ dichloromethane (1:1 v/v) and this fraction was discarded. The non-ortho PCBs were eluted with another 10 ml hexane/ dichloromethane (1:1 v/v). This eluate was carefully evaporated to dryness and redissolved in 5/d n-nonane. This fraction could directly be used for G C / M S analysis.
GC/MS analysis of non-ortho PCBs Residue analysis of the non-ortho PCBs was performed on a QMD 1000 H R G C / L R M S Quadrupole operating with NCI detection. In the selected ion mode the two most abundant ions of the chlorine isotope cluster of each molecular ion (M +2 or M +4) were scanned. A 60 m J & W DB5 fused silica column (i.d. 0.25 mm; stationary phase 0.1/~m) and 2 m retention gap was used. 2/al of a sample was injected on-column. The carrier gas was helium at 250 Kpa. The column temperature program was: T 1 = 160 ~ 1 min; Rate 30~ to 250~ for 20 min. Ionisation energy was 70 eV. Identification of the non-ortho PCBs was based on the retention time of the 13C-labeled internal standards and comparison of the isotope ratios with theoretical values. Detection of PCB #77 proved to be highly unsensitive when using NCI mode and omitted from further analysis. In contrast, the detection of the PCBs #126 and #169 was largely improved with NCI compared to the EI § mode. Detection limits for PCBs #126 and #169 were approximately 20 and 5 p g / g lipid.
3
3.1
Results Concentrations
In Table 1 the average PCB concentrations in the Dutch human milk samples are listed. In the appendices (--' pp. 81, 82) 1,2 and 3 the PCB concentrations of the individual milk samples are given. The PCBs #153, #138, #118 and #180 were the most common congeners present with average concentrations ranging between 30 and 140 ng/g lipid. The PCB congeners #28, #52, #101, #105, #167, #157 were present between I and 10 ng/g lipid. PCB #156 was present in slightly higher concentrations being approximately 15 ng/g lipid. The PCBs #126 and #169 were present at much lower levels than the congeners mentioned above. Concentrations of the PCBs #126 and #169 were approximately 90 and 60 p g / g lipid respectively. When average concentrations for the above congeners were calculated for the three different locations, no statistically significant differences were observed. In addition, no statistical difference could be observed in lipid content between the three locations ( ~ See Table 1).
3.2
PCB mixture composition
All PCB chromatograms of the individual milk samples showed a striking qualitative similarity for congeners having five or more chlorine atoms. Consequently, the ratios between different congeners were determined with linear regression. Calculations were performed using the formula [PCBx] Xcoeff" * [PCBMarker ] + Constant ( ~ Table 2).
75
Predictor PCBs in Human Milk
Research Articles
Table 1: AveragePCB concentrations in ng/g lipid in human milk from three differentlocations in the Netherlands (PCB#126 and #169 in pg/g lipid) 3 Locations
F~CB C,ongeners
s.d., n = 32
Average
Average
s.d., n -- 12
s.d., n = 1 2
Average
7.6
6.2
8.0
2.8
5.5
2.0
9.5
9.2
#r
5.8
2.4
6.4
2.4
5.4
2.0
5.7
2.6
#101
2.1
1.1
2.7
1.8
1.8
0.6
1.9
0.6
#118
30.5
12.5
32.2
9.5
29.1
8.4
30.8
16.8
#153
127.8
50.3
139.3
67.0
123.5
31.6
124.6
51.4
#105
7.4
2.8
7.7
1.8
7.1
2.2
7.4
3.7
#138
101.5
41.4
116.2
53.6
95.5
24.6
97.9
43.2
#167
4.9
2.2
5.4
2.6
4.6
1.4
4.9
2.5
#156
15.8
6.1
16.4
8.0
15.6
4.1
15.5
6.2
#157
2.7
1.2
3.0
1.5
2.7
0.8
2.7
1.2
#180
70.2
27.3
75.6
37.4
69.0
20.5
67.8
24.5
#126
95
45
93.5
49.6
89.8
29.3
102.2
52.9
#169
61
28
58.2
35.8
63.1
23.2
61.8
26.4
lipid
2.8
0.7
2.8
1.1
2.4
0.5
3.0
0.5
Table 2: Correlation coefficients, regression line parameters and significance levels for the relationships between PCB 153 and 118 with the most significant non and mono-ortbo PCBs in Dutch human milk PCB Y
R-value Constant
1531
105
0.794
153
118
153
126
153~
X-Coeff. number p-value
t-value
1.70
0.044
32
Predictor PCBs in Human Milk
Research Articles
Individual PCBs as Predictors for Concentrations of Non and Mono-Ortho PCBs in Human Milk':" 1Martin van den Berg, 1Theo L. Sinnige, 2Mats Tysklind, 1A.T.C. (Bart) Bosveld, 3Marcel Huisman, 4Corinne KoopmansEssenboom, Sjanna G. Koppe 1Research Institute of Toxicology, University Utrecht, P.O. Box 80.176, 3508 TD Utrecht, The Netherlands 2Institute of Environmental Chemistry, University of UmeCi, S-90187 Umefi, Sweden 3Academic Medical Hospital, Department of Obstetrics and Gynecology, P.O. Box 30001, 9700 RG Groningen, The Netherlands 4Sophia's Children's Hospital and Erasmus University, Department of Pediatrics, P.O. Box 70029, 3000 LL Rotterdam, The Netherlands 5Academic Medical Center, Department of Neonatology, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands
Corresponding author: Dr. Martin van den Berg
Abstract 32 Dutch human milk samples were analyzed for PCBs with either HRGC-ECD or HRGC-LRMS in the NCI mode. Samples were collected from three different locations in The Netherlands: Amsterdam, Rotterdam and Groningen. Quantitatively, no differences could be observed between the three localities, while in addition the congener specific pattern showed a striking similarity for all individual samples. Only principal component analysis revealed slight individual differences. Based on similarities in the PCB profiles, linear relationships were calculated between 2,3'4,4',5-PnCB (#118) or 2,2'4,4'5,5'HxCB (#153) and the most relevant non and mono-ortho PCBs exhibiting dioxinlike activity. These PCBs included 2,3,3',4,4'-PnCB (#105), 3,3',4,4'5-PnCB (#126) 2,3,3',4,4',5-HxCB (#156), 2,3,3',4,4',5';HxCB (#157), 2,3',4,4',5,5"-H• (#167) and 3,3',4,4',5'5-HxCB (#169). Good linear relationships were observed between individual PCBs. Based on the results of this study, PCB #118 can be used to predict concentrations of the PCBs #105 and #126. PCB #153 can be used as a predictor for the PCBs #156, #157, #167 and #169, but also for the total toxic equivalencies (TEQs) of non and mono-ortho PCBs
present in human milk. This method using certain PCBs as predictors for other toxicological relevant congeners, can be useful and cost effective, e.g. for epidemiological studies. However, before applied a number of conditions should be met. These are: 1) A stable composition of the PCB matrix should be established. 2) A possible time dependent change in composition of the matrix should first be excluded when used over different time periods.
1
Introduction
Chlorinated biphenyls (PCBs), dibenzo-p-dioxins (PCDDs) and dibenzofurans (PCDFs) are widespread contaminants in the global environment. PCBs have been produced commercially as late as the 1970's in many industrialized countries [1,2]. PCDDs and PCDFs have been formed as unwanted byproducts in industrial and combustion processes [3,4].
*) This research project was supported by The Dutch Ministry of Housing, Physical Planning and the Environment. Grant no. 361359.
ESPR-Environ. Sci. & Pollut. Res. 2 (2) 73-82 (1995) 9 ecomed publishers, D-86899 Landsberg, Germany
Due to their chlorine substitution and hydrophobic properties, these groups of compounds have a high persistency in the environment. These extreme physico-chemical properties lead to slow degradation in the environment resulting in biomagnification through the foodchain. As a result PCDDs, PCDFs and PCBs can be found at significant levels in species higher in the foodchain, humans included [ 4 - 11]. Although a significant number of PCDD, PCDF and PCB congeners has been released in the environment, uptake by living organisms and further transport through the food chain results in a significant change of the complex mixtures due to factors like molecular size and stereospecific requirements for metabolism [12,13]. Although 210 PCDD and PCDF congeners can be formed in different abiotic processes only a very limited number (17), those with a 2,3,7,8-substituted pattern, are usually found in the biotic samples. This biopersistency of 2,3,7,8-substituted PCDDs and PCDFs is a direct result of the presence of chlorine atoms on the four lateral positions, which block the most sensitive positions for metabolic breakdown towards more polar compounds [13]. For PCBs the structural requirements for biopersistency are more complex and governed both by the number and position of the chlorine atoms in the biphenyl molecule. In general, the absence of two vicinal hydrogen atoms and the presence of two adjacent chlorine atoms on the meta and para positions increase the resistance against metabolic breakdown in many species [12]. Due to these more complex structural requirements for facilitated metabolism and the high number of possible congeners (209), a more complex mixture accumulates in environmental biota. In general, 50 to 75 PCB congeners can be found in tissue samples from e.g. toppredators like fish-eating birds and mammals [5,6]. The mechanism of action of PCDDs, PCDFs and PCBs has been well studied during the last decades. It is generally assumed, that most of the biologic and toxic effects are mediated by a single receptor protein, the Ah-receptor, present in almost every species studied so far [14]. The structural requirements necessary for low capacity and high af-
73
Predictor PCBs in Human Milk
finity binding to this Ah-receptor are those of congeners with chlorines on the lateral positions. For PCDDs and PCDFs these include congeners with four chlorine substituents on the 2,3,7,8 positions. For PCBs those congeners with at least two adjacent chlorine atoms on the meta and para positions (3,4 or 5) act as Ah-receptor agonists. An increasing ortho substitution pattern decreases the Ah-receptor binding affinity and therefore reduces the "dioxin-like" activity of PCBs [15]. Based on this single receptor mediated mechanism, it has been postulated that for these compounds additivity of mixtures is a direct result. For PCDDs and PCDFs numerous in vivo and in vitro experiments have shown additivity for biologic and toxic effects [ 1 5 - 1 8 ] . For combinations of PCBs with PCDDs or PCDFs, antagonistic as well as synergistic effects have been reported [13 - 15]. However, the magnitude of the effect is minor compared to other uncertainties, e.g. species to species extrapolation, already present in the risk assessment. As a result of this prevalent additive effect, the toxic equivalency concept (TEQ) has been developed for risk assessment purposes. In this method each congener with Ah-receptor mediated activity is given a weighing factor relative to 2,3,7,8-TCDD, the biological most active compound. The congener specific weighing factors are called Toxic Equivalency Factors (TEFs) and can be used to calculate the total TEQs by simply adding up the products of congener specific concentrations and their TEF values [15]. During the last five years several studies concluded that non and mono-ortho PCBs with adjacent meta and para substitution patterns contribute significantly to the total TEQs of the biotic samples, including human milk. The PCBs contributing mostly to the total amount of dioxin- equivalents were 3,3',4,4'5-PnCB (#126), 2,3',4,4',5-PnCB (#118) and 2,3,3',4,4'-PnCB (#105) [19- 22]. Due to this significant contribution of these PCB congeners to the total TEQs of the mixture, the requirement for adequate chemical analysis of these PCBs became obvious. As a result several methods have been developed for sample clean-up and chemical analysis of especially the non-ortho PCBs [23 - 30]. Similar to PCDD and PCDF analysis, these methods proved to be costly and time consuming using advanced analytical methods like high resolution gas chromatography (HRGC) with or without mass-spectrometry (MS). GC or G C / M S analysis showed that complex mixtures of PCBs, PCDDs and PCDFs were remarkably stable in qualitative composition within one species in the same time period and area [20,22,25,31]. This stable mixture composition raises the question whether or not for PCBs some common and easy to measure congeners, e.g. 2,2'4,4'5,5'-HxCB (#153), can be used as marker compounds to predict the concentration of certain non and mono-ortho PCBs. Such a method could be cost-effective in e.g. epidemiological studies with populations in which the composition of the PCB mixture has been well established. Several groups have made an attempt for this approach but results are not conclusive [20,25]. In this study, we report on the PCB composition in Dutch human milk and the possibility to use certain congeners to predict either individual non and mono-ortho PCB concentrations or total PCB-TEQs.
74
Research Articles
2
Materials and M e t h o d s
Reagents Carbosphere activated carbon, 8 0 - 100 mesh with surface area 1000 m2/g was obtained from Alltech (Deerfield, USA) and pretreated as described earlier [24]. Basic alumina and Florisil were obtained from E.Merck (Darmstadt, FRG). Bio-beads SX-3 was purchased from Bio-Rad laboratories (Richmond, USA). 13C-labeled 3,3',4,4'- TCB (#77), 3,3',4,4'5PnCB (#126), 3,3',4,4',5,5'- HxCB (#169) and 12C-2,2'4,4',6,6'HxCB (#155) were obtained from Cambridge Isotope Laboratories (Woburn, M_A, USA) and used as internal standards for GC-ECD or GC/MS analysis. A reference m~nare of the PCBs (C.N. Schmidt BV, Amsterdam, The Netherlands) listed in Table 1 was used for congener specific identification on GC- ECD. All PCBs used had a purity > 99 %. All solvents and other reagents were of analytical grade.
Samples Individual human milk samples were collected from three different locations in The Netherlands. Samples were collected at The Academic Medical Center in Amsterdam (8), Academic Medical Hospital in Groningen (12) and Sophia's Children Hospital in Rotterdam (12). All samples were stored at -20 ~ prior to analysis.
Sample extraction Extraction of the human milk samples was done as described earlier by LIEM and coworkers [24]. Before extraction of 50 to 150 ml milk, 1 ml hexane containing 250 ng PCB #155 and 1 ml toluene containing 1.2 ng 13C-PCB #77, #126 and #169 were added to the sample. The percentage of lipid was determined gravimetrically. Clean-up of PCBs (except PCBs #126 and 169) 0.2 gram of milk fat was used for the column clean-up preceding the PCB analysis on GC-ECD. The clean-up procedure used for this PCB analysis was done as described earlier [6, 32] with the following slight modifications. A column (length 30 cm, i.d. 1 cm) was filled with 10 grams of 1.2 % deactivated Florisil, topped with 1 gram Na2SO4, was prewetted with 15 ml hexane. The fat sample was dissolved in approximately 1 ml hexane in a small beaker and transferred to the top of the column. The beaker was rinsed with 3 x I ml hexane also added to the top of the column. PCBs were eluated from the Florisil column with 46 ml hexane. The eluate was concentrated to approximately 1 ml under a gently flow of nitrogen and transferred to a GC vial. 50/al of n-nonane was added to the vial and the 9volume was further concentrated to less than 50/al, which was directly used for GC-ECD analysis. HRGC-ECD analysis The following standard PCBs, found to be major components in human milk, were detected and quantitated by GC-ECD analysis: 2,4,4'-TrCB (#28), 2,2',5,5'-TCB (#52), 2,2',4,5,5'PnCB (#101), 2,2',3,4,4',5'-HxCB (#138), 2,2',4,4',5,5'-
E S P R - E n v i r o n . Sci. & Pollut. Res. 2 (2) 1995
Research Articles
HxCB (#153), 2,2',3,4,4'5,5'-HpCB (#180). From the penta- and hexachlorinated mono-ortho PCBs, the following congeners were detected in the milk samples: 2,3,3',4,4'PnCB (#105), 2,3'4,4',5-PnCB (#118), 2,3,3',4,4',5-HxCB (#156), 2,3,3',4,4',5'-HxCB(#157), 2,3',4,4',5,5'-HxCB (#167). The presence of 2',3,4,4'5-PnCB (#123) could not fully be excluded as under these GC conditions this congener might coelute with PCB #118. PCB analyses were done on a Carlo Erba 5360 Mega Series Gaschromatograph fitted with an electron capture detector using a 2 m retention gap and a 60 m J & W DB5 fused silica column (i.d. 0.25 mm; stationary phase 0.1 /am). 2/al of a sample was injected splitless at a temperature of 300 ~ Helium gas was used as a carrier at 200 Kpa, while the makeup gas was a mixture of argon : methane = 90 : 10 at 150 Kpa. The column temperature program w a s : T 1 = 130 ~ 2 min; Rate1 = 5 ~ to 180 ~ Rate2 - 1 ~ to 250 ~ Rate 3 = 10 ~ to 275 ~ 10 minutes at 275 ~ The detector temperature was set at 365 ~ Detection limit was approximately 4 pg/injection, equivalent with approximately 500 p g / g lipid, depending on the congener. Analytical recoveries for the above PCB congeners ranged from 94 to 104 + 4 % relative to the internal standard PCB #155. PCB #155 could not be detected in milk samples without addition of this internal standard.
Clean-up of the non-ortho PCBs #77, #126 and #169 Basically the Carbosphere carbon clean-up procedure was used to separate the non-ortho PCBs as described earlier [24,32]. However, a number of modifications were adapted to reduce interfering residue components during H R G C / M S analysis. The remaining milk fat sample was dissolved in 10 ml dichloromethane and added on top of the Carbosphere column. The beaker was rinsed with an additional 2 x 10 ml dichloromethane, also added to the column. The Carbosphere column was refluxed for 1 hour with 25 ml dichloromethane and this fraction was discarded. Directly afterwards, the column was rinsed with 20 ml toluene and followed by a 2.5 hour reflux with 25 ml toluene. This fraction contained the non-ortho PCBs. The toluene eluate was carefully evaporated to dryness under a gentle flow of nitrogen. The residue was again dissolved in 5 ml hexane and added on top of a column (length 30 cm, i.d. 1 cm) containing 5 gram basic alumina and 1 gram anhydrous Na2SO 4. The beaker was rinsed with 2 x 2.5 ml hexane also added to the Alumina column. The non-ortho PCBs were eluted from the column with 50 ml hexane/dichloromethane (1:1 v/v). This eluate was slowly concentrated under nitrogen to approximately 0.5 ml. Initially this fraction was used directly for G C / M S analysis but it was found that still too many interfering components (either using El + or NCI mode) were present. Consequently, an additional clean-up step, involving gel permeation chromatography with Bio-beads SX-3 was introduced. 3 grams of Bio-beads SX-3 were swollen in an excess of a hexane/dichloromethane (1:1 v/v) mixture. The Bio-beads were slurried in a glass column (length 30 cm, i.d. 1 cm) equipped with a stopcock and plugged with glasswool. The Bio-beads were given at least an hour to sink and settle after which the Carbosphere fraction (concentrated to approximately 0.5 ml)
ESPR-Environ. Sci. & Pollut. Res. 2 (2) 1995
Predictor PCBs in Human Milk
was added on top of the Bio-beads column. The beaker was again rinsed with 2 x 0.5 ml hexane/dichloro- methane (1:I v/v). The column was then eluted with 9 ml hexane/ dichloromethane (1:1 v/v) and this fraction was discarded. The non-ortho PCBs were eluted with another 10 ml hexane/ dichloromethane (1:1 v/v). This eluate was carefully evaporated to dryness and redissolved in 5/d n-nonane. This fraction could directly be used for G C / M S analysis.
GC/MS analysis of non-ortho PCBs Residue analysis of the non-ortho PCBs was performed on a QMD 1000 H R G C / L R M S Quadrupole operating with NCI detection. In the selected ion mode the two most abundant ions of the chlorine isotope cluster of each molecular ion (M +2 or M +4) were scanned. A 60 m J & W DB5 fused silica column (i.d. 0.25 mm; stationary phase 0.1/~m) and 2 m retention gap was used. 2/al of a sample was injected on-column. The carrier gas was helium at 250 Kpa. The column temperature program was: T 1 = 160 ~ 1 min; Rate 30~ to 250~ for 20 min. Ionisation energy was 70 eV. Identification of the non-ortho PCBs was based on the retention time of the 13C-labeled internal standards and comparison of the isotope ratios with theoretical values. Detection of PCB #77 proved to be highly unsensitive when using NCI mode and omitted from further analysis. In contrast, the detection of the PCBs #126 and #169 was largely improved with NCI compared to the EI § mode. Detection limits for PCBs #126 and #169 were approximately 20 and 5 p g / g lipid.
3
3.1
Results Concentrations
In Table 1 the average PCB concentrations in the Dutch human milk samples are listed. In the appendices (--' pp. 81, 82) 1,2 and 3 the PCB concentrations of the individual milk samples are given. The PCBs #153, #138, #118 and #180 were the most common congeners present with average concentrations ranging between 30 and 140 ng/g lipid. The PCB congeners #28, #52, #101, #105, #167, #157 were present between I and 10 ng/g lipid. PCB #156 was present in slightly higher concentrations being approximately 15 ng/g lipid. The PCBs #126 and #169 were present at much lower levels than the congeners mentioned above. Concentrations of the PCBs #126 and #169 were approximately 90 and 60 p g / g lipid respectively. When average concentrations for the above congeners were calculated for the three different locations, no statistically significant differences were observed. In addition, no statistical difference could be observed in lipid content between the three locations ( ~ See Table 1).
3.2
PCB mixture composition
All PCB chromatograms of the individual milk samples showed a striking qualitative similarity for congeners having five or more chlorine atoms. Consequently, the ratios between different congeners were determined with linear regression. Calculations were performed using the formula [PCBx] Xcoeff" * [PCBMarker ] + Constant ( ~ Table 2).
75
Predictor PCBs in Human Milk
Research Articles
Table 1: AveragePCB concentrations in ng/g lipid in human milk from three differentlocations in the Netherlands (PCB#126 and #169 in pg/g lipid) 3 Locations
F~CB C,ongeners
s.d., n = 32
Average
Average
s.d., n -- 12
s.d., n = 1 2
Average
7.6
6.2
8.0
2.8
5.5
2.0
9.5
9.2
#r
5.8
2.4
6.4
2.4
5.4
2.0
5.7
2.6
#101
2.1
1.1
2.7
1.8
1.8
0.6
1.9
0.6
#118
30.5
12.5
32.2
9.5
29.1
8.4
30.8
16.8
#153
127.8
50.3
139.3
67.0
123.5
31.6
124.6
51.4
#105
7.4
2.8
7.7
1.8
7.1
2.2
7.4
3.7
#138
101.5
41.4
116.2
53.6
95.5
24.6
97.9
43.2
#167
4.9
2.2
5.4
2.6
4.6
1.4
4.9
2.5
#156
15.8
6.1
16.4
8.0
15.6
4.1
15.5
6.2
#157
2.7
1.2
3.0
1.5
2.7
0.8
2.7
1.2
#180
70.2
27.3
75.6
37.4
69.0
20.5
67.8
24.5
#126
95
45
93.5
49.6
89.8
29.3
102.2
52.9
#169
61
28
58.2
35.8
63.1
23.2
61.8
26.4
lipid
2.8
0.7
2.8
1.1
2.4
0.5
3.0
0.5
Table 2: Correlation coefficients, regression line parameters and significance levels for the relationships between PCB 153 and 118 with the most significant non and mono-ortbo PCBs in Dutch human milk PCB Y
R-value Constant
1531
105
0.794
153
118
153
126
153~
X-Coeff. number p-value
t-value
1.70
0.044
32
