Maternal and early postnatal polychlorinated biphenyl exposure in relation to total serum immunoglobulin concentrations in 6-month-old infants

Abstract

Animal data indicate that developmental tetrachlorodibenzo-p-dioxin exposure alters immune function; however, the potential immunotoxicity of dioxin-like and non-dioxin-like polychlorinated biphenyls (PCBs) in the developing infant is an understudied area. The aim of the current study is to examine the association between maternal and early postnatal PCB concentrations in relation to total infant serum immunoglobulin concentrations determined at 6-months-of-age. We selected 384 mother-infant pairs participating in a birth cohort study in Eastern Slovakia. PCB concentrations of several congeners were determined in maternal and cord serum samples and in infant serum samples collected at 6-months-of-age using gas chromatography with electron capture detection. Total immunoglobulin (Ig) G, A, and M concentrations were determined by nephelometry, and IgE concentrations were determined by enzyme-linked immunoassay. Linear regression models with adjustment for potential confounding factors were used to estimate the associations between maternal, cord, and 6-month infant PCB concentrations and total serum immunoglobulins. The median maternal serum concentration of PCB-153 was 140 ng/g lipid, ≈10-fold higher than concentrations in childbearing-age women in the United States during the same period. Maternal, cord, or 6-month infant PCB concentrations were not associated with total serum immunoglobulin levels at 6 months, regardless of the timing of PCB exposure, PCB congener, or specific immunoglobulin. In this population, which has high PCB concentrations relative to most populations in the world today, we did not observe any association between maternal and early postnatal PCB concentrations and total immunoglobulin measures of IgG, IgA, IgM, or IgE.

Keywords::

• Epidemiology
• cohort
• DAG
• roma

Introduction

Polychlorinated biphenyls (PCBs) are lipophilic compounds that do not easily degrade in the environment and, once in the body, are not readily excreted (Grandjean et al., 2008; Verner et al., 2009). While diet is the primary source of exposure for most adults, PCB body burdens during infancy are the result of in utero and lactational exposure. Indeed, models of lifetime PCB exposure indicate that serum PCB concentrations are greatest during the time period prior to weaning (Verner et al., 2008).

Some evidence indicates that prenatal and early postnatal exposure to PCBs may alter immune function in the infant and child. For example, higher maternal and/or postnatal PCB concentrations have been associated with reduced thymus size in infants (Park et al., 2008), increased risk for respiratory and middle ear infections (Weisglas-Kuperus et al., 2000, 2004; Dallaire et al., 2004, 2006; Glynn et al., 2008), altered immune cell counts and phenotypes (Weisglas-Kuperus et al., 2000, 2004; Glynn et al., 2008), and reduced antibody responses to childhood vaccines (Weisglas-Kuperus et al., 2000; Heilmann, 2006, 2010). Although functional measures such as T-cell-dependent antibody response have been considered the “gold standard” for developmental immunotoxicity testing (Luster et al., 2005; Dietert and Holsapple, 2007), some studies have considered other immune outcomes, such as total serum immunoglobulin (Ig) concentrations as measures of potential immunotoxicity, since these measures appear to predict morbidity in later childhood. For instance, IgE concentrations measured during early postnatal life are associated with future atopy and respiratory morbidity (Martinez et al., 1995; Sherrill et al., 1999; Klinnert et al., 2001), and results from studies of children with autism suggest that IgG, IgM, and IgA concentrations may be different than those without autism (Gupta et al., 1996; Croonenberghs et al., 2002; Ashwood et al., 2003; Trajkovski et al., 2004; Dietert and Dietert, 2008; Heuer et al., 2008). Thus, early immune perturbations may serve as valuable biomarkers of later morbidity, when clinical endpoints-which often require lengthy and/or detailed follow-up-are not available. However, few data exist concerning developmental exposure to PCBs and immunoglobulin concentrations during early infancy.

We therefore examined PCB concentrations determined in maternal, cord, and 6-month infant serum in relation to total IgG, IgA, IgM, and IgE measured in infants at 6-months-of-age. Participants in our study reside in an area with significant environmental contamination from PCBs (Kocan et al., 2001) and therefore have high (and varied) PCB body burdens compared to most contemporary populations (Kocan et al., 1994; Longnecker et al., 2003), making these individuals ideally suited for studying health outcomes associated with PCB exposure.

Methods

Sample selection, specimen collection, and follow-up

The details of our sample selection and data collection procedures have been published previously (Jusko et al., 2010). Briefly, women living in two districts in Eastern Slovakia between 2002 and 2004 were approached at the time they visited their local hospital in each district to deliver their child. In total, 1134 mother-infant pairs were enrolled. Following the informed consent process, women were asked to donate two 9 mL serum tubes of blood. After delivery of the infant, cord blood was also collected.

When children were ≈6-months-of-age, the families were invited back to their local hospital’s Department of Pediatrics for a follow-up examination. At that time, ≈9 mL of blood was collected from the infant for subsequent organochlorine, lipid, and immunochemistry analyses. Of the original cohort of 1134, 971 (86%) mother-infant pairs still remained in the study at the 6-month follow-up. The Institutional Review Boards at the University of California at Davis and the Slovak Medical University each approved the study protocol.

PCB and lipid measurement

Concentrations of 15 individual PCB congeners were determined in maternal, cord, and 6-month infant serum samples by the Department of Toxic Organic Pollutants at the Slovak Medical University in Bratislava. The PCB congeners determined were International Union of Pure and Applied Chemistry numbers 28, 52, 101, 105, 114, 118, 123⁺¹⁴⁹, 138⁺¹⁶³, 153, 156⁺¹⁷¹, 157, 167, 170, 180, and 189. The procedure for organochlorine determination in serum samples involved extraction, cleanup, and quantitation by high-resolution gas chromatography with electron capture detection (Conka et al., 2005). Maternal, cord, and 6-month infant total serum lipid concentrations were determined using the formulas of Akins et al. (1989) and Takayama et al. (1977).

Individual congeners were included in our statistical evaluation if fewer than 20% of samples were below the limit of detection (LOD) for that congener. This resulted in the selection of six maternal PCB congeners (PCB nos. 118, 138⁺¹⁶³, 153, 156⁺¹⁷¹, 170, and 180), and four congeners in both cord and 6-month infant serum samples (PCB nos. 138⁺¹⁶³, 153, 170, and 180). When an individual value was less than LOD, we assigned the value as the LOD divided by the square root of 2 (Lubin et al., 2004). For maternal, cord, and 6-month PCB concentrations, the corresponding “sum” variable is the arithmetic sum of the above 6, 4, and 4 congeners, respectively.

Determination of immunoglobulin concentrations

Immunologic assays were performed on 40% of the 971 infants (N = 384) still enrolled in the study at 6 months. This number was based on power calculations that incorporated the estimated partial correlation coefficients between PCB and antibody concentrations from previous studies (Weisglas-Kuperus et al., 2000; Heilmann et al., 2006). The 384 infants were selected using stratified, random sampling as presented previously (Jusko et al. 2010). At the analysis stage, normalized weights were applied to all regression models to account for this sampling design (described below) so that the estimates could be interpreted as if the 6-month subsample were drawn at random from the overall cohort at 6 months.

From the infant specimens collected at the 6-month visit, total serum IgG, IgA, IgM, and IgE concentrations were measured by the Department of Immunology and Immunotoxicology at the Slovak Medical University in Bratislava, Slovakia. Total IgG, IgA, and IgM were measured by rate nephelometry using a Beckman Coulter IMMAGE Immunochemistry System (Beckman Instruments, Inc., Fullerton, CA). Total serum IgE was analyzed by commercial sandwich enzyme-linked immunoassay kit (Biogema, Kosice, Slovakia). Extremely high and low values of IgE were reanalyzed using rate nephelometry on the same instrument as above. Concentrations of IgG, IgA, and IgM are reported in grams per liter (g/L) while results for IgE are given in international units per milliliter (IU/mL).

Data collection

At birth, mothers completed a questionnaire that included socio-demographic information, questions related to maternal health and medication use, family living environment, past pregnancies, and tobacco use. From the infant’s medical record, child’s birth weight, and gestational age were abstracted. At the child’s 6-month visit, mothers completed a questionnaire that included questions about the home environment of the child, the child’s growth, breastfeeding habits, and smoking habits of persons in the household.

Statistical analyses

Selection of potential confounding variables

Directed Acyclic Graphs (DAGs) were used to select potential confounding variables. Separate DAGs were created for perinatal PCB exposure (maternal and cord) and postnatal PCB exposure (6-month infant PCBs). DAGs for each immunoglobulin were not created as the causal structure of these models was regarded to be similar. Thus, two different sets of confounders were selected for our models: (i) ethnicity (Roma vs other), maternal smoking before or during pregnancy (yes/no), and maternal age at child’s birth (years) for maternal and cord PCB models; and, (ii) ethnicity (Roma vs other), maternal smoking at 6 months (yes/no), maternal age at child’s birth (years), infant sex, and infant age at the time of 6-month blood draw (days).

Multivariate methods

Separate linear regression models that included total serum IgG, IgA, IgM, or IgE concentrations as the dependent variable and measures of total and congener-specific maternal, cord, or 6-month infant PCB concentrations as the independent variable of interest, were fit. In all models, immunoglobulin concentrations were transformed by natural-log to ensure normality. PCB concentrations, which were expressed in ng/mL on a “wet weight” basis, were also natural-log transformed to reduce the potential influence of outlying values. Results for the regression models are expressed as the percent change in a particular immunoglobulin concentration for a change in PCB concentration across the interquartile range, i.e., the 25^th to the 75^th percentile. Regression analyses were carried out using the SURVEYREG procedure in SAS (version 9.2), with applied normalized weights to account for the sampling structure of the immunology subset. To reduce the potential for residual confounding by age, the regression analyses were restricted to the 350 (91%) infants who had blood collected within 1 month of their 6-month birthday.

Results

Study subjects

In this study, 74% of women were between 20 and 30 years-of-age at the time of their child’s birth, and 36% of women reported smoking before or during their pregnancy (Table 1). Approximately 20% of families were identified as being of Roma ethnicity. Three percent of women reported no breastfeeding, while 41% of women breastfed their child for the first 6 months of life. Comparing mother-infant pairs followed-up at 6-months-of-age to the original cohort recruited at birth, mother-infant pairs lost to follow-up at 6 months tended to be younger, of Roma ethnicity, and to have had at least one previous pregnancy beyond 20 weeks of gestation.

Table 1.  Characteristics of infants and mothers in cohort study at birth and at 6-months, and those in the subset selected for immunologic analysis at 6 months.

Characteristic (%)*	Full cohort at birth (n = 1134)	Full cohort at 6 months (n = 971)	Immunology subset (n = 384)
Infant sex
Male	51.9	51.3	51.0
Female	48.2	48.7	49.0
Birth weight (g)
<2500	4.8	5.5	4.4
2500–3499	56.9	55.5	56.6
≥ 3500	38.4	39.1	38.9
Gestation length (weeks)
< 37	3.8	3.7	3.3
37–41	93.4	93.5	93.7
≥ 42	2.8	2.9	3.0
Parity
0	41.5	45.1	41.3
1	32.7	30.5	33.2
2–4	25.5	24.2	25.3
Missing	0.3	0.3	0.2
Breastfeeding duration (months)
0	—	4.4	3.4
1–< 6	—	51.0	47.1
6	—	36.7	41.4
Missing	—	7.8	8.1
Ethnicity of child
Romani	20.8	18.0	19.5
Slovak/Eastern European	79.2	82.0	80.5
Maternal age at child’s birth (years)
18–< 20	8.4	6.8	8.6
20–30	75.0	74.2	73.7
> 30	16.7	19.0	17.7
Maternal smoking before or during pregnancy
Yes	36.2	38.0	35.5
No	63.8	62.0	64.5
Maternal smoking at 6-month follow-up
Yes	—	24.2	21.9
No	—	72.9	74.4
Missing	—	2.9	3.7
*Percents may not sum to 100 because of rounding.

PCB concentrations

As reported previously (Jusko et al. 2010), maternal PCB concentrations were determined for all of the 384 mother-infant pairs in our subset, and for 376 and 246 of the cord and 6-month infant specimens, respectively. On a per-lipid basis, the median maternal PCB-153 concentration was 140 ng/g lipid (25th and 75th percentiles: 108 and 228 ng/g lipid, respectively). The median sum of maternal PCBs 118, 138⁺¹⁶³, 153, 156⁺¹⁷¹, 170, and 180 was 455 ng/g lipid (25th and 75th percentiles: 335 and 720 ng/g lipid, respectively). The median sum of cord PCBs was 343 ng/g lipid (25th and 75th percentiles: 224 and 535 ng/g lipid, respectively), and for 6-month infant sum PCB concentration the median concentration was 361 ng/g lipid (25th and 75th percentiles: 94 and 787 ng/g lipid, respectively).

Immunoglobulin concentrations

Of the 384 mother-infant pairs selected for immunologic analysis, IgG, IgA, IgM, and IgE concentrations were completed for 384, 373, 384, and 369 6-month infant specimens, respectively. Weighted median values for IgG, IgA, and IgM were 4.5, 0.21, 0.65 g/L, respectively, and 5.2 IU/mL for IgE. IgG, IgA, and IgM concentrations were moderately associated (0.43 ≤ r ≤ 0.58; P < 0.0002), but weaker correlations between IgE and IgG, IgA, and IgM were observed (r = 0.19, 0.22, and 0.09, respectively).

Association between PCB and immunoglobulin concentrations

Table 2 presents the results for maternal, cord, and 6-month infant sum PCB concentrations in relation to total serum IgG, IgA, IgM, and IgE concentrations at 6-months-of-age (congener-specific associations were not observed; therefore, only the sum PCB results are presented). Overall, effect sizes were small and had wide confidence intervals. There was also no age-specific pattern of association, i.e., stronger effect estimates with maternal vs 6-month infant PCB concentration.

Table 2.  Adjusted percent change in total serum immunoglobulin concentrations associated with a change in the sum of maternal, cord, or infant PCB concentrations across the interquartile range.

PCB measure	IgG (g/L)			IgA (g/L)			IgM (g/L)			IgE (IU/mL)
	N	% Change	95% CI	N	% Change	95% CI	N	% Change	95% CI	N	% Change	95% CI
Maternal PCB*	350	1.2	−4.2–7.0	339	−2.0	−9.9–6.5	350	0.1	−6.1–6.6	337	7.4	−8.7–26.5
Cord PCB*	343	0.4	−6.0–7.3	332	−2.7	−12.3–8.0	343	−0.4	−7.6–7.3	331	7.4	−9.2–27.1
6-Month PCB^†	228	6.4	−1.5–14.9	221	−2.4	−14.4–11.3	228	−6.1	−13.0–1.3	224	2.3	−23.0–35.9

*Adjusted for Roma ethnicity (yes/no), maternal smoking before or during pregnancy (yes/no), and maternal age at the time of the child’s birth (years).

^†Adjusted for Roma ethnicity (yes/no), maternal smoking at child’s 6-month visit (yes/no), maternal age at the time of the child’s birth (years), infant sex, and age of the infant at 6-month blood draw (days).

IgA, IgE, IgG, and IgM, Immunoglobulin A, E, G, and M; PCB, polychlorinated biphenyl.

In a secondary analysis, models were also adjusted for breastfeeding duration (months), since breastfeeding is associated with higher immunoglobulin A and E concentrations in the infant (Kanariou et al., 1995; Grandjean et al., 2010) as well as increased PCB concentrations in the infant. However, no noteworthy associations after adjustment for this potential confounder were observed.

Discussion

In this population, with substantial PCB exposure, an association was not observed between maternal, cord, or infant serum PCB concentration and total serum concentration of IgG, IgA, IgM, or IgE determined at 6-months-of-age. These results are discussed below in terms of existing epidemiologic studies, and potential alternative hypotheses that could explain our null findings.

At least three previous cohort studies have examined the association between maternal/perinatal PCB exposures and immunoglobulin concentrations during infancy and childhood. In one study, data published from a different cohort of mother-infant pairs from the Slovak Republic showed increased cord IgE concentrations with increasing levels of placental PCB-118, but no association with any other dioxin-like or non-dioxin-like congener was observed (Reichrtova et al., 1999). While these data may indicate specificity for dioxin-like PCBs, the PCB-118-IgE association may have been attenuated (or non-existent) had adjustment for potential confounders occurred. In another study, mother-infant pairs from the Faroe Islands were followed from birth to 7 years-of-age when total serum IgE was assessed (Grandjean et al., 2010). Despite the considerable serum PCB concentrations in this population, no association was observed between maternal PCB exposures and total IgE in the child. A prospective cohort study by Dewailly et al. (2000) also measured IgG, IgA, and IgM concentrations in infants at 3-, 7-. and 12-months-of-age and found no association between these parameters and prenatal PCB exposure. Similarly, in the current study, no association was observed between PCB concentrations measured in maternal, cord, or infant serum and immunoglobulin concentrations at 6-months-of-age; furthermore, this study did not reveal any evidence of congener-specific associations that might indicate, for example, Ah receptor-mediated immunotoxicity (Hogaboam et al., 2008). These (essentially) null findings from cohort studies contrast with results from three cross-sectional analyses that observed some association between PCB and immunoglobulin concentrations. In one study of Flemish adolescents, Van Den Heuvel et al. (2002) observed inverse associations between total IgG and non-dioxin-like PCB congeners 138, 153, and 180. In a German study, a positive association between PCB concentrations and IgM levels in 7–8-year-old children was noted (Karmaus et al., 2005). In addition to assessing PCB exposure during pregnancy, the Faroe Islands cohort determined PCB concentrations in children at 5 and 7 years-of-age; unlike prenatal exposure, increasing PCB concentrations at 5 and 7 years (sum of non-dioxin-like PCB congeners 138, 153, and 180) were associated with higher total IgE at 7 years-of-age (Grandjean et al., 2010).

One candidate explanation for the divergent findings across studies, at least in terms of in utero exposure, may be attributable to the degree of PCB exposure in each population, though this possibility appears to be an unlikely explanation. Among those studies discussed above, prenatal PCB concentrations in the Faroe Islands cohort were greatest, followed by concentrations in the current and former Slovak populations, and the Northern Quebec study (Longnecker et al., 2003) - all of which showed essentially null results with regard to in utero exposure and total immunoglobulin concentrations. In addition, approximately 95% of mothers breastfed their child at least 1 month in our study, and 41% of mothers reported breastfeeding their child at 6-months-of-age, suggesting that peak PCB exposures were occurring during lactation in our sample (i.e., at 6-months-of-age). However, we observed no association between these 6-month infant PCB concentrations and immunoglobulin measures.

Another possible explanation of our null findings, and of the discrepant results between the cohorts and cross-sectional studies described above, is that circulating IgG and IgA concentrations measured during infancy are highly variable, especially compared to measures taken during toddlerhood and adulthood (Stiehm and Fudenberg, 1966). Variability in IgG concentrations at 6-months-of-age may be the result of (i) still present, but declining maternal IgG and (ii) rapidly increasing IgG production by the infant. Thus, even if PCB concentrations were a cause of altered immunoglobulin concentrations at 6-months-of-age, variability in total immunoglobulin measures may limit power and thus detection of associations with PCBs (or other environmental toxicants). Such a hypothesis is consistent with the results of the studies discussed here; those which assessed immunoglobulin concentrations later in childhood and into adolescence (van den Heuvel et al., 2002; Karmaus et al., 2005; Grandjean et al., 2010) observed associations with PCB concentrations while those studies which assessed immunoglobulin concentrations during infancy (Dewailly et al., 2000; and the current study), did not observe associations between PCB measures and immunoglobulin concentrations.

While larger variability in total immunoglobulin concentrations is a measurement issue which concerns statistical power, a slightly different but related issue in using total immunoglobulin concentrations as a measure of potential immunotoxicity concerns lack of specificity. For instance, a measure of “total” immunoglobulin concentration includes all antibodies in an extremely large (> 10⁸) antibody repertoire. Therefore, if maternal and early life exposure to a toxicant causes subtle perturbation in the functional capacity of the immune system, a measure of total serum immunoglobulins may “miss” detection of immunotoxicity.

The present study has some additional limitations. First, immunoglobulin concentrations beyond 6-months-of-age were not assessed. Doing so, and in the context of a longitudinal model, would have increased power to detect associations, and as noted above, immunoglobulin concentrations measured later in life would be less variable. Second, statistical power was reduced for analyses that included 6-month infant PCB concentrations as the exposure of interest, owing to a smaller number of samples analyzed for PCB concentrations at that time-point. Further, besides PCB-118, other dioxin-like congeners that may be associated with altered immunoglobulin concentrations were not measured, as they exhibit stronger Ah receptor activity.

Strengths of the current study include the ability to analyze both pre and postnatal measures of PCB exposure, which may be particularly relevant in this population where, 41% of mothers are breastfeeding at 6 months, and PCB concentrations in the infant are likely greatest at this age. Compared to some previous studies, the current investigation also collected extensive medical and demographic data longitudinally, which allowed adjustment for factors such as prenatal or postnatal exposure to maternal smoking and breastfeeding duration. Adjustment for these factors should have reduced bias in the estimates of association. Further, PCB exposures varied widely in this study compared to most contemporary populations (Longnecker et al., 2003), thereby increasing power to detect an association with immunoglobulin concentrations.

Immunoglobulin concentrations may be valuable markers of later morbidity, and as such, the goal of the current study was to determine whether in utero and early postnatal PCB concentrations are associated with altered immunoglobulin concentrations. The current findings suggest that in utero and early postnatal PCB exposures are not associated with immunoglobulin concentrations in infants at 6-months-of-age, despite the fact that maternal PCB concentrations in our study were substantial, and several-fold higher than contemporary concentrations in the United States. Our results need to be interpreted with caution, however, since total immunoglobulin concentrations during infancy are highly variable, and a true association may have been obscured by the “noise” inherent in this outcome. Future studies of environmental exposures may be more successful in detecting associations in children during later childhood when immunoglobulin measures are less variable, and nearer to adult concentrations.

Acknowledgements

The authors wish to thank Drs. Allen Silverstone and Troy Torgerson for helpful discussions about the developing immune system, and Jan Petrik, Zhiwei Yu, and Hye-Youn Park who was each instrumental in getting the early parts of this project organized.

Declaration of interest

This research received support from National Institutes of Health grants T32-ES007262, T32-RR023256, U01-ES016127, and R01-CA096525 and from the Intramural Research Program of the National Institutes of Health, National Institute of Environmental Health Sciences. Support was also provided by a dissertation award from the University of Washington Department of Epidemiology and a Fulbright Grant from the US State Department.