Introduction

Human serum is contaminated with numerous manufactured chemicals released into the environment during past decades (Arctic Monitoring and Assessment Programme 2003). Several xenobiotic compounds have weak agonistic or antagonistic actions on steroid receptors in invitro assays (Bonefeld-Jorgensen 2004; Sohoni and Sumpter 1998) and in animal models (Gray 1998). The estrogen receptor (ER), androgen receptor (AR), and aryl hydrocarbon receptor (AhR) are expressed throughout the male genital tract, and sex hormone signaling plays a pivotal role for development and regulation of male reproductive function (Hess 2003; Holdcraft and Braun 2004; Schultz et al. 2003). Effects on male reproductive organs or reproductive function in adult rats have been observed after exposure to weak xenohormones such as polychlorinated biphenyl (PCB) (Hsu et al. 2003), dichlorodiphenyltrichloroethane (DDT) (Ben Rhouma et al. 2001), and polychlorinated dibenzo-p-dioxins (Chahoud et al. 1992; Gray et al. 1995; Simanainen et al. 2004). Also in humans, exposure to chemical compounds with endocrine-disrupting properties has been suggested to be related to the apparent decline in semen quality (Swan et al. 2003). However, it remains unknown whether the low-level exposure to xenohormones that virtually all humans experience has implications for reproductive health (Safe 2000; Sharpe and Irvine 2004; Storgaard et al. 2006; Toft et al. 2004).

In large-scale epidemiologic studies addressing health risks related to hormonal active xenobiotics, it is practically and economically unfeasible to measure serum levels of more than a few compounds. It is of considerable interest that techniques have been developed to examine the xenobiotic activity of sex hormone receptors in serum fractions free of endogenous hormones (Fernandez et al. 2004; Hjelmborg et al. 2006; Rasmussen et al. 2003). Thus it has become possible to examine human health outcomes in relation to the integrated receptor activity of hundreds of xenobiotics that are found in human serum.

This article is to our knowledge the first report of semen quality in relation to xenohormone and AhR activity. We report cross-sectional relations of serum xenobiotic receptor activities and semen quantity and quality in four geographically different study groups that were selected to obtain high contrast in body burdens of PCBs and the main DDT metabolite dichlorodiphenyldichloro-ethylene (p,p´- DDE) (Toft et al. 2005a).

Materials and Methods

We addressed pregnant couples in 19 towns and settlements all over Greenland, in Warsaw (Poland), and in Kharkiv (Ukraine). Pregnant women and their partners were consecutively enrolled during antenatal visits. Moreover, Swedish fishermen were enrolled separately and independent of current pregnancy. Altogether 798 men provided a fresh semen sample; the percentage of semen providers relative to all men who were encouraged to deliver a semen sample was 79% in Greenland (201 men), 7% in Sweden (191 men), 29% in Warsaw (198 men), and 33% in Kharkiv (208 men). Blood samples were collected from the participating men within 1 week of the semen sample collection, except for a subgroup of 116 men from Greenland, who had their blood sample collected up to 1 year in advance. Venous blood samples were collected in 10-mL vacuum tubes; and after centrifugation the serum samples were stored at –80°C until analysis. The local ethics committee in each of the four participating countries approved the study, and each participant gave written informed consent before the study. Details of study design, selection of populations, and data collection have been published previously (Toft et al. 2005a).

Receptor-mediated chemical-activated luciferase gene expression (CALUX) assays of serum extracts were performed in a subset of men who provided semen samples. A total of 365 men were available with one, two, or all three receptor activity values; these included 262 subjects for the AR assay, 338 for the AhR assay, and 358 for the ER assay. Due to the limited amount of serum available, not all receptor assays were performed on each sample. Data on both xenobiotic activity and semen quality were available from 319 men. Characteristics of the study groups are provided in Table 1. The 319 men did not differ significantly from the remaining 479 men delivering a semen sample regarding sperm concentration, motility, morphology, age, or period of abstinence before collection of the semen sample (data not shown).

Measurements of xenobiotic receptor activity in serum. For estrogenic and androgenic activity determination, the serum fraction (F1) containing persistent organochlorine pollutants (POPs) and free of endogenous estrogens and androgens was obtained by solid phase extraction–high performance liquid chromatography (SPE-HPLC) extraction. SPE was carried out using Oasis HLB (hydrophilic-lipophilic balance) extraction cartridges (vol 6 mL; 500 mg HLB sorbent; Waters, Milford, MA, USA). Extracted compounds were collected using a VAC ELUT SPS 24 vacuum manifold (Varian, Harbor City, CA, USA). The HPLC system consisted of an Alliance 2695 separations module with a 300-µL injection loop, equipped with a 2996 Photodiode Array Detector and a Fraction Collector II (Waters). Separation was performed on a Spherisorb Si 60 analytical column 250 x 4.6 mm inner diameter, 5 µm particle size (Waters) as described by Hjelmborg et al. (2006).

Extraction of lipophilic POPs from serum to be tested for AhR activity was performed by ethanol and hexane followed by cleaning on a Florisil + sodium sulfate column (Ayotte et al. 2005), at Le Centre de Toxicologie, Sante Foy, Quebec, Canada.

Measurements of the xenobiotic-induced receptor activities are described in detail by Bonefeld-Jorgensen et al. (2006), Krüger et al. (2007), and Long et al. (2006). For the estrogenic, the androgenic, and the AhR activity assays, all samples were tested in triplicate in two sets of tests designed to test the basal response on the receptor assay and the response when a physiologic level of the respective ligand is present, respectively. The test of basal xenobiotic activity of the serum extract alone [termed XER (xenoestrogen receptor activity), XAR (xenoandrogen receptor activity), and AhR activity] was designed to test primarily for agonistic effects, but if the response on the assay was below the reference level (response of the solvent control) an antagonistic effect is indicated. On the other hand, the test for activity when the active ligands [17β-estradiol (E₂), methyltrienolone (R1881), or 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD)] were present in a concentration giving 40–50% of maximum induction (termed XERcomp, XARcomp, and AhRcomp) was designed to test primarily for antagonistic effects; but if response on the assay higher than reference values was observed, an additive or synergistic effect is indicated. Solvent controls and/or samples from a pooled serum sample of Danish men and women were run in parallel for each assay. Table 2 shows the coefficient of variations (CVs) between the 3–5 aliquots from the same serum samples, between solvent control samples with and without ligand added in a concentration of 40–50% maximum activity, and the interassay variability between pooled serum samples.

The estrogenic, androgenic, and AhR-mediated dioxin-like activities were determined using receptor mediated CALUX assays. For each assay, a cell line with the respective receptor and the luciferase reporter vector was employed as described in previous research (Bonefeld-Jorgensen et al. 2005, 2006; Krüger et al. 2007; Long et al. 2006). Briefly, the estrogen response was measured in the stable transfected MVLN human breast cancer cell line carrying the estrogen-response-element-luciferase reporter vector (kindly provided by M. Pons, INSERM, Montpellier, France) (Bonefeld-Jorgensen et al. 2006). The androgen receptor activity was determined in the Chinese Hamster Ovary cells (CHO-K1) transiently co-transfected with the MMTV-Luc reporter vector (kindly provided by R.M. Evans, Howard Hughes Medical Institute, San Diego, CA, USA) and the AR expression plasmid pSVAR0 (kindly provided by A.O. Brinkmann, Erasmus University, Rotterdam, the Netherlands) (Krüger et al. 2007). The AhR-activity was determined in stable transfected mouse hepatoma cell line Hepa1.12cR carrying the AhR-luciferase reporter vector provided by M.S. Denison (University of California, Davis, CA, USA) (Long et al. 2006). The luciferase activity was measured in a LUMIstar luminometer (Ramcon, Denmark) in 96 well plates. The luciferase activity was calculated as relative light units (RLU) per microgram cell protein (relative to the respective solvent control) and finally presented as RLU per milliliter serum. The reference values for the solvent control activity corresponding to the amount of serum extract added for the ER and AR assay was 3.14 RLU/mL serum, and 6.67 RLU/mL serum for the AhR assay. The estradiol equivalence (XER-EEQ) and TCDD toxic equivalence (AhR-TEQ) was calculated on samples that show values significantly higher than solvent control using a standard dose–response curve of 17β-estradiol and TCDD, respectively. For details, see Bonefeld-Jorgensen et al. (2006) and Long et al. (2006). The validity of these assays has been confirmed by testing the agonistic or antagonistic response of natural hormones and a series of chemicals with endocrine-disrupting activity (Bonefeld-Jorgensen et al. 2001, 2006; Fernandez et al. 2004; Hjelmborg et al. 2006; Krüger et al. 2007; Long et al. 2006; Pliskova et al. 2005; Rasmussen et al. 2003).

Collection and analysis of semen samples. Semen samples were collected by masturbation at the residence or in privacy in a room at the hospital. The subjects were asked to abstain from sexual activities for at least 2 days before collecting the sample and to note the actual abstinence time. If collected at home, the sample was kept close to the body to maintain a temperature close to 37°C when transporting it to the laboratory immediately after collection.

The samples were analyzed for motility and concentration according to the World Health Organization (WHO 1999) manual for basic semen analysis (WHO 1999). Briefly, for each sperm sample, the sperm concentration was determined on two aliquots of diluted semen samples (1:10 or 1:20) using an Improved Neubauer Hemacytometer (Paul Marienfeld, Bad Mergentheim, Germany). If the difference between the counts on the two aliquots exceeded 10% of the sum, two new assessments were made. Using a microscope mounted with a heated stage (37°C), the sperm cell motility was determined by counting the proportion of a) fast progressive sperm; b) slowly progressive sperm; c) local motile sperm; and d) immotile sperm on 100 sperm within each of two fresh drops of semen, placed on a preheated (37°C) clean glass slide, and covered with a cover slip. All semen samples were analyzed by one researcher in each country, and all semen analyzers had been trained in a series of three workshops held before and during the sample collection at the Fertility Centre, Malmö University Hospital, Sweden. This center is accredited by the European Academy of Andrology and participates in the Nordic Association of Andrology and the European Society of Human Reproduction and Embryology quality control program. The median interindividual CV was 8.1% for sperm concentration assessment and 11.1% for motility (grade A+B) assessment (Toft et al. 2005b).

The morphology of the sperm from all the countries in this project was determined centrally by two technicians at the Fertility Centre, Malmö University Hospital, on Papanicolaou-stained smears using the WHO (1999) criteria.

Statistical analyses. The effects of serum xenobiotic activities on semen quality are not easy to predict because of the highly interconnected hormonal regulation of spermatogenesis. Thus, a broad strategy for statistical analysis was employed. Visual inspection of scatterplots of the crude associations between exposure and outcome did not indicate any threshold effect for any of the associations. However, to allow for analysis of nonmonotonic response across the whole range of receptor activity, multiple linear regression analyses were performed in two receptor activity subgroups showing lower or higher receptor activity compared with the reference value of their respective solvent controls, representing agonistic and antagonistic responses, respectively. If the homogeneity test did not indicate different associations across the study populations, we computed common estimates across all four study populations. When significant associations or heterogeneity was indicated, the associations in the single populations were evaluated by Spearman's correlation analysis.

Before inclusion in the linear regression models sperm concentration, percentages of sperm cells with normal morphology, age, and abstinence time were transformed by the natural logarithmic function, which improved normality and homogeneity of variance, as indicated by inspection of Q–Q plots. The percentage of motile sperm cells was not transformed. We decided, apriori, that sperm concentration would be adjusted for duration of sexual abstinence before delivery of the semen sample and age. Analyses of motility were restricted to the 95% of samples for which the analysis was initiated within 60 min after collection. To decide which of the potential confounders (listed in Table 1) to adjust for in the multivariate models, we used the change in estimate method suggested by Greenland (1989). None of the β-coefficients were changed more than 10% when these potential confounders were included one by one in the models, except for high alcohol consumption (> 21 drinks/week). Because of the limited number of participants with high alcohol consumption and the lack of information on all the Swedish participants, we made alternative analyses excluding high alcohol consumers and achieved similar results. Therefore, the results are presented without adjustment or restriction regarding alcohol consumption. All statistical analyses were performed using SAS software (version 9.1.3; SAS Institute Inc., Cary, NC, USA).

Results

The test of homogeneity across populations indicated country differences in the association of sperm concentration to XARcomp activity, AhR activity, and AhR-TEQ; furthermore, the proportion of normal sperm seemed to differ among countries in the subgroup with XARcomp activity < 3.14 (Table 3). Geographic differences with significant associations between the continuous exposure markers and semen outcomes were only found for XARcomp activity and AhR activity, which showed negative and positive associations, respectively, to sperm concentration in Warsaw, but no significant associations in the other populations [see Supplemental Material, Table 1 (http: //www.ehponline.org/docs/2007/9352/suppl.pdf)]. In all other tested associations, no indication of heterogeneity across populations was found, so combined analysis could be performed. When the data from the four study groups were combined (Table 3), a statistical significant positive association of sperm concentration and XER activity was found both below [40% increase per unit increase in XER activity; 95% confidence interval (CI), 1 to 79%] and above the reference level (14%; 95% CI, 2 to 28%), but not across the whole range of activity (9%; 95% CI, –1 to 20%). The correlation analysis in the individual study groups indicated that the association below reference level was driven mainly by a positive association in Warsaw, and the association above reference level was driven mainly by a positive association in Sweden [see Supplemental Material, Table 1 (http: //www.ehponline.org/docs/2007/9352/suppl.pdf)]. A scatterplot of the association is given in Figure 1A. Furthermore, sperm motility was positively associated to XER activity in the subgroup below the reference level with an increase of 10% motile sperm per unit increase in XER activity. This association seemed also to be driven by a positive association in Warsaw. A scatterplot of sperm motility and XER activity is given in Figure 1B.

Figure 1. Sperm concentration (A) and sperm motility (B) in relation to XER activity in the four populations. Reference value = 3.14. *XAR activity of the F2.1 fraction significantly higher than for the F1 fraction. #Significantly lower than the R1881 EC50 solvent control.

Table 1.

Table 2.

Table 3.

In the groups stratified at the reference level, no statistical significant difference in sperm count, motility, or morphology was observed between the groups with low or high receptor activity.

Disscusion

The present study suggested a positive association of serum xenoestrogen activity and sperm concentration and motility across the four study groups. Furthermore, we found geographic differences in some of the tested associations, with the most marked effects in the population from Warsaw. Clearly, the statistically significant but rather weak associations must be confirmed in future studies before any strong conclusion about associations can be made.

The positive association of sperm concentration and XER activity found at XER activity < 3.14 could be caused by adverse effects of antiestrogenic compounds on sperm count, which is plausible because of the known essential role of estrogen receptor function in male reproduction (Hess 2003). However, the assay specifically designed to test for competitive effects did not confirm an association between effects on this assay and sperm count across study populations. The positive association of sperm concentration and XER activity > 3.14, indicating a stimulating effect of exogenous estrogenic compounds on sperm counts, is contrary to the expected, but could be hypothesized to be caused by an antiapoptotic role of estrogens on germ cells (Pentikainen et al. 2000) or a direct stimulatory effect of estrogens on spermatogenesis, because ERs are present in male germ cells (Lambard and Carreau 2005).

It is well known that there is cross-talk between ER, AR, and AhR (Morrow et al. 2004; Pascussi et al. 2004) that may lead to other responses on the receptors in vivo, compared with the exvivo tests used in the present study, where the response on the single receptors are tested. This further complicates the interpretation of the associations of the semen quality and receptor activities. However, by evaluating the combined responses on the different receptors we might get closer to net effects invivo.

For the group from Warsaw, a predominantly net estrogenic serum activity was observed where 21% of the samples had induced estrogenic effects, and antiestogenic effects were found only on 7% of the samples, compared with 1–14% agonistic activity and 19–71% antagonistic activity in the other study groups (Bonefeld-Jorgensen et al. 2006). This may explain why the associations of XER activity and sperm count and motility were found mainly in Warsaw. Whether the negative association between XARcomp and semen concentration in Warsaw reflects an increase of the androgenic activity affecting the sperm concentration negatively can only be hypothesized. Similarly, whether the positive correlation between the AhR activity (and AhR-TEQ) and semen concentration for the Warsaw study group can be explained by an increased metabolism of chemicals with adverse effects on semen concentration can at this step only be a theory.

Only three of the 50 associations (6%) tested by linear regression differed significantly from unity, which is close to the 5% positive findings, which were expected to occur under the null hypothesis of no difference (Table 3). Therefore, most if not all of the observed associations might be chance findings.

In the present study we included populations with both a large within- and betweenpopulation exposure contrast to POPs, which are known to interfere with ERs, ARs, and AhRs (Bonefeld-Jorgensen et al. 2001; Pliskova et al. 2005). The study groups from Greenland and Sweden represent highly exposed populations, whereas the other populations reflect the sort of exposure generally found in different parts of Europe. Except for the Swedish fishermen, the included populations were selected to reflect the general population in the regions. People with occupational exposure or people living in accidentally polluted areas may have higher levels of exposure to POPs or other compounds that may interfere with ER, AR, or AhR.

In the present study, the difference in net RLU activity in the assays was only 2–3 fold between samples with low activity (p5) and high activity (p95), indicating that the actual contrast in biologic response was limited. However, the exposure contrast between the samples with low level (p5) and high level of calculated ER-EEQ and AhR-TEQ was in the range of 5- to 10-fold, indicating substantial differences in the amount of chemicals present in the samples with low and high activity (Bonefeld-Jorgensen et al. 2006; Long et al. 2006). The limited contrast in biologic activity may be one of the reasons for the lack of consistent associations found in the present study.

Agonistic or antagonistic ER and AR activity was found in some individuals from each population, but with large variation between countries. For example, 71% of the samples from Greenland showed antagonistic XERcomp activity, whereas antagonistic XERcomp activity was found in only 7% of the samples from Warsaw (Bonefeld-Jorgensen et al. 2006). However, agonistic AhR activity was found in almost all of the serum samples from the included populations in the present study (97%; Long et al. 2006), and therefore we would especially expect to detect effects of this exposure marker across populations if it was associated with adverse effects on semen quality. In rats, administration of a single dose of TCCD in a concentration of 0.05 µg/kg during gestation reduced offspring sperm count by 25% (Gray et al. 1997), but adult exposure to TCDD required a dose of 3 µg/kg to have effect on male testis (Chahoud et al. 1992). Thus, it seems that male reproductive function may be affected by AhR-inducing agents, but it is likely that the most sensitive period to reproductive disturbances is the fetal period where small alterations of receptor activity may be of crucial importance for development of reproductive organs. In the present study we were not able to determine whether fetal exposure to compounds with effects on xenobiotic activity is affecting adult semen quality, but we found no indication that xenobiotic AhR agonistic activity was related to reduced sperm counts or impairment of other semen characteristics in adult men.

We expected the overall number of subjects included in the present study to be sufficient to detect associations between xenobiotic activity and semen quantity and quality, even though some misclassification on both exposure and outcome may appear. However, the statistical power to detect effects on semen quality within populations may be limited, particularly for sperm concentration, which is known to show considerable intra- and interindividual variation (Bonde et al. 1996).

If subfertile men with low or high receptor activities were selectively declining to participate, the findings would be biased. Although the participation rate was low in three of the four regions, selection bias is unlikely, because the men had no knowledge about either semen quality or xenobiotic serum receptor activity. Furthermore, the data did not indicate markedly different associations in the study group from Greenland, where the participation rate was high compared with that of the other study groups.

The ex vivo xenohormone assays have been validated thoroughly (Bonefeld-Jorgensen et al. 2006; Hjelmborg et al. 2006; Krüger et al. 2007) and because no consistent associations between xenohormone activity and natural estrogen or testosterone levels were found, it is unlikely that the xenohormone activities were influenced by contamination of endogenous steroid hormones (Bonefeld-Jorgensen et al. 2006; Hjelmborg et al. 2006; Krüger et al. 2007).

Negative findingsmay be biologically plausible because most xenobiotic hormonal actions of single compounds are weak in comparison with endogenous hormones (Safe 2000; Sharpe and Irvine 2004). In the present study, the median estimated estrogen equivalents (XER-EEQ) among subjects with agonistic estrogenic serum activity (13%) was on average 0.7 pg/mL serum, which is 4% of the median estradiol value (19 pg/mL) measured in the males in the present project (Giwercman et al. 2006). Only about 2% of the total estradiol in men is free and not bound to sex hormone–binding globulin (SHBG) or albumin (Van Pottelbergh et al. 2004), whereas xenohormones can bind to SHBG or albumin with much lower affinity, and therefore are bioavailable to a larger extent (Crain et al. 1998). A more direct measure of the response of the xenoestrogens can be seen as the further increase in the XERcomp assay, which has been determined to be 21% (Bonefeld-Jorgensen et al. 2006). The activity of xenoestrogens is thus considerable, but also the natural variation of estrogen level among men is high (Giwercman et al. 2006), so it is likely that the homeostatic processes of the body can compensate for these changes in estrogenic activity, and the production and maturation of sperms will not be affected. However, the xenobiotic serum estrogenicity was in the percentage range of physiologic levels in males and far higher than expected when considering the measured concentrations and the weak estrogenic potency of individual POPs—some five to six orders of magnitude lower than 17β-estradiol (Bolger et al. 1998). Because only the serum fraction including the POPs and not containing endogenous estrogens was used for the analysis, this may indicate that mixtures of several estrogenic compounds may cause actions much higher than simple summation of effects (Rajapakse et al. 2002).

In conclusion, in the present study we found that ex vivo estrogenic, androgenic, or dioxin-like activity in serum samples from the general population in Europe and among Inuits was not consistently and strongly associated with adult semen quality. Future analysis should investigate whether disturbances of fetal ER, AR, or AhR activity cause more severe reproductive effects.